Lidar system with radiating waveguide array or radiating waveguide face and planar lens for parallel detection

Abstract

The invention relates to a lidar system for detecting the surroundings, in order to transmit and / or receive light beams, comprising a waveguide array consisting of a plurality of radiating waveguides, preferably of the same type and arranged in parallel, each waveguide having a plurality of coupling points or radiating waveguide surfaces, the waveguide surface has a plurality of, in particular strip-shaped coupling structures, preferably coupling points or coupling structures for the radiation are located in an approximately equidistant grid, and the radiation direction is caused to oscillate in a first spatial direction by a change in the frequency and / or the effective refractive index of the wave in the radiation waveguide or in the radiation waveguide surface, the radiating waveguide array or the radiating waveguide surfaces are fed from the output end of the planar lens via the connecting waveguides or the connecting waveguide surfaces, and the radiating direction is rotated in a second spatial direction, preferably perpendicular to the first spatial direction, by means of a change in the frequency and / or the effective refractive index of the waves in the connecting waveguides or the connecting waveguide surfaces. The planar lens has a plurality of input ends which implement different radiation directions in a second spatial direction and which are coupled to the transmit-receive channels running in parallel, so that the radiation waveguide array or the radiation waveguide surface can be used for simultaneous detection of different radiation directions, preferably one of the two spatial directions is horizontal, and preferably one of the two spatial directions is horizontal. The other is vertical, preferably an arrangement comprising an array of radiating waveguides or radiating waveguide faces and connecting waveguides or connecting waveguide faces together with planar lenses is implemented on a photonic chip.

Description

Technical Field

[0001] This invention relates to a lidar system, particularly suitable for detecting the surrounding environment in motor vehicle applications. According to the invention, for transmitting and / or receiving signals, the lidar system has a radiating waveguide array or radiating waveguide surface and a planar lens, thereby enabling detection of different radiation directions. Furthermore, different configuration schemes for two-dimensional variations in radiation direction are proposed. Background Technology

[0002] More and more motor vehicles are equipped with driver assistance systems. These systems use sensor systems to detect the surrounding environment and, based on the identified traffic conditions, initiate automatic vehicle responses and / or provide instructions to the driver, especially issuing warnings. Here, system functions are categorized into comfort functions and safety functions.

[0003] These systems are now moving in a more advanced direction. They will not only assist the driver, but the driver's tasks will be increasingly performed autonomously by the vehicle, meaning that the driver's control will be increasingly replaced; this is known as autonomous driving.

[0004] Autonomous driving, in particular, requires sensors to provide highly accurate and machine-interpretable information about the surrounding environment. Radar systems, with their limited angular accuracy and separation capabilities, cannot currently meet these high detection requirements, either alone or in combination with camera systems. Therefore, lidar systems are used, offering high angular resolution (horizontal and vertical) similar to camera devices, while also providing distance information and separation capabilities per pixel. Currently, so-called time-of-flight lidar systems are primarily used, treating electromagnetic radiation as particles and thus only directly measuring distance, not relative velocity. Coherent lidar systems are increasingly gaining attention; these systems (like radar systems) treat electromagnetic radiation as waves, thus allowing direct measurement of relative velocity via the Doppler effect. Other advantages of coherent lidar systems include robustness to interference from other sources (e.g., by other lidar systems or sunlight) and higher sensitivity at greater distances, enabling longer operational ranges. Furthermore, coherent lidar systems are said to have greater potential for high semiconductor integration, potentially reducing manufacturing costs.

[0005] In coherent lidar systems, modulated electromagnetic waves are transmitted, meaning they must vary with time in at least one of their amplitude, frequency, or phase parameters; otherwise, distance measurement is impossible. The most commonly used modulation in coherent lidar systems is linear frequency modulation (FMCW = Frequency Modulated Continuous Wave), which typically consists of two frequency ramps with opposite slopes. However, this modulation suffers from ambiguity, especially with multiple reflections in the same radiation direction, and producing highly linear frequency changes is extremely complex. Phase modulation (e.g., pseudo-random variations of discrete phase values ​​at a fixed transmission frequency) avoids or minimizes these drawbacks, but digital analysis of the received signal is more complex, and existing schemes have some limitations, particularly in terms of sensitivity and range. To date, no known modulation scheme can achieve or allow continuous frequency variations within the data acquisition range of a pixel and across pixels, which would be advantageous for varying the radiation direction using waveguides. Variations in radiation direction are typically achieved mechanically in two spatial directions, which is cumbersome, results in excessively large structures, and has robustness limitations. In most cases, real-valued mixers are used (because they require significantly less work compared to complex-valued mixers); however, determining the sign of the received frequency is often difficult or impossible, thus requiring two assumptions about relative velocity and, if necessary, object distance. Current coherent lidar systems do not fully utilize the integration potential of semiconductors. Summary of the Invention

[0006] The task, solution and advantages of the present invention

[0007] The objective of this invention is to provide a simple method for achieving parallel detection of different radiation directions and two-dimensional radiation direction changes in a lidar system.

[0008] This task is essentially accomplished using the lidar system described in claim 1. Advantageous embodiments of the invention are claimed in the dependent claims. The core concept here is to achieve parallel detection of different radiation directions using a radiating waveguide array or radiating waveguide surface fed by a planar lens.

[0009] The advantages of this invention are particularly that it enables parallel detection of different radiation directions and two-dimensional radiation direction changes in a simple manner, and can be fully realized through semiconductor integration.

[0010] The lidar system for detecting the surrounding environment according to the present invention is characterized in that, firstly, for transmitting and / or receiving, it comprises an array of multiple or a large number of radiating waveguides, preferably of the same type and arranged in parallel, each waveguide having multiple coupling points, or having a particularly wide radiating waveguide surface having multiple, particularly strip-shaped coupling structures, wherein preferably, the coupling points or coupling structures for radiation are located in an approximately equidistant grid; secondly, the radiation direction oscillates in a first spatial direction by changing the frequency and / or effective refractive index of the wave in the radiating waveguide or the radiating waveguide surface; and thirdly, the radiating waveguide array or the radiating waveguide surface is connected by connecting waveguides or connecting waveguide surfaces. Fourth, the radiation direction is oscillated along a second spatial direction by varying the frequency and / or effective refractive index of the wave in these connecting waveguides or these connecting waveguide surfaces, the second spatial direction preferably being perpendicular to the first spatial direction. Fifth, the planar lens has multiple inputs that realize different radiation directions in the second spatial direction and are connected to parallel-operating transceiver channels, so that the radiating waveguide array or the radiating waveguide surface can be used to simultaneously detect different radiation directions. Sixth, preferably, one of the two spatial directions is horizontal and the other is vertical. Seventh, preferably, this arrangement is implemented on a photonic chip.

[0011] Furthermore, the lidar system can operate in a coherent manner, wherein the modulation, particularly phase modulation, is superimposed with continuous frequency variations, wherein the continuous, particularly linear frequency variations can be implemented within pixels and across pixels.

[0012] Advantageously, the different radiation directions realized by the multiple input ends of the plane lens are located in a coarse grid in the second spatial direction, while the radiation directions are scanned only over a small range by varying the frequency or effective refractive index in the connecting waveguide or in the connecting waveguide surface, thereby advantageously reducing the range required for frequency and / or effective refractive index variations.

[0013] Furthermore, liquid crystal material can be present in the radiation waveguide or the radiating waveguide surface, the adjacent region of the connecting waveguide and / or the feed waveguide, particularly in a planar manner above it. The optical properties of the liquid crystal material are affected by the applied voltage, thereby changing the effective refractive index of the radiation waveguide or the radiating waveguide surface, the connecting waveguide and / or the feed waveguide, and thus achieving a change in the radiation direction in the first and / or second spatial directions.

[0014] Advantageously, in the adjacent region of the connecting waveguide or the connecting waveguide surface, particularly above it, there are two complementary triangular regions with liquid crystal material, which are controlled by two complementary voltages consisting of a DC component and a variable component of opposite polarity.

[0015] In an advantageous design of the invention, a liquid crystal material for changing the effective refractive index is present on the radiating waveguide or radiating waveguide surface of the waveguide array, particularly in a planar manner above it, wherein different frequencies are used in the coarse grid to change the radiation direction along a first spatial direction, while the radiation direction is scanned only in a small area by the change in effective refractive index, so that the variability of the reduction in effective refractive index is sufficient.

[0016] Furthermore, the length of the connecting waveguide varies at least approximately linearly at the input end of the connected radiating waveguide or the radiating waveguide surface of the array, thereby achieving scanning in the second spatial direction by changing the frequency. For this purpose, the connecting waveguide preferably consists of one or more straight and parallel segments and one or more bends having exactly the same shape or at least the same bending angle.

[0017] Advantageously, amplitude and / or phase arrangements, i.e., so-called amplitude tapering and / or phase tapering, are provided on the radiating waveguide array or the radiating waveguide surface, preferably in two spatial directions, in order to minimize or avoid side lobes in the radiation as much as possible. The amplitude tapering is achieved by coupling of different strengths, while the phase tapering is preferably achieved by an imprecise equidistant arrangement of the coupling structure and / or an imprecise equidistant arrangement of the radiating waveguide, or by the input end and / or the output end of the planar lens in the radiating waveguide surface.

[0018] In a preferred embodiment of the present invention, a second waveguide of the same type is provided parallel to the radiating waveguide and parallel to the connecting waveguide ground. This second waveguide of the same type is not coupled to the output end of the lens, thereby avoiding or at least reducing the effects caused by inter-waveguide coupling.

[0019] Advantageously, the planar lens is achieved by a liquid crystal layer above a wide waveguide surface, which may be composed of multiple regions, wherein tolerances or frequency dependence can preferably be compensated by one or more control voltages.

[0020] In an advantageous design of the invention, the transceiver channels can be switched sequentially among a set of multiple lens inputs, thereby achieving different radiation directions in the second spatial direction.

[0021] In addition, different phase modulations are used for different transmit and receive channels, especially to achieve robustness to coupling in the circuit section before the lens input.

[0022] Preferably, the system comprises multiple radiating waveguide arrays or radiating waveguide surfaces, associated connecting waveguides or connecting waveguide surfaces, and a plane lens. The radiating waveguide arrays or radiating waveguide surfaces, associated connecting waveguides or connecting waveguide surfaces, and plane lenses are preferably connected to the same modulated laser source in series or parallel operation and have different structures, so that they achieve different radiation directions in at least one spatial direction at the same frequency, thereby advantageously reducing the range required for frequency and / or effective refractive index variation.

[0023] In an advantageous design of the invention, multiple radiating waveguide arrays or radiating waveguide surfaces and associated connecting waveguides or connecting waveguide surfaces and planar lenses are provided, and in order to perform radiation deflection, prisms or large, preferably common, prism-shaped sub-regions are respectively provided above the radiating waveguide arrays or radiating waveguide surfaces, particularly to achieve a large detection range and / or high resolution even in the edge regions.

[0024] Preferably, especially in lidar systems where orientation errors, inaccurate frequency knowledge, and / or inaccurate relationship between the effective refractive index and the corresponding control variables lead to deviations in the radiation direction, these deviations are determined based on the measured radial relative velocity of a stationary object so that they can be subsequently taken into account and / or corrected.

[0025] Advantageously, the road surface is used as a stationary object, and its angle in the vertical direction is preferably determined based on the measured distance and the sensor mounting height.

[0026] Advantageously, a device is provided for changing the direction of radiation, which is used to compensate for orientation errors and / or, in particular, to adaptively adjust the detection range according to traffic conditions.

[0027] In another advantageous design of the invention, particularly to ensure eye safety, the device for changing the radiation direction is monitored and / or the change in radiation direction is monitored, and the monitoring is achieved by examining the changes in the received signal in each corresponding spatial direction related to the object's reflection or the proportion of internal reflection and coupling and the reflection of the overlay layer. Attached Figure Description

[0028] Figure 1 A coherent lidar system employing binary phase modulation is shown.

[0029] Figure 2 The pseudo-random binary phase modulation process is shown.

[0030] Figure 3 The real-value portion of a low-frequency analog received signal is shown.

[0031] Figure 4 The two-dimensional correlation of the signals received by the two objects is shown.

[0032] Figure 5 The linear change in transmission frequency during phase modulation is shown.

[0033] Figure 6 The receiving frequency is shown as a time shift relative to the transmitting frequency in the absence of Doppler frequency shift (i.e., relative velocity is zero).

[0034] Figure 7 The two-dimensional correlation of the received signals of two objects is shown, where phase modulation is superimposed with linear frequency variation.

[0035] Figure 8 illustrates a cost-optimized fixed-route digital circuit for implementing two-dimensional correlation filtering; wherein, Figure 8a An overview diagram is shown. Figures 8b to 8d The three modules are shown in detail.

[0036] Figure 9 An implementation of a complex multiplier for exemplary rotation factor cost optimization is shown.

[0037] Figure 10 shows a fixed-route digital circuit for cost optimization of two-dimensional correlation filtering, which uses frequency shifting and decimation before the Fast Fourier Transform (FFT); where, Figure 10a An overview diagram is shown. Figures 10b to 10d The three modules are shown in detail.

[0038] Figure 11 The implementation of the rotation factor for frequency shift is shown.

[0039] Figure 12 The curve of the squared error of the transmission frequency relative to an ideal linear change is shown.

[0040] Figure 13 The relationship between the receiving frequency range and the distance to the object is shown, and the range of ambiguity caused by the unknown sign of the real-value mixer is marked with shade.

[0041] Figure 14 The relationship between the receiving frequency range and the target distance is shown when the modulation bandwidth is reversed.

[0042] Figure 15 The relationship between the receiving frequency range and the target distance is shown when the modulation bandwidth is reversed and doubled.

[0043] Figure 16 A waveguide with equidistant coupling points is shown, which is used for focusing and scanning by frequency variation in a first spatial direction.

[0044] Figure 17The linear curve of the radiation angle changing over time and the corresponding transmission frequency used to rotate the radiation angle are shown.

[0045] Figure 18 The variation of the radiation angle over time and the corresponding transmission frequency are shown at different scanning speeds.

[0046] Figure 19 The arrangement of the waveguide and lens is shown from two different angles for focusing in a second spatial direction.

[0047] Figure 20 It shows Figure 19 The arrangement also includes a prism-shaped device made of an electrically controlled dielectric constant material for scanning in a second spatial direction.

[0048] Figure 21 The arrangement of the waveguide and transparent one-dimensional liquid crystal array is shown from three different angles, which is used for focusing and scanning in the second spatial direction.

[0049] Figure 22 The arrangement of the waveguide and reflective one-dimensional liquid crystal array is shown from three different angles, which is used for focusing and scanning in the second spatial direction.

[0050] Figure 23 A two-dimensional liquid crystal array is shown.

[0051] Figure 24 An arrangement of 32 parallel and equidistant waveguides is shown.

[0052] Figure 25 A two-dimensional pixel field is shown for 32 identical and equidistant waveguides with small spacing.

[0053] Figure 26 A two-dimensional pixel field is shown for 32 identical and equidistant waveguides with a spacing 32 times that of the former.

[0054] Figure 27 A two-dimensional pixel field is shown for 32 identical waveguides with different spacings.

[0055] Figure 28 A two-dimensional pixel field of 32 small-spaced, equidistant waveguides is shown, and the field is continuously scanned stepwise in the second spatial direction.

[0056] Figure 29 A two-dimensional pixel field of 32 small-pitched, equidistant waveguides is shown, and is continuously scanned in the second spatial direction.

[0057] Figure 30 The radial component of the relative velocity of a stationary object with an elevation angle of 0° is derived.

[0058] Figure 31 The radial component of the relative velocity of a stationary object at any position and angle is derived, as well as the angle at which the road surface is seen in the vertical direction.

[0059] Figure 32 The irradiated area of ​​the road surface is shown.

[0060] Figure 33 The implementation of three equidistant phase values ​​is shown, with a switch between the three line segments of different lengths.

[0061] Figure 34 The implementation of four equidistant phase values ​​is shown, wherein there is a switchable inverter and a switch between two line segments of different lengths.

[0062] Figure 35 A modulation sequence according to the prior art is shown, which is synthesized from two sub-sequences.

[0063] Figure 36 A novel modulation sequence with two periodically nested subsequences is shown.

[0064] Figure 37 Using two objects as an example, two Fast Fourier Transforms (FFTs) of the first modulated subsequence are shown.

[0065] Figure 38 Using two objects as an example, two correlations of the second modulated subsequence are shown.

[0066] Figure 39 A new modulation sequence with two periodically nested subsequences is shown as an alternative.

[0067] Figure 40 A novel modulation sequence according to the present invention is shown, comprising two periodically nested subsequences and superimposed linear frequency modulation.

[0068] Figure 41 A radiating waveguide array is shown, which, according to the invention, is fed through a common, meandering, and therefore extremely long waveguide, thereby enabling two-dimensional scanning by frequency.

[0069] Figure 42 The first scanning mode for two-dimensional scanning by frequency is shown.

[0070] Figure 43 A second scanning mode for two-dimensional scanning by frequency is shown, wherein a negative angle in the first spatial direction leads to ambiguity in the second spatial direction.

[0071] Figure 44An arrangement with a meandering, extended, and curved feed waveguide is shown.

[0072] Figure 45 An arrangement with an extended and straight meandering feed waveguide is shown to achieve inherent geometric consistency.

[0073] Figure 46 An arrangement of meandering feed waveguides with an extended, straight, yet tilted configuration is shown to achieve inherent geometric consistency.

[0074] Figure 47 A meandering feed waveguide is shown, with each bend period coupled to multiple connecting waveguides and further coupled to multiple radiating array waveguides.

[0075] Figure 48 The overall scanning mode is shown, which is formed by three independent arrays with slightly different structures and feed waveguide lengths shortened to one-third.

[0076] Figure 49 and Figure 47 In contrast, a radiating wide waveguide surface is used instead of a radiating waveguide array, which gradually widens from its feed end.

[0077] Figure 50 A radiating waveguide with a voltage-controlled liquid crystal layer is shown, which covers the waveguide to affect its effective refractive index and thus its radiation angle; the left side shows a cross-section of the waveguide and its surrounding environment, and the right side shows a longitudinal section.

[0078] Figure 51 A voltage-controlled liquid crystal layer is shown covering a radiating waveguide array. This liquid crystal layer allows the radiation direction to be changed in a first spatial direction, while scanning is achieved in a second spatial direction via a meandering feed waveguide.

[0079] Figure 52 It was shown Figure 51 Possible scanning patterns for the arrangement.

[0080] Figure 53 A voltage-controlled liquid crystal layer is shown covering a pure direct-feed waveguide. This liquid crystal layer can change the radiation direction in a second spatial direction, while the radiating waveguide array can scan the frequency in a first spatial direction.

[0081] Figure 54 A voltage-controlled liquid crystal layer is shown that simultaneously covers the feed waveguide and the radiating waveguide array, through which the radiation direction in two spatial directions can be changed.

[0082] Figure 55A voltage-controlled triangular liquid crystal layer is shown covering the connecting waveguide, through which the radiation direction in the second spatial direction can be changed; in order to achieve scanning in the first spatial direction, another liquid crystal layer is provided above the radiating waveguide array.

[0083] Figure 56 Two complementary triangular liquid crystal layers are shown covering the connecting waveguide, which are controlled by complementary voltages; in addition, to avoid coupling effects, the connecting waveguide is stretched and separated and fed by a cascaded shunt network.

[0084] Figure 57 and Figure 56 In contrast, the cascaded shunt network fed in is replaced with a straight waveguide arranged at an angle.

[0085] Figure 58 An arrangement according to the invention is shown, employing a planar lens to achieve parallel transmission and reception, a radiating waveguide array for frequency scanning in a first spatial direction, and a connecting waveguide located between the two, above which are two complementary triangular voltage-controlled liquid crystal layers for scanning in a second spatial direction.

[0086] Figure 59 and Figure 58 In contrast, the connecting waveguide and the radiating array waveguide are replaced with wide waveguide surfaces that gradually widen from the feed end.

[0087] Figure 60 and Figure 58 In contrast, scanning is also performed using a voltage-controlled liquid crystal layer covering the waveguide array in the first spatial direction.

[0088] Figure 61 The diagram shows a voltage-controlled liquid crystal layer positioned only above the radiating waveguide array, while scanning in the second spatial direction is achieved via frequency using a connecting waveguide with a linearly varying length.

[0089] Figure 62 A method for parallel transmission and reception using signals containing multiple frequencies is shown, in which different radiation directions are simultaneously achieved in a first spatial direction by means of a radiating waveguide array; while scanning in two spatial directions is achieved by a voltage-controlled liquid crystal layer covering the radiating waveguide array and connecting the waveguides. Detailed Implementation

[0090] Figure 1A schematic diagram of a coherent lidar system 1.1 is shown. A coherent signal with a wavelength range of approximately λ = 1550 nanometers (nm) is generated using a laser source 1.2; the coherence length is at least several microseconds, and the frequency is constant according to existing technology. The signal then enters a switchable inverter 1.3, which can be used to change the sign of the signal, which is equivalent to a phase shift of 180°. The change of signal sign only occurs within a fixed grid, for example, 3.33 nanoseconds (ns). According to existing technology, the change of signal sign is pseudo-random, i.e., within T... m = After 3.33 nanoseconds, the frequency or probability of changing the sign of the signal is only 50%. Figure 2 This diagram shows a curve of the modulation sequence b(n), which consists of the values ​​+1 and -1, also known as binary; it is expressed in terms of N... m = 4096 cycles, meaning it repeats once every 13.6 microseconds (µs). The modulated signal passes through amplifier 1.4, circulator 1.5 (i.e., transceiver converter), is transmitted through transceiver unit 1.6, and is partially reflected back by an object 1.7; the delay depends on the object distance r and is therefore variable.

[0091]

[0092] Where, c = 3·10 8 meters per second (m / s) is the speed of light. The frequency shift depends on a radial relative velocity v, and is therefore variable, generated by the Doppler effect.

[0093]

[0094] The signal is then received by transceiver unit 1.6 and transmitted to the subsequent receiving path via circulator 1.5. In a complex-valued mixer 1.8 (also known as an IQ mixer), the modulated received signal is superimposed on the unmodulated laser signal and converted into a complex-valued low-frequency signal by means of photodiode unit 1.9; as shown in equation (1b), the frequency f of this signal is... e Corresponding to Doppler frequency shift f D :

[0095]

[0096] The modulation of this signal delays the signal propagation time relative to the transmitted signal. Figure 3 The low-frequency analog received signal e is shown in the case of an object. a The real value part of (t) is then sampled in the analog-to-digital converter unit 1.10 at a sampling frequency f. s =300MHz, that is, per T s Sampling and digitization are performed at 3.33 nanoseconds, resulting in the real-valued portion of the result. Figure 3The point is marked in the middle; due to the sampling time T here. s = 3.33 nanoseconds and modulation time T m = 3.33 nanoseconds is equal, therefore the number of sampled values ​​N per modulation period is equal. s Also related to the grid length N m Equal, i.e., N s = N m = 4096, therefore N = 4096 is used without indexing for either of them below. The complex-valued sampled signal e(n), also known as the received sequence, can be described as follows:

[0097]

[0098] Here, it is assumed that the propagation time t0 is the modulation time T. m = m0, which is an integer multiple of 3.33 nanoseconds, and therefore also the sampling time T. s = Multiples of 3.33 nanoseconds:

[0099]

[0100] Furthermore, corresponding to the Doppler frequency shift,

[0101]

[0102] Both are integers; 'a' is the complex amplitude of the received sequence, and 'exp' represents the exponential function. It is the imaginary unit.

[0103] According to equation (3), the complex-valued receiver sequence e(n) refers to a single object without vertical extension and an ideal receiver. In reality, there may be multiple objects and / or extended objects, and additional noise r will be generated in the receiver. e (n), especially thermal noise; then the resulting received sequence is

[0104]

[0105] Among them, “sum i=1、......、I " represents the sum function of I non-vertically extended single objects at indices i = 1, ..., I.

[0106] The discrete propagation time m of I objects can be determined from the received sequence e(n) for time periods n = 0, 1, ..., N-1. 0.i and discrete Doppler frequency shift k 0、iTo obtain the most accurate measurement results possible, that is, to achieve the best possible separation of signal and noise, thereby obtaining the maximum sensitivity and effective range of the lidar system, the so-called optimal filtering method is required. This involves filtering the received sequence e(n) and the two-dimensional space ê of the normalized received sequence with the possible ideal amplitude of a single object. m、k Correlation filtering between (n):

[0107]

[0108] Here, M-1 corresponds to the assumed or maximum object distance of interest, and the Doppler shift k is assumed to take all values. Thus, the two-dimensional correlation E m、k The conclusion is

[0109]

[0110] Here, "conj" represents the composition of conjugate complex values, while the modulation sequence b(n) remains unchanged in this process due to its real value. Correlation At object position (m, k) = (m 0、i k 0、i The value at this point is the peak value (often referred to as the power peak). Figure 4 It shows that in (m) 0、1 k 0、1 ) = (300, 3846) and (m 0、2 k 0、2 The two-dimensional relative magnitudes of the received amplitudes at (101, 1000) correspond to their distances r1 = 150 m and r2 = 50.5 m and radial relative velocities v1 = -14.2 m / s and v2 = 56.8 m / s (for the asymmetric range k = 0, ..., N-1 chosen above for k, negative Doppler frequencies are assigned for k > N / 2, and thus negative relative velocities are assigned; negative relative velocities indicate the relative moving away of the objects).

[0111] To determine the distance r of the object i and radial relative velocity v i It is necessary to determine the two-dimensional correlation. The peak value of the quantity, wherein only the peak value above the detection threshold is used in order to distinguish it from system noise. According to equations (3b) and (3c), r i and v i The location of the peak value, i.e., the discrete propagation time m, can be used to determine this. 0、i and discrete frequency shift k 0、i The calculation is as follows:

[0112]

[0113]

[0114] Therefore, the distance and relative velocity of multiple objects can be directly and explicitly determined from a modulation sequence. This has significant advantages compared to the linear frequency modulation with two frequency slopes of opposite signs commonly used in coherent lidar systems—in which ambiguity / uncertainty is unavoidable when there are multiple objects.

[0115] The calculation of this two-dimensional correlation and the downstream analysis and evaluation are performed in the digital signal processing unit 1.11. It represents a high computational workload of order N·M·N. However, the above equation (6) can also be regarded as a discrete Fourier transform by multiplying e(n)·b(nm), n =0, ..., N-1; it needs to be determined for each m = 0, ..., M-1; the discrete Fourier transform (DFT) is calculated by fast Fourier transform (FFT):

[0116]

[0117] Where k = 0, ..., N-1 are the output dimensions of the Fast Fourier Transform (FFT), i.e., discrete frequencies, reducing the computational workload to the order M·N·log2(N).

[0118] Phase modulation using superimposed linear frequency modulation

[0119] According to existing technology, it has been assumed that the frequency of the applied phase modulation sequence b(n) is constant, i.e., it does not change. The following will consider a method for continuously varying the frequency according to the present invention, wherein an ideal linear variation is first assumed; such as... Figure 5 As shown, the frequency of laser source 1.2 and the transmission frequency f TX (t) for a duration of T pm = N·T m = 13.7 microseconds modulation period at B = 800 MHz (i.e., at a slope B / T) pm = 58.6 MHz / microsecond) increases linearly. Therefore, the transmission frequency f is generally... TX Regarding (t):

[0120]

[0121] There is no Doppler shift (i.e., the initial assumption is that the relative velocity is zero), such as Figure 6 As shown, the frequency f of the received signal RX (t) Due to the corresponding delay in propagation time t0:

[0122]

[0123] Therefore, it is related to the transmission frequency f TX (t) is shifted downwards. Frequency shift f caused by propagation time. r for

[0124]

[0125] According to equation (1a), the propagation time t0 is:

[0126]

[0127] For the example of an object distance r = 150 meters, a propagation time t0 = 1 microsecond, and the above modulation values, the frequency shift caused by the propagation time is f. r = 58.6 MHz. According to equation (11b), the frequency shift f caused by the propagation time r It is proportional to the distance r between objects.

[0128] Total frequency shift and the frequency of the received signal after mixing f e The Doppler shift f according to equation (1b) D And the component f caused by propagation time according to equation (11b) r composition:

[0129]

[0130] This differs in particular from the case of a constant transmission frequency in the prior art initially considered. Since the phase modulation sequence with a propagation time offset still acts on the receiving frequency, equation (3a) continues to apply to the sampled and digitized receiving sequence e(n), where, for the discrete receiving frequency k0, equation (3c) now applies instead:

[0131]

[0132] And using the discrete propagation time m0 according to equation (3b) and :

[0133] k0 = N·T s ·(f D +f r ) = N·T s ·2·v / λ - m0·T s ·B, (13b)

[0134] Here, the discrete receiving frequency k0 is initially assumed to be an integer. Therefore, the optimized filtering method is to normalize the two-dimensional space of the received sequence e(n) with the possible ideal amplitude of a single object. The method of filtering based on the correlation between them is still effective, characterized by a two-dimensional correlation E. m、k Equations (6) and (8) are used here; here, equation (8) applies:

[0135]

[0136] That is, fast Fourier transform is performed by the corresponding products between the received sequence e(n) and the corresponding offset modulation sequences b(nm) to achieve cost-effective computation. Figure 7 The two-dimensional correlation E of the two object examples above is shown. m、k The values ​​of the quantities are given, with the distance between the two objects being r1 = 150 meters and r2 = 50.5 meters, and the radial relative velocities being v1 = -14.2 m / s and v2 = 56.8 m / s; the discrete time shift is m. 0、1 = 300 and m 0、2 = 101 remains unchanged, while the discrete frequency shift is due to the propagation time component -m according to equation (13b). 0、1 ·T s B = -800 and m 0、2 ·T s B = -269 becomes k 0、1 = 3846 - 800 = 3046 and k 0、2 = 1000 - 269 = 731.

[0137] To determine the peak position of the correlation (m) 0、i， k 0、i Determine the distance r between objects i Equation (7a) still applies:

[0138] r i = m 0、i ·c·T m / 2, ((7a))

[0139] In order to determine the radial relative velocity v i The discrete frequency shift k caused by propagation time must be considered. 0、i The components—which can be derived using equation (13b):

[0140] v i = λ / (2N·T s )·(k 0、i + m 0、i ·T s (B); (14)

[0141] Unlike the original equation (7b), it is necessary to start from the discrete frequency shift k 0、iSubtract the value due to propagation time -m 0、i ·T s ·B, this value is related to the discrete propagation time m 0、i Proportional.

[0142] The relevant system methods and the advantages of combining phase modulation and frequency variation will be explained in detail in subsequent sections.

[0143] Implementing the computational logic of two-dimensional correlation

[0144] First, we will discuss how to implement the two-dimensional correlation E in the digital signal processing unit 1.11. m、k The calculation. So far, we have considered the received sequence e(n) for time periods n = 0.1, ..., N-1, and the corresponding correlation E. m、k This refers to a single detection direction, that is, the horizontal and vertical directions associated with a single pixel. In fact, in each detection cycle initially assumed to last 100 milliseconds, approximately 160,000 detection directions, or pixels, are covered; this is typically achieved through a parallel transceiver, i.e., a combination of parallel pixel detection and scanning, i.e., sequential pixel detection. A parallel transceiver means... Figure 1All elements 1.4 to 1.10 of the lidar system 1.1 (except for the common focusing rotation device and radiation rotation device in the transceiver unit 1.6) exist multiple times, for example, 32 times. Scanning can be accomplished by, for example, continuous mechanical movement (e.g., a mirror motion) or electronic means, by continuous change or by sequential switching (the possibility of electronic scanning will be explained later). In continuous scanning, pixels may also partially overlap, i.e., the latter part of the N values ​​of the received sequence e(n) of a pixel is also used as the former value of the next pixel. If we now assume M = 500 (corresponding to the maximum distance of 249.5 meters in the above explanation), the Fast Fourier Transform (FFT) of length 4096 in equation (8) must be calculated 800 million times per second. Microprocessors or DSPs (Digital Signal Processors) cannot perform so many Fast Fourier Transform (FFT) calculations. Such processors typically operate at clock frequencies in the 1 GHz range, meaning that almost the entire 4096-bit FFT must be computed per clock frequency. However, modern processors typically perform a maximum of 100 multiplications and additions per clock frequency, even with parallel vector computing units, which is far fewer than the number of operations required for a 4096-bit FFT. Therefore, many FFT calculations can only be implemented using specialized computational logic in hardware. Since the FFT algorithm consists of many sub-units called butterfly operations, multiple butterfly operations with programmable multipliers are typically implemented for dedicated hardware FFT computational logic, as the twiddle factors required for multiplication vary with the butterfly operation sequence. However, implementing programmable multipliers is extremely complex.

[0145] Since the 800 million Fast Fourier Transforms (FFTs) per second roughly correspond to a 1 GHz clock frequency that such computational logic can achieve, programmable multipliers can be omitted by implementing each butterfly structure of the FFT and each adder and multiplier (for the corresponding twiddle factor) contained therein directly in the computational logic. As shown in module 8.4 of Figure 8, the 4096-fold FFT consists of log2(4096) = 12 sequential stages, each with 2048 butterfly structures. Each butterfly structure determines two output values ​​from two complex input values ​​through complex addition and subtraction and a complex multiplication (in Figure 8, an example of a butterfly structure in the first FFT stage is highlighted with a thick line). Because many sequential computations cannot be completed in a single clock cycle, intermediate storage must be inserted in registers; in Figure 8, such intermediate memories are inserted between the 12 FFT stages, represented by module z. -1In practice, the number of intermediate memories can be much greater, as the input feed of the first stage, for example, can be very long, potentially requiring additional intermediate memories (which receive the values ​​from their inputs each clock cycle). Therefore, the computation is performed in a so-called pipeline—the Fast Fourier Transform (FFT) calculation spans multiple clock cycles, with the computation circuit containing multiple FFT data points; the FFT input data is fed into the pipelined computation circuit each clock cycle, resulting in the FFT output appearing at the output of the computation circuit several clock cycles later, thus yielding a new FFT result each clock cycle.

[0146] The main work of implementing this arithmetic logic is done by the multiplier. Specifically, the product between the complex-valued signal and the twitch factor...

[0147]

[0148] This constructs a unit pointer (value = 1); typically, four real-valued multipliers are required. Each of these real-valued multipliers is usually implemented by adding multiple shifted values. However, a high precision factor is not needed here; for example, an error as high as 1 / 32 can be tolerated, even when using quantized values.

[0149]

[0150] Here, "round" indicates the rounding function. This rounding is used to adjust the correlation E. m、k The noise generated at the output is below the required dynamic range, and typically also below the receiver noise; moreover, the signal loss caused by this noise is negligible. Therefore, it is only necessary to implement multipliers with factors of ±1 / 16, ±2 / 16, ..., ±15 / 16. Taking a multiplier with a factor of 7 / 16 as an example; since...

[0151]

[0152] This can be achieved by subtracting the input value shifted four bits to the right from the input value shifted one bit to the right—assuming a binary representation is used here; the representation of 7 / 16 above is called the canonical symbol code (CSD). Except for the factors ±11 / 16 and ±13 / 16, all the above factors can be achieved by at most one addition or subtraction according to equation (16); to avoid using two additions or subtractions for these factors ±11 / 16 and ±13 / 16, they are approximated as ±10 / 16 and ±14 / 16, which still results in acceptable quantization noise.

[0153] Figure 9 The complex multiplier of the twitch factor is shown in the figure.

[0154]

[0155] From a 9-bit binary input value After performing multiplication, the output values ​​are of the same length. The example values ​​have already been input. For the real and imaginary parts of the output value, each is implemented using two real-value multipliers and one adder; then, each part adds the two results of the real and imaginary parts. Since the negative values ​​of the real and imaginary parts of the input value are also needed, inversion is required. Here, inversion is achieved by bitwise inversion, i.e., omitting the extra 1, the so-called least significant bit (LSB); the resulting error can be compensated by adding a correction value to the input value of the Fast Fourier Transform (FFT)—to this end, the effect of the missing 1 during inversion can be determined at the output of the FFT and converted to the input via the Inverse Discrete Fourier Transform (DFT). When right-shifting the binary value (used to implement the multipliers), the latter part is simply omitted, i.e., no rounding is performed; the resulting error is the average error, which can also be compensated by adding a correction value to the input value of the Fast Fourier Transform (FFT). When shifting right, the bit length of the value remains unchanged; therefore, in the double-compensated representation considered in this paper, it is only necessary to extend the highest bit of the input value accordingly, i.e., copy that bit. The right shift itself can be easily implemented through appropriate wiring, so no additional work is required. Since the value of the rotation factor to be multiplied is always 1, the input and output values ​​of the multiplier have the same numerical range; there is no need to extend additional bits upwards.

[0156] In a butterfly configuration, each of the two complex values ​​is added and subtracted; therefore, the resulting value can be twice the input value, necessitating an upward expansion of the numerical range by one bit. This increases the bit length of the 12 levels of the Fast Fourier Transform (FFT) by 12. However, through addition and subtraction, the noise component in the values ​​originating from receiver noise also increases, with an average increase of √2. Thus, the noise amplitude doubles after each of the two stages. Consequently, the least significant bit (LSB) can be omitted in each second stage (i.e., scaling by a factor of 0.5); the resulting quantization noise is less than the effect of receiver noise because the FFT input range is chosen such that receiver noise already has multiple LSBs there. The effect of simply omitting the LSB (i.e., not rounding), i.e., the resulting average error, can be compensated by re-adding the correction value to the FFT input value. In the circuit shown in Figure 8, this scaling is omitted in the last stage because there are no other computational steps inside the Fast Fourier Transform (FFT) that would benefit from the reduction in bit length; therefore, the bit length is increased from 8 bits / bit at the input to 15 bits / bit at the output by the Fast Fourier Transform (FFT).

[0157] As shown in Figure 8, implementing Fast Fourier Transform (FFT) in a structure with frequency decimation (division-in-frequency FFT) results in shorter bit lengths for the longest traces and preceding nontrivial multiplication operations (the latter two stages do not contain multiplication operations; the factor -ĵ only represents the corresponding wiring). Furthermore, this structure avoids reordering the input data in long traces; here, for further processing, there is no need to reorder the output data to its natural order. Therefore, implementing this structure requires less effort compared to the alternative, time-decimation FFT structure (time-division-in-Time FFT).

[0158] According to equation (8), to determine the correlation E m、k This requires applying a Fast Fourier Transform (FFT) to the product of the received sequence e(n) and the shifted modulation sequence b(nm). Due to the periodicity of the modulation sequence b(n) (its period is N), the product of the unshifted modulation sequence b(n) and the periodically shifted received sequence e(nm) can also be calculated. N The product of (n+m)), where "mod N " represents the modulus function of modulus N, and then a Fast Fourier Transform (FFT) is applied to it:

[0159]

[0160] These related values ​​differ in phase from those in equation (8), but are identical in magnitude, and only this magnitude is relevant to further evaluation; therefore, for simplicity, the same equation notation is used here (this relationship stems from the time shift of the Fourier transform). The periodically shifted received sequence e(mod) N The product of (n+m) and the modulation sequence b(n) is constructed in module 8.2 of Figure 8 and implemented using a switchable inverter. For a value of 1 in b(n), the input value remains unchanged; for a value of -1, it is simply inverted bit by bit—inverting actually requires adding a least significant bit (LSB), but the effect of omitting this addition is compensated by adding a correction value to the input value of the Fast Fourier Transform (FFT) in module 8.3. It is important to note that a switchable inverter is only needed when the modulation sequence can change—if the modulation sequence does not change, hard inversion can be implemented, but only during inversion. The cyclic shift of the received sequence is initially loaded into z by the received sequence value. -1 Register chain implementation.

[0161] Prior to this, in module 8.1, a correction value c1(n) was added to the received sequence to compensate for the effects of coupling and reflection in the lidar system's internal environment or its immediate vicinity; this will be discussed in more detail later.

[0162] As described above, to simplify quantization calculations, inversion is performed using pure truncation and pure bit-by-bit inversion; the resulting average error is compensated for by adding a correction value c2(n) in module 8.3 before the Fast Fourier Transform (FFT). This correction stage can also be implemented after the Fast Fourier Transform (FFT), rather than before.

[0163] After the Fast Fourier Transform (FFT), that is, after the correlation E is formed m、k The results will then be further processed. First, the magnitude of each of the N=4096 complex values ​​will be constructed in module 8.5. Since high precision is not required here, the complex values ​​can be processed... The value of |i| is approximated using the following method:

[0164]

[0165] Here, "max" and "min" represent the maximum value function and the minimum value function, respectively; this calculation can be performed with fewer logical operations.

[0166] The resulting N=4096 values ​​are used for both the summation operation in module 8.6 and the maximum value calculation in module 8.7. Both modules are cascaded; each of the 12 levels calculates the sum or maximum value of corresponding pairs of values. The registers required between levels are not represented.

[0167] To distinguish between the peak values ​​of correlations generated by objects and noise peaks, a summation calculation is needed to estimate the noise level. Since the peak values ​​of correlations generated by objects are very few, meaning that most values ​​represent only noise, dividing the summation result by 4096 (i.e., right-shifting by 12 bits) provides a good estimate of the noise level.

[0168] The maximum value, which in the frequency shift dimension (in addition to the Doppler component caused by relative velocity, there is also a component caused by propagation time), is determined as the magnitude of the Fast Fourier Transform (FFT) output value associated with N=4096 and the relevant index k for each corresponding time shift m (corresponding to distance). If this maximum value is at least 3 times higher than the estimated noise, it is considered to be generated by an object; according to the corresponding time shift m=m 0、i and the corresponding frequency shift index k=k 0、i The distance and relative velocity of each corresponding object i can be determined using equations (7a) and (14), and its reflectivity can be determined from the relevant level / level. As shown in module 8.7, if only the absolute maximum value is determined, only the object with the highest reflectivity in each corresponding pixel can be determined at a distance. If the very rare case of two objects with different relative velocities in a pixel at a distance (within about half a meter) is to be covered, the corresponding maximum values ​​of multiple value blocks can also be output due to the cascaded structure of the maximum value search, such as 8 modules of the same length. With the input data of the maximum value search set accordingly, multiple modules can also be used to interpolate the peak values ​​in the Fast Fourier Transform (FFT) to more accurately determine the frequency shift; because typically, the peak value appears in two adjacent Fast Fourier Transform (FFT) values ​​(as it is not at the integer index k0 as previously considered), two values ​​can be obtained by placing them in different modules of the maximum value search by appropriately arranging or distributing the input data.

[0169] It's also important to note that no window function is used for the Fast Fourier Transform (FFT) here; that is, the FFT input value is not multiplied by a bell curve. This approach is only necessary or meaningful when two objects with similar relative velocities but significantly different reflectivity appear at the same distance within a pixel and need to be separated. In particular, if no window function is used at the FFT input and the frequency shift index k0 is not an integer (i.e., the peak value is split between two adjacent FFT values), the sensitivity of the FFT output will decrease (i.e., the ability to detect objects with weak reflectivity and at a greater distance will be reduced). This effect can be mitigated by choosing an FFT length longer than its input signal, i.e., adding zeros to the input signal, a process known as zero-padding.

[0170] Given that the index is determined in the maximum value search, it's important to note that, due to the cascaded implementation, the index can be built bit-by-bit, starting from the least significant bit (LSB). In each comparison output of two values, in addition to the current maximum value, there is an index value whose bit length corresponds to the level number. The resulting index is related to the linear numbering at the maximum value search input; since the numbering at the Fast Fourier Transform (FFT) output is mixed with the frequency shift index k, a conversion / mapping is required later.

[0171] For the output dimension of the Fast Fourier Transform (FFT), i.e. the discrete frequency, the asymmetric region k = 0, ..., N-1 has been considered so far; however, the actual frequency shift k can be assumed to have two signs, typically limited by a previous low-pass filter, for example as part of an analog-to-digital converter, where, for simplicity, this limit is assumed to be k = -N / 2, ..., +N / 2. Thus, by subtracting N, the upper half of k = 0, ..., N-1 can be mapped to a negative value.

[0172] In the design scheme discussed here (sampling time T) s = 3.33 nanoseconds and a linear frequency modulation bandwidth B = 800MHz), the relative velocity range at the Fast Fourier Transform (FFT) output is approximately ±419 km / h when the target distance is zero (according to equation (14), where k 0、i = ±N / 2), and when the maximum object distance is 249.5 meters, the relative speed range is approximately -147...+690 km / h (according to equation (14), where k 0、i = ±N / 2 and m 0、i= M-1 = 499); therefore, the possible range of relative velocities, or the range of functional interest, is completely covered, even significantly beyond the coverage area except for negative velocities at long distances—how can uniform over-coverage of the relative velocity range, i.e., all distances, be achieved later, in order to shorten the Fast Fourier Transform (FFT) length by means of subsequent decimation. Generally, if the possible frequency shift range is smaller than the frequency range of the Fast Fourier Transform (FFT), i.e., the sampling and modulation times are shorter than the range considered for neighboring locations, decimation can be used before the Fast Fourier Transform (FFT); in the simplest case, this type of decimation is performed by calculating a partial sum of the product of the offset received sequence and the modulation sequence.

[0173] As mentioned above, the received signal undergoes low-pass filtering after the mixer—this can be done in a dedicated filter or as part of the analog-to-digital converter (especially when the converter is a triangular sigma converter). To obtain optimal sensitivity (i.e., optimal signal-to-noise ratio), an optimal filter related to the modulation scheme should be used: the impulse response of the low-pass filter also has a rectangular curve, provided the modulated signal on the transmitting side is rectangular and retains its shape in the received signal (i.e., after mixing). After low-pass filtering, the received signal element will obtain a triangular curve with double the length—but this curve can only be accurately obtained when the frequency shift is zero. For other frequency shifts, the larger the frequency shift, the less ideal the filtering effect (in terms of maximum sensitivity) becomes with this low-pass filter; by shortening the sampling time T... s Increasing the sampling frequency (i.e., increasing the sampling frequency) can reduce the relative range of frequency shift, thereby reducing the signal-to-noise ratio loss in low-pass filtering, where the modulation time T m Alternatively, a time higher than the sampling time can be selected to combine with decimation before the Fast Fourier Transform (FFT) without significantly increasing the required computational workload. Through low-pass filtering, the correlation E... m、k Typically, peak values ​​appear at two consecutive discrete distances m, allowing for interpolation to determine distances more accurately.

[0174] At the output of the fixed-wiring digital circuit shown in Figure 8, relevant information is received every clock cycle, approximately every nanosecond, indicating the presence of an object at each corresponding pixel and at each considered corresponding distance, and the relative velocity of the object. The received sequence e(n) of the pixel is loaded into the register once, and then kept constant through cyclic pushing over M = 500 clock cycles (corresponding to 500 different displacements m and different distances). For example, all 160,000 pixels will be evaluated sequentially within a detection cycle lasting 100 milliseconds.

[0175] The logic of the digital circuit shown in Figure 8 is primarily composed of addition—requiring approximately 300,000 adders with an average length of 12 bits. As the semiconductor technology of digital circuits continues to shrink in size, such large-scale fixed wiring becomes possible, both in terms of cost and power consumption. Compared to frequency modulation in existing technologies, phase modulation shifts more of the implementation cost of coherent lidar systems to the digital domain (because the analog portion becomes simpler, as discussed later, eliminating the need for high-precision linear and complex-valued receivers with frequency variations; and also, as discussed later, an easily achievable scan can be achieved by superimposing the frequency variations of phase modulation); the rapid advancements in semiconductor technology used in digital circuits lead to more optimized costs, making the solution more economical.

[0176] Now let us observe an alternative choice structure as shown in Figure 8. According to equation (6), the two-dimensional correlation E m、k This can also be viewed as a one-dimensional temporal correlation between the received sequence e(n) and the sequence b(n)·exp(-j2π·n / N·k), which is calculated for each k as follows:

[0177]

[0178] Among them, "CC" m "" indicates the cyclic correlation between two sequences of length N, where m = 0, ..., N-1 are the dimensions of the correlation output (the discrete propagation time m component in the twitch factor has been omitted above, as it only affects the phase of the result and has no effect on the amplitude value, which is the only concern here). Since N>M in the considered design, the cyclic correlation will handle more discrete distances m than required.

[0179] Cyclic correlation in the time range is equivalent to a product of discrete Fourier transforms in the frequency range:

[0180]

[0181] Among them, IFFT m Let m = 0, ..., N-1 represent the inverse fast Fourier transform (IFT), where m = 0, ..., N-1 are its output dimensions (it is assumed here that the inverse discrete Fourier transform (DFT) is implemented through a fast Fourier transform (FFT)). According to the frequency shift theorem of the Fourier transform, the factor applied to the modulation sequence b(n) in the time range... This represents a shift within the frequency range, i.e., the Fourier transform:

[0182]

[0183] Due to the frequency shift theorem of Fourier transform and the cyclic property of discrete Fourier transform, the correlation value can be further converted as follows:

[0184]

[0185] This relationship can be transformed into a structure similar to that shown in Figure 8. The input value of this structure is the Fast Fourier Transform (FFT) of the received sequence that needs to be calculated in advance. It is multiplied by the Fast Fourier Transform (FFT) of the pre-determined modulation sequence b(n) in the form of a cyclic offset k. Then, the Inverse Fast Fourier Transform (IFFT) is performed, which differs from the Fast Fourier Transform (FFT) only in the sign of the rotation factor; the output dimension of the Inverse Fast Fourier Transform (IFFT) is the distance dimension m (in the structure shown in Figure 8, it is the frequency shift dimension k), that is, for a discrete frequency offset k, the two-dimensional correlation E m、k Located at distance dimensions m = 0, ..., N-1. Then, perform a single-value calculation again, followed by summation and maximum value calculation along the distance dimension. Although multiple reflections at different distances may exist within a pixel, the probability that these reflections have the same frequency shift (consisting of relative velocity and distance values) is very low; therefore, determining the absolute maximum value here is sufficient. For example, consider the case of fog plus the possibility of stationary objects in a pixel: although all reflections have the same relative velocity, they have different frequency shifts due to different distances. This is achieved by using the index m = m 0、i and k=k 0、i The maximum value above the detection threshold can be used to determine the distance and relative velocity of each corresponding object i again with the help of equations (7a) and (14).

[0186] If the same modulation sequence b(n) is always used, then the product E(mod) is achieved. N A multiplier of (lk)·B(I) can be implemented with fixed wiring—through approximation and circuit-switched data (CSD) representation—requiring very little implementation effort. If the modulation sequence changes, a programmable multiplier may be required, which significantly increases the implementation effort.

[0187] As mentioned above, assuming N > M (N ≈ 8M), it's common to process more discrete distances m = 0, ..., N-1 than needed. This can be avoided by decimation before the Inverse Fast Fourier Transform (IFFT)—if the decimation factor is 8, N = 4096 values ​​will become only 512 values, which will be fed into the IFFT; in the simplest case, decimation is achieved by adding the corresponding 8 adjacent values. Thus, the dimensionality of the IFFT changes from the initial 4096 to 512, significantly reducing the implementation effort; its length of 512 still covers the entire distance range of length M = 500.

[0188] However, it should be noted that such a structure requires N = 4096 clock cycles per pixel (this is the amount of shift that the Fast Fourier Transform (FFT) of the input signal must calculate); this is 8 times more than the structure shown in Figure 8, but the structure shown in Figure 8 has reached its maximum achievable value at a clock frequency of approximately 1 GHz. Therefore, this alternative structure of equation (22) must be constructed multiple times, which greatly offsets the advantage of the shorter inverse Fast Fourier Transform (IFFT) length. This alternative structure is only meaningful when the product of the sampling length N (based on the modulation length design considered so far) and the number of pixels is small, since the structure only needs to be constructed a few times, preferably once.

[0189] Optimize computational logic using twiddle factor and decimation method.

[0190] As mentioned above, the total frequency shift of the received signal after mixing, i.e., frequency f e It consists of the Doppler shift and the propagation time correlation component generated by linear frequency modulation:

[0191]

[0192] For the discrete frequency k0 of the received sequence e(n), the following equation applies:

[0193]

[0194] The component caused by propagation time (in the above equation, the second component of distance r or discrete propagation time m0) causes the assessment of the relevant frequency range to depend on and increase with distance (this is relative to the frequency value with the largest amplitude, especially at larger distances). This means that the computational structure shown in Figure 8 actually requires a longer Fast Fourier Transform (FFT) compared to pure phase modulation, i.e., without additional linear frequency modulation, and therefore requires more computational effort. This problem can be solved by eliminating the frequency component caused by propagation time -m0·T before the Fast Fourier Transform (FFT). s ·B is to be solved; now let's briefly derive why we need to determine the two-dimensional correlation E according to equation (6). m、k The conversion is as follows:

[0195]

[0196] The output dimension of the Fast Fourier Transform (FFT) is now discrete frequency. Therefore, according to equations (13b) and (24), the peak position of the Fast Fourier Transform (FFT) is... It corresponds only to the relative velocities of the respective objects:

[0197]

[0198] The product of the received sequence e(n) and the offset modulation sequence b(nm) is multiplied by a rotation factor before the Fast Fourier Transform (FFT).

[0199]

[0200] Multiplication is performed. This multiplication is carried out in module 10.8 of the computational structure shown in Figure 10. That is, before the Fast Fourier Transform (FFT) begins, the correction value c2(n) with the average value error inside the Fast Fourier Transform (FFT) is truncated and reversed, and the addition operation implemented in module 10.3 is performed. The modules in Figure 10 that are basically the same as the computational structure shown in Figure 8 (mainly different only in dimensions) are numbered in the same way. mark.

[0201] While keeping the dimension N = 4096 of the Fast Fourier Transform (FFT) constant, for discrete frequencies In terms of coverage, it is k = -N / 2, ..., N / 2, the relative speed v ranges approximately ±419 km / h (source: ,in, k = ±N / 2), which is significantly higher than the relevant speed range, for example, v 最小 (v min ) = -80 km / h to v 最大 (v max = +280 km / h — For objects moving towards each other (positive sign of v), the functional correlation of relative velocity magnitude is higher than that for objects moving away (negative sign). Frequency k The corresponding asymmetric range can be shifted by the corresponding offset frequency.

[0202]

[0203] Frequency is

[0204]

[0205] The transformation is performed using a symmetric range; the frequency of this time correction and centering can be derived using equations (23) and (27):

[0206]

[0207] When using this frequency for Fast Fourier Transform (FFT) k In this case, the two-dimensional correlation E m、k Similar to deriving its representation from equation (24):

[0208]

[0209] Therefore, before the Fast Fourier Transform (FFT), the product sequence e(n)·b(nm) must be combined with the modified rotation factor.

[0210]

[0211] Multiply.

[0212] For v as an example 最小 = -80 km / h to v 最大 = Relative speed range of +280 km / h, at this frequency of the Fast Fourier Transform (FFT) output. k The symmetric range is k = -881, ..., +881. This range is less than half the length of the Fast Fourier Transform (FFT), N / 2 = 2048, so a decimation of 2 times can be performed before the FFT. In the simplest case, as shown in module 10.9 of Figure 10, this can be achieved by adding two consecutive values ​​out of a total of N values ​​(the required bit width increases by 1 because two values ​​need to be summed). Subsequently, the N / 2 summed values ​​constitute the input values ​​of the FFT, and module 10.3 is placed before them to add the correction value c2(n). Since the length is halved, N / 2 = 2048, the FFT is reduced by one stage, that is, only 11 stages (see FFT module 10.4 in the computational structure shown in Figure 10). The frequency of the FFT output. k Extends only to the range k =0, ..., N / 2-1. The subsequent modules 10.5-10.7, used to calculate quantities, sums, and maximum values, also have only half the dimension, N / 2 = 2048.

[0213] As mentioned above, the computational structure needs to cover approximately 160,000 pixels; based on current considerations, each pixel requires calculating an offset of M = 500 m between the received sequence and the modulation sequence, which requires M = 500 clock cycles. With the maximum achievable clock cycle for the computational logic being approximately 1 GHz, a single detection cycle of only about 100 milliseconds can be achieved (this also needs to consider costs such as data loading and system reconfiguration). However, in practice, a cycle time of 50 milliseconds is often required, which necessitates implementing twice the computational structure. To avoid this, the modulation time T can be... m Effective doubling, thus requiring only half the offset for each pixel to be calculated; also possible via T mWith the constant value of 3.33 nanoseconds, each corresponding modulated sequence b(n) is defined as having two consecutive values ​​equal to achieve this effective doubling:

[0214]

[0215] Therefore, the sampling time T s and modulation time T m They remain equal; therefore, the received sequence e(n) and the modulated sequence b(n) are still correlated with the same discrete time n. The only difference between the above considerations and the equations is that, for the offset m, now only one of every two values ​​needs to be considered, i.e., only even values ​​m = 0, 2, 4, ..., M-2 are considered. The two-dimensional correlation E in the calculation structure is then calculated. m、k The representation (where the offset is the received sequence, not the modulated sequence) is similar to Equation (17) using Equation (30) above, taking into account the decimation and the inconsistency of the time shift (because only for E) m,k When the value of the quantity is of interest:

[0216]

[0217] According to equation (33), the periodically shifted received sequence e(mod) N The product between (n+m) and the modulation sequence b(n) is described in module 10.2; according to module 10.1, the addition of the upstream correction value c1(n) remains unchanged from the initial structure in Figure 8.

[0218] Since each pixel in the structure shown in Figure 10 only requires M / 2 = 250 clock cycles, which is half of the initial structure shown in Figure 8, a half-cycle time of 50 milliseconds can be achieved. In addition, since the data dimension of the main module is halved, the workload is also only about half of the original.

[0219] For decimation, the simplest implementation to date is by adding two consecutive values. The resulting first-order low-pass filter (i.e., of length 2) has a large transition region and a gentle slope; this leads to two results: firstly, in the relevant frequency range... Significant changes in level / level (in) The variation between 2.15 dB is significant. On the other hand, due to interference from high-frequency noise (in... At a rate as high as 2.15 dB, sensitivity loss occurs after decimation. A higher-order low-pass filter allows for a clearer design of its transition region; for a third-order low-pass filter with four coefficients [0.5, 1, 1, -0.5], the level / level difference is only 0.91 dB, with a maximum sensitivity loss of 1.88 dB. This third-order low-pass filter requires three additions but still no multiplications, as the coefficient of 0.5 can be shifted one bit to the right, i.e., it can be achieved through pure wiring (the negative sign of the coefficient -0.5 can be simplified by bit-by-bit reversal). Using higher-order low-pass filters and coefficients not in the quadratic grid (i.e., those themselves require one or more additions) allows for a sharper edge. It should also be noted that frequencies above ±881 Hz can, in principle, be considered and defined. k (Therefore, the relative speeds outside the corresponding range of -80 km / h, ..., +280 km / h can also be calculated); however, level loss and sensitivity loss increase, and... The above may result in some ambiguity (because the frequency will be reflected in the range of ±1024).

[0220] Now, let's explain the twitch factor d described by equation (31). n、m This means that multiplication is performed using unit vectors with phases spanning the entire angular range from 0 to 2π. The phase of the rotation factor does not need to be arbitrarily precise; instead, it can be chosen from a finite set of phases that is closest to d. n、m The actual phase. The simplest method is to realize only four phases: 0, π / 2, π, and 3π / 2, i.e., the rotation factor is... This eliminates the need for multiplication; however, the resulting impact is significant, namely, the effective loss of sensitivity and the generation of noise or spurious peaks in the Fast Fourier Transform (FFT) output. Therefore, a method with eight phase values ​​of 0, π / 4, ..., 7π / 4 will be considered below, resulting in a rotation factor of...

[0221]

[0222] Therefore, in addition to the simple rotation factors ±1 and ±ĵ, there are also rotation factors. .factor It can be approximated by 1-2 -2 = 0.75 is achieved, which is effectively achieved through addition (except for reversing and right shifting by 2 bits, which can be easily implemented without much effort).

[0223] Due to the rotation factor d n、m This varies with the offset m, that is, it changes within the clock cycle of the computation logic; therefore, a programmable structure is needed to implement them, as exemplified in... Figure 11 The n shown in module 11.2 (there are a total of N = 4096 such structures). Through the rotation factor d n、m Multiplication will result in the real part i Re (n, m) and imaginary part Transform complex input values ​​into values ​​with real-valued parts. Re (n, m) and the imaginary part o Im The complex output value of (n, m) depends on one of the eight twitch factors.

[0224]

[0225] When used, logic 11.21 controls six toggle switches (switching between 0 and their corresponding input values) and four switchable bit inverters. In this logic, the required ten binary switch signals are generated by the corresponding discrete phase values ​​at the input terminals. That is, it is calculated from 3 digits. The corresponding values ​​p n、m That is, according to equation (34):

[0226]

[0227] In module 11.1, this is determined using an integrator (i.e., a first-order recursive structure); the linear part of the offset m (which gradually increases during the clock cycle of the computation structure) is implemented using the value in the recursive accumulator register R2, and the constant part (- k 偏移 This is achieved by initially loading the integrator register RI from register R1. The two values ​​in registers R1 and R2... The scaling factor is 2 compared to the values ​​in equation (36). 11 This means that calculations are actually performed with 11 decimal places, especially when integrating the value in the stepwise integration register R2, to achieve sufficient precision (this value is integrated M / 2 = 250 times, meaning the initial rounding error of the register value will be amplified by a factor of 250). It should be noted that since only even values ​​of m are used, further calculations are needed... (Additional factor 2). Since the integrator is implemented using 14-bit binary two's complement arithmetic, it ignores overflows exceeding the effective value of 8, thus essentially performing the modulo operation in equation (36). By using the above 3 bits, i.e., the 3 most significant bits (MSB), the range p is achieved. n、m= 0, ..., 7 quantization; in fact, this is not rounding according to equation (36), but rather truncation is implemented, which is different from rounding to an average of 0.5 - because the average error is calculated for all N rotation factors (n = 0, ..., N-1), and therefore only represents a constant phase shift, which is not important from a functional point of view (a constant phase shift of the Fast Fourier Transform (FFT) input data will only result in a constant phase shift of the Fast Fourier Transform (FFT) output data, so its magnitude, i.e. its only relevant quantity, will not change).

[0228] according to Figure 11 As shown, this is achieved with the rotation factor d. n、m The multiplication operation requires only a small amount of work. Alternatively, one could invest more effort by using a standard complex-valued multiplier (consisting of four real-valued multipliers) to calculate a 3-bit programmable factor (i.e., for values ​​-1, -0.75, ..., 0.5, 0.75); since adding 1 is not feasible, scaling the rotation factor, for example, with a factor of 0.875 might be more advantageous. Similar to... Figure 11 The factors used by the four real-valued multipliers can be derived from their respective numerical values ​​p using a logic method. n、m The value can also be longer than 3 bits (because multipliers with such complex values ​​can implement more than 8 different twitch factors).

[0229] Compensation of nonlinear frequency curves

[0230] To date, the transmission frequency f TX (t) is assumed to be an ideal linear curve:

[0231]

[0232] In reality, modulation is not perfect; for example... Figure 12 As shown, it typically has a one-squared error.

[0233]

[0234] And based on the relevant transmission frequencies, the following results were obtained:

[0235]

[0236] Without Doppler shift, the receiving frequency f RX (t) relative to the transmission frequency The propagation time t0 was offset; the frequency shift f caused by the propagation time between the transmitting and receiving frequencies. r (t), then we get:

[0237]

[0238] The last two terms describe the error.

[0239]

[0240] The frequency shift related to propagation time (which, according to equation (11a), is not included in the ideal linear frequency shift) is:

[0241]

[0242] After mixing and digitization, this error is directly transferred to the received sequence e(n), that is, the error k transferred to the discrete receiving frequency. 0、Q And similar to equation (13a), we get:

[0243]

[0244]

[0245] By receiving N values ​​of the sequence e(n), the first two terms are constant, indicating a constant frequency error, and are correlated in two dimensions E. m、k The peak position of the magnitude only moves in the frequency dimension k; conversely, the last term changes linearly with discrete time n, meaning the frequency changes in the received sequence, which will affect the two-dimensional correlation E. m、k This causes the peak value in the frequency dimension k to broaden—the width is approximately Q·2·N. 2 ·T s 2 ·T m • m0. The displacement and broadening of the peak value increase with increasing discrete propagation time m0 (i.e., with increasing object distance). Although the effect of displacement can be considered when the frequency k0 is simply converted into relative velocity, the broadening leads to reduced sensitivity (reduced peak value height), reduced accuracy in determining the peak value position and relative velocity, and poorer separation of two objects at the same distance with similar relative velocities.

[0246] The following explains how to avoid broadening of the peak value. Based on the frequency error f in equation (39)... r、Q (t), the phase error can be obtained as follows

[0247]

[0248] The integration constant is omitted in this derivation because the constant phase component is independent of the function. Let the time be discrete-time n, where t = n·T s , t0 = m0·T m And T pm = N·T m Therefore, we can conclude that:

[0249]

[0250] For objects with discrete running time m0, the two-dimensional correlation E can be calculated. m、k At this time, the product sequence e(n)·b(nm) is multiplied by the rotation factor, and the negative value of the aforementioned phase error is added to each offset m to correct the phase error.

[0251]

[0252] (Note that in the method shown in Figure 10, m can only use a subset of even values). These twist factors can be implemented according to equation (26) or (31) with the aforementioned twist factor d. n、m The implementation of this is combined to compensate for the frequency shift that depends on the propagation time and, if necessary, the asymmetric Doppler frequency range—then a rotation factor with the sum of the phases of two separate rotation factors must be implemented. According to equation (43), d n、m、Q The first two phase components are linear in offset m, therefore, in implementing something similar to... Figure 11 When dealing with twiddle factors, they can be implemented as an additional part of register R2, which is the input to the integrator:

[0253]

[0254] The additional part consists of two subsequent parts, which are proportional to the magnitude Q of the second-order frequency error. The subsequent part of equation (43), which is proportional to the square of the offset m, is generally negligible, as it is significantly smaller than the other parts since M << N; however, to implement it, a second-order integrator is required—therefore, a third register is also needed, which is integrated into an upstream integrator whose output is used as another input. Figure 11 In the integrator shown.

[0255] Frequency error correction can, in principle, be performed after the Fast Fourier Transform (FFT), rather than before; it should be performed before the FFT along with the rotation factor d. n、m、Q Multiplying, n=0, ..., N-1, is equivalent to convolving the spectra of these twitch factor sequences (i.e., Discrete Fourier Transform (DFT) or Fast Fourier Transform (FFT)). In the case of the considered quadratic frequency error, this spectrum has a wide magnitude peak near zero (the magnitude peak width increases with offset m); therefore, the convolution can be restricted to a summation of a few values, weighted with complex-valued factors—though this implementation is far more complex than the correction methods previously described with the Fast Fourier Transform (FFT) above, also because these weighting factors depend on the offset m.

[0256] So far, we have considered the quadratic error of the transmission frequency as described in equation (37). The above considerations can also be applied to the general error f of the transmission frequency. TX、Q (t), which is the general linear curve deviation. To correct the transmission frequency error, equation (41a) and t = n·T are used. s and t0 = m·T m In this case, the general form of the rotation factor multiplied by the product sequence e(n)·b(nm) is:

[0257]

[0258] These twist factors can be recombined with the twist factor d in equation (31). n、m This is combined to compensate for frequency shifts dependent on propagation time and asymmetric Doppler frequency ranges. For similar... Figure 11 The implementation of discrete phase value p n、m The value range of is 0, 1, ..., 7 (i.e., the length is 3 bits), which is an extension of equation (36), and it involves an ideal linear frequency modulation:

[0259]

[0260] Used to generate these discrete phase values Figure 11 The module 11.1 shown can be replaced by registers to which the respective pre-calculated phase values ​​are written. Besides using registers solely for storing 3-bit phase values, a long register can be used to write all m discrete phase values, i.e., M / 2 = 250 3-bit values ​​in the example considered above, where m clock cycles are run on this register, i.e., cyclic pushing. The advantage of this is that the writing process does not need to be performed when calculating pixels; when multiple pixels have the same transmission frequency error, only one write is needed (followed by cyclic pushing). This method of pre-calculating register values ​​can also be used with… Figure 11 The methods shown are used in combination, that is, using an integrator circuit, so that these registers only need to store the parts that cannot be implemented with an integrator circuit. In some cases, this may result in a single register being used to generate multiple phase values ​​p. n、m That is, for different n and / or m, thereby reducing the number of registers required and the amount of pre-computation work (the same applies to microcontrollers).

[0261] To compensate for a nonlinear frequency curve, the error, i.e., the deviation from a linear curve, must be understood. As mentioned above, uncompensated frequency curve errors particularly cause broadening of the magnitude peaks in the frequency dimension k; for the case of a quadratic error, there is a fixed relationship between the width of the magnitude peaks and the error magnitude Q (see above), thus Q can be calculated. A nonlinear frequency curve between pixels can also cause angular errors (discussed in detail later, including how such angular errors can be determined); if these angular errors are known, the errors in the frequency curve can be inferred from them—not only across pixels but also within pixels, because the error curve is usually quadratic, at least in some regions.

[0262] Compensation for internal coupling and overlay reflection

[0263] As mentioned above, in the computational logic modules 8.1 and 10.1 with fixed wiring as shown in Figures 8 and 10, the received sequence is further supplemented with a correction value c1(n) to compensate for the effects of coupling and reflection within the lidar system or its immediate surroundings, particularly in a coverage layer. It is worth noting that radar sensors and coherent lidar sensors are based on the same coherent operating principle, but differ in their electromagnetic frequency ranges, thus offering advantages in compensating for such effects. We will now further explain how to determine these correction values ​​through phase modulation in a coherent lidar system, and how to easily implement these correction values.

[0264] Assume that, on the one hand, the discrete distance m = 0 is exactly at the position where the distance is zero; on the other hand, the distances of these couplings and reflections are also negligible (therefore, the receiving frequency f...). e = 0; Doppler components are also zero), in this case, the level P generated by these couplings and reflections in the received sequence can be determined by the average of the product of the received sequence e(n) and the unoffset modulation sequence b(n). (The signal portion of the received sequence originating from the actual object has a lower level on the one hand, and on the other hand, it corresponds to the offset modulation sequence and usually has a receiving frequency f.) e (≠ 0, therefore largely averaged). Thus, the correction value c1(n) is obtained by multiplying the level P, thus determined, by -1, and by multiplying the unoffset modulation sequence b(n) by a value ±1. The determination of this average value P (by summing N = 4096 values ​​e(n)·b(n) and shifting the result 12 bits to the right) can be performed either outside or within the fixed-wire computational logic. If implemented within the computational logic, an additional module can be implemented, or the two-dimensional correlation E at discrete distance m = 0 and discrete frequency k = 0 can be used directly in the existing computational logic. m、kThe value corresponds to the sum of e(n)·b(n). If the correlation result is used, the calculation logic is implemented in a pipeline, so the calculation lasts for several clock cycles; if the level value / level value calculated by two-dimensional correlation is to be used in the same pixel, the offset of the input sequence in module 8.2 or 10.2 must be paused, i.e., the clock cycle of that module must be paused. Alternatively, the level value of the previous cycle or the adjacent pixel that has been processed before can be used (as long as the value P between pixels does not change significantly). For such a determined level value P, the fixed-route calculation logic assigns a correction value c1(n) to the value ±P; if the same modulation sequence b(n) is always used, fixed-route for each corresponding positive and negative sign can be implemented (the negative sign is preferably implemented only by bit-by-bit inversion); if b(n) changes, a switchable inverter is required.

[0265] Due to the losses in the received pulse waveform described above, especially the reflections from the overlay layer, the propagation time may be non-negligible. Therefore, even at discrete distances m = 1, signal components may still be visible if necessary. The level in the received sequence e(n) depends not only on the respective values ​​of b(n) but also on the previous value b(n-1). Therefore, two average values ​​must be determined: one is the average of the time values ​​n when b(n) and b(n-1) have the same sign, and the other is the average when b(n) and b(n-1) have different signs (this can be achieved using configurable inverters and switches if the modulation sequence b(n) may change). Then, in the fixed-wiring logic, the correct sign is assigned to the correction value c1(n) based on whether b(n) and b(n-1) have the same or different signs at their respective n values. In principle, these two level values ​​can also be determined by the sums of e(n)·b(n) and e(n)·b(n-1) for all n, where these sums can be determined by an explicit calculation or obtained from two-dimensional correlation. If the received pulse is severely attenuated and / or offset such that the value of the received sequence e(n) is affected by the values ​​of three adjacent b(n), the above method should be extended accordingly.

[0266] Through the linear frequency variation superimposed by phase modulation, the received sequence, especially the received sequence reflected by the overlay, may have a small frequency; even if only a small portion of a single period passes through the N = 4096 values ​​of a pixel, the compensation mechanism assuming zero received frequency and constant phase will no longer be entirely effective. To reduce the frequency effect, the level value can be determined segment by segment, for example, into four segments of length N / 4 = 1024. As an alternative, if the received frequency is known to be small, it can be eliminated by multiplying by a corresponding rotation factor before determining the level value, and then these rotation factors can be applied again when determining the correction value c1(n); however, this is quite cumbersome. The level value can also be redetermined from two-dimensional correlation; here, an interpolation can also be used to determine the unknown frequency.

[0267] If the effects of coupling and reflection remain at least approximately constant over time and / or within a small pixel region, they can be averaged by the detection period and / or pixel region. However, typically the effects are not constant across all pixels; for example, the phase position of the overlay reflection will differ due to variations in radiation direction.

[0268] The sign of the received frequency in a real-value mixer is determined by using linear frequency modulation.

[0269] The following sections will explain the advantages and systematic approach of combining phase modulation and frequency variation.

[0270] exist Figure 1 In the lidar systems considered so far, the mixer is configured as a complex-valued mixer, which significantly increases the workload of the receiver path (almost doubling) compared to a real-valued mixer (which has only one real-valued output). With a real-valued mixer, only the magnitude of the received frequency can be determined, not its sign, because of the correlation E... m、k There are two peak values ​​(m0, +k0) and (m0, -k0). In the prior art, if the transmission frequency is constant, only the magnitude of the radial relative velocity can be determined, but its sign cannot be determined. According to the prior art, the sign can be determined by tracking, i.e., tracking over multiple acquisition cycles, and / or by rationality verification, such as whether the measured magnitude of the radial relative velocity corresponds to a stationary object, but this also has some drawbacks. These drawbacks can be reduced by superimposed linear frequency modulation of phase modulation. According to equation (12):

[0271]

[0272] After mixing and digitization, the receiving frequency f e The Doppler frequency shift f generated by the relative velocity v D and the frequency shift f caused by linear frequency modulation r Composition, where f rProportional to the object distance *r* determined by phase modulation. Related to the range of relative velocities. The corresponding receiving frequency range f e Therefore, it shifts with the distance r from the object; Figure 13 What is shown is v 最小 = -80 km / h and v 最大 = +280 km / h, and the linear frequency modulation design scheme considered above, i.e., modulation bandwidth B = 800 MHz at T pm = 13.7 microseconds. Taking a target distance of r = 200 meters as an example; the relevant frequency range extends to f e = -106.8, ..., 22.3 MHz. Since only the magnitude of the received frequency can be determined in a real-value mixer, but not its sign, therefore in f... e Within the range of -22.3, ..., 22.3 MHz, this frequency corresponds to the relative speed range v = 156, ..., 280 km / h, thus leading to ambiguity—for example, it may be impossible to distinguish between relative speeds v = 156 km / h and v = 280 km / h, or v = 186 km / h and v = 250 km / h; for all measured values ​​|f e The receiving frequency of 22.3 MHz can only use a negative sign, so the relative speed can be clearly determined. Figure 13 In the diagram, the range of ambiguity in the received frequency is indicated by shading. For a maximum object distance of 249.5 meters, almost the entire relevant frequency range f e = -126.2, ..., 2.9 MHz are all negative values, so there will be no ambiguity / uncertainty except for the relative speed range v = 265, ..., 280 km / h.

[0273] To reduce ambiguity in the receiving frequency range, a negative modulation bandwidth can be selected, which is the same value as B = -800 MHz mentioned above. This means that the transmission frequency decreases linearly over time. Figure 14 This indicates the corresponding relationship for the received frequency; for object distances r > 73.4 meters, there is no longer any ambiguity. However, the magnitude of the maximum received frequency is now higher than the positive modulation bandwidth (198.1 MHz instead of 126.2 MHz), thus requiring a higher sampling frequency, at least approximately f. s = 450 MHz (instead of f) s = 300 MHz). Increasing the modulation bandwidth B can further reduce the ambiguous distance range; for twice the modulation bandwidth and a negative sign, i.e., B = -1600 MHz, according to Figure 15As shown, ambiguity only occurs when the object distance r < 36.7 meters; however, in this case, the sampling frequency needs to be further increased. It is also worth mentioning that, as explained above, if the frequency shift caused by propagation time before the Fast Fourier Transform (FFT) is eliminated by multiplying by the appropriate rotation factor, and then decimation is performed accordingly, then a higher sampling frequency (and more sampled values ​​per pixel, with the data acquisition time remaining constant) can be used to calculate the Fast Fourier Transform (FFT) length of the two-dimensional correlation while keeping it constant.

[0274] By utilizing the frequency shift effect of linear frequency modulation, which depends on propagation time, the ambiguity problem in relative velocity determination when determining the distance to nearby objects can be reduced by real-valued mixers. Since a trajectory is usually set during tracking, the ambiguity problem can be resolved by assigning the detection results generated in a single detection cycle to the trajectory.

[0275] If the lowest possible sampling frequency is to be used, only a small modulation bandwidth can be used, which may lead to ambiguity in the determination of relative velocity over the entire distance range. This can be resolved by changing the modulation bandwidth during the detection period, especially by alternating its sign. For the receiving frequency, with a modulation bandwidth of +B, the following can be derived:

[0276]

[0277] In the case of inversion modulation bandwidth -B:

[0278]

[0279] Among them, the frequency shift depends on the propagation time.

[0280]

[0281] Here, the value is for +B and is generally known because the object distance r is determined by phase modulation. The Doppler frequency shift f is obtained by adding these two equations. D for

[0282]

[0283] However, when using a real-value mixer, only the magnitude / absolute value |f of two received frequencies can be measured. e、1 | and |f e、2 The frequency shift f caused by the forward propagation time will now be addressed. r (i.e., B < 0) This leads to the derivation of how to definitively determine the relative velocity v. If the Doppler frequency shift f... D The value is lower than f r (i.e., |f) D | < f r ), then according to equation (47), fe、1 > 0, i.e., f e、1 = |f e、1 |, and f e、2 < 0, i.e., f e、2 = -|f e、2 The difference in the received frequency values ​​can be expressed using equation (47):

[0284]

[0285] Furthermore, in the considered |f D In the case of | < fr, for the difference:

[0286]

[0287] For the Doppler frequency shift f D ≥ f r Both receiving frequencies are non-negative, i.e., f e、1 = |f e、1 | and f e、2 = |f e、2 Therefore, using equation (47), we obtain:

[0288]

[0289] Regarding the Doppler frequency shift f D ≤ -f r Both receiving frequencies are negative, i.e., f e、1 = -|f e、1 | and f e、2 = -|f e、2 Therefore:

[0290]

[0291] Therefore, the difference |f| between the measurements of the two receiving frequencies can be obtained. e、1 | - |f e、2 The text clearly distinguishes and defines three situations: :

[0292]

[0293] For each of these three cases, the signs of the two receiving frequencies are clearly defined (see the corresponding sections above), so the Doppler frequency shift f can be explicitly determined using equation (48). D :

[0294]

[0295] So far, the frequency shift f caused by propagation time has been considered. r The positive sign (i.e., since the modulation bandwidth B of equation (11b) is < 0); a similar consideration applies to the negative sign (i.e., the modulation bandwidth B > 0):

[0296]

[0297] These two relationships can be expressed using the sign of the modulation bandwidth B, V. B = ±1 In summary, then, for the radial relative velocity v = λ / 2·f D It can be clearly concluded that:

[0298]

[0299] This also needs to take into account the calculation error and the slight changes in relative speed and / or distance between the two detection cycles. The value will slightly exceed the actual limit. Therefore, "≥" or "≤" should be used for comparisons instead of "=". Thus, the relative velocity v can be explicitly determined by first calculating the difference between the two received frequencies. and with The comparisons are then performed, and the corresponding equations are applied to the three resulting regions to calculate v. Only when f... r = 0, that is, when the distance is zero, the sign of the relative velocity v cannot be determined, because at this time the second and third cases in the above equation (52) cannot be distinguished and the first case cannot occur (this ambiguity can also be explained by the fact that the receiving frequency consists only of the Doppler frequency shift in terms of magnitude alone). Due to the error and change between the two detection cycles, even if f r The value of f is very small, which can lead to the inability to determine the sign of the Doppler frequency shift, thus making it impossible to determine the relative velocity; however, f r The very small value corresponds to the area directly in front of the sensor, where no object is normally present, and even if an object is present, the sign of its relative velocity is known from historical records and / or is approximately zero in most cases. It is important to note that, in order to derive equation (52), the Doppler frequency shift f, obtained by averaging from two detection cycles according to equation (48), is effectively used. D Of course, a Doppler frequency shift f based on only one detection period can also be used. D Calculations are performed according to equation (47a) or (47b). Furthermore, the frequency shift caused by a slightly different distance in the two detection cycles can be considered; then, different values ​​of f are used in equations (47a) and (47b). r、1 and f r、2 And add their difference as a smaller component in equation (48).

[0300] Therefore, the ambiguity of relative velocity can be resolved by comparing the received frequencies measured in two detection cycles. As with normal tracking, this requires a cross-cycle allocation of the detection results. However, in traditional methods of resolving ambiguity through tracking, significantly more cycles are typically required because, in the aforementioned cases, the measured distance curve—the distance change measured across cycles—is compared to the distance change expected by each corresponding velocity assumption; this requires multiple cycles when the received frequency magnitude is small, thus resulting in minimal difference between the two velocity assumptions.

[0301] By changing the modulation bandwidth, for example by alternating the sign of the signal, the velocity ambiguity can be resolved not only in two detection cycles. Alternatively, each pixel can be acquired in a single detection cycle with two different modulation bandwidths B, either with different values ​​or signs; however, given the hardware, this would halve the number of pixels or require doubling the number of parallel transmit / receive paths. To avoid this, closely spaced pixels with different modulation bandwidths B, particularly adjacent pixels, can be acquired, rather than identical pixels; since the actual object is typically expanded and detected across multiple pixels, and the relative velocity and distance are at least approximately the same, equation (52) can be applied again to explicitly determine the relative velocity v, provided that the two receiving frequencies belong to two pixels with different modulation bandwidths B, either with different signs. For closely spaced pixels, particularly adjacent pixels, a different B approach is to select different B values ​​in systems with different scan planes, particularly adjacent scan planes.

[0302] Continuous scanning is performed by changing the frequency.

[0303] The following describes one possibility for performing continuous scanning in a detection plane. Here, devices whose radiation direction (for transmission and reception) depends on the frequency are used; examples of such devices include dispersive materials, lattice structures, or waveguides, with the latter being the focus of this paper (of course, the method shown can also be applied similarly to other devices). Figure 16 A waveguide 16.2, 1 cm in length, implemented in a photonic semiconductor chip 16.1, is shown. Its side feed 16.3 has a frequency of f, and the corresponding spacing of the equidistant coupling points 16.4 is λ0 / 2, where λ0 = 1550 nm (outdoor wavelength to f0 = c / λ0 = 194 THz). The coupling points are located at... Figure 16The waveguide is represented in a midpoint shape, but it can also have some scalability. Coupled points are used to couple out the wave during transmission and to couple in the wave during reception. The waveguide structure is periodic, meaning that each pair of adjacent coupled points has the same shape. Therefore, the corresponding phase difference Δφ(f) between two adjacent coupled points depends on the frequency f used and is a constant. Thus, the phase of the wave transmitted at the k-th coupled point (k = 0, ..., K-1, where K = 1290°) is...

[0304]

[0305] in," mod 2π "" indicates a modulus of 2π, which is a symmetric modulus function mapped to the symmetric region -π, ..., +π; the composition of the modulus takes into account the fact that there are usually many wavelengths between two coupling points, and also to make the correlation between phase difference and frequency as strong as possible, which can be achieved through Figure 16 The meandering structure shown is implemented (as shown in the figure, the meandering structure can extend inwards or be parallel to the chip surface). If the phase difference Δφ(f) is an integer multiple of 2π (i.e., there are an integer number of wavelengths between each coupling point), then the radiation direction is perpendicular to the waveguide. In other cases, i.e. mod 2π In the case where (Δφ(f)) ≠ 0, such as... Figure 16 As shown, the radiation direction is tilted at an angle of γ1(f), which is determined by the radiation length difference Δl and the phase difference between adjacent coupling points. mod 2π (Δφ(f)) compensation definition; since the phase difference between the radiation length difference Δl and the outdoor space wavelength λ = c / f is -2π·Δl / λ, and Δl = sin(γ1(f))·λ0 / 2, where λ0 = c / f0, therefore we can conclude that:

[0306]

[0307] Here, "asin" represents the arcsine function. For example... mod 2π (Δφ) = -π / 2 and at least approximately λ = λ0, then the angle γ1 = 30°. The radiation angle applies to both transmitting and receiving—which also conforms to the reciprocity principle.

[0308] Continuous spatial scanning is achieved through continuous frequency variations; that is, when acquiring pixel data, the frequency also changes at least in an approximately linear manner. Therefore, as... Figure 5 As shown, the frequency variation superimposed on the phase modulation is a direct result of the frequency variation scanning. As mentioned above, the required sampling frequency f... sIt increases with the modulation width B, which is the frequency change during one pixel, which effectively limits the sampling frequency—for the pixel duration T considered here. pm = 13.7 microseconds and a maximum effective range of slightly less than 250 meters, its magnitude should be |B| ≤ 2GHz (for f s (≤ 800MHz). If (e.g., in the horizontal direction) more than 500 pixels are to be scanned, the total frequency needs to be changed by about 1 THz, which is equivalent to about 0.5% of the average frequency f0 = c / λ0 = 194 THz (at an outdoor space wavelength of λ0 = 1550 nm). For a hypothetical scan range of -20°, ..., +20°, the phase difference Δφ(f) between two adjacent coupling points must change by about +0.34π, ..., -0.34π, or about 0.68π. This requires a high frequency-dependent sensitivity of the phase difference Δφ at a 0.5% frequency change. In addition to using a meandering waveguide curve, a high-dispersion waveguide can also be used within the frequency range used.

[0309] First, assume that the radiation angle γ1 varies linearly with time within the range of -20°, ..., +20°, i.e.

[0310]

[0311] Furthermore, the phase difference Δφ(f) has a constant slope over the small frequency range under consideration:

[0312]

[0313] Here, it is considered that the radiation direction should be zero at the intermediate frequency f0 (i.e., mod 2π (Δφ(f0) = 0). According to equation (54), we can conclude that:

[0314]

[0315] Therefore, the desired time-dependent frequency curve f(t) is obtained as follows:

[0316]

[0317] For a scan duration of 4.5 milliseconds (ms) (500 partially overlapping pixels, spaced at 9 microsecond intervals) and a pre-defined frequency variation of 1 THz (→ F1 = 2.13 / THz), Figure 17 The diagram shows the curves of the radiation angle γ1(t) and frequency f(t) over time. It is important to note that there is actually a small frequency difference in the transmission and reception frequency f(t) (due to the Doppler effect and propagation time shift), where the resulting difference in the radiation angle is much smaller than the radiation width itself.

[0318] According to equation (56), the linear change of the radiation angle γ1(t) with time implies that the angular resolution is constant across the entire detection range with equal pixel spacing that varies with time. In particular, for lidar sensors pointing forward, i.e. towards the driving direction, different angular resolutions may be more advantageous: a higher angular resolution is needed along the driving direction (γ1 = 0), while a faster scanning speed can be achieved in the outward direction (γ1 = ±20°; γ1 here refers to the horizontal direction as an example). Figure 18 This example illustrates the point. It's worth noting that the faster the scanning speed, the larger the target area scanned by the lidar radiation. Therefore, the signal received within a single pixel acquisition duration will lose coherence (the object will be detected at least partially different points at different times). This reduced coherence slightly decreases sensitivity and thus the effective range, but this is not significant in the outer regions where the effective range requirement is lower. On the other hand, Doppler accuracy will slightly decrease (because the peak values ​​in the two-dimensional correlation of the Doppler dimension will be slightly blurred), but this is also not crucial because Doppler accuracy is typically extremely high, i.e., the relative speed is extremely high.

[0319] Even with a constant scan rate, the frequency curve is not perfectly linear in time according to equation (56) according to equation (58) (especially due to the inclusion of a sinusoidal function). With an inconstant scan rate, the nonlinearity of the frequency curve in time becomes more pronounced (see also...). Figure 18 Another reason for the temporal nonlinearity of the frequency curve is that the waveguide has a strong dispersion property (i.e., according to equation (57), the linear relationship between the phase difference Δφ(f) and the frequency f no longer holds). In a detection scan (i.e., 500 pixels in the example above), this kind of nonlinearity of the temporal frequency curve will cause a correlated nonlinearity in the frequency curve even in a single pixel, i.e., a frequency error f. r、Q (t), especially according to the quadratic form of equation (39b). As mentioned above, this type of frequency error can be addressed by the rotation factor d. n、m The corresponding correction part is compensated; in the case of quadratic error, it is only necessary to adjust according to equation (44). Figure 11 The structure shown includes register R2. By adjusting the value of this register, the modulation bandwidth B, which varies with the pixel (due to the slope of the frequency curve f(t), can also be considered). Thus, the output dimension of the Fast Fourier Transform (FFT) always corresponds to the same relative velocity.

[0320] In principle, almost any scanning curve can be achieved through a corresponding frequency variation curve—a major advantage of scanning via frequency variation. Within each cycle, the detection range can be adjusted, especially according to traffic conditions; thus, a wider detection range is more important at lower vehicle speeds (e.g., urban traffic) than at higher vehicle speeds (e.g., highways). Problems of improper sensor adjustment (e.g., the sensor is not facing the direction of travel but offset 2° to the right) can be easily resolved by adjusting the frequency range used accordingly, essentially by slightly shifting the frequency range.

[0321] Instead of the continuous scanning considered so far, stepwise scanning could also be used in principle, where the frequency changes with each pixel, but remains constant within each pixel (in which case a phase modulation with a constant frequency, i.e., without superimposed frequency modulation, could also be used). However, this stepwise scanning itself has some drawbacks (nevertheless, it should not be ruled out as a possible implementation), including:

[0322] - The inability to achieve pixel overlap results in a reduced data acquisition time per pixel for a given hardware and cycle time. Furthermore, the data acquisition time is further reduced due to the time required for frequency switching (especially before the frequency reaches a new value) and the subsequent wait for the maximum propagation time of the received signal (approximately 1.67 microseconds at a maximum distance of 249.5 meters). In the design to date, the considered pixel interval is approximately 9 microseconds, resulting in a data acquisition time of only 6 microseconds, more than half that of continuous scanning with a pixel duration of 13.7 microseconds. The main drawback of this significant reduction in data acquisition time per pixel is a decrease in sensitivity (approximately 3.5 dB under the aforementioned conditions), leading to a reduction in effective range (approximately 19%); the decrease in resolution and accuracy for determining relative speed is not critical. The loss would be even greater if the pixel pitch were reduced by, for example, a factor of 2 to halve the number of parallel transceiver paths (replacing 32 paths with 16).

[0323] In a real-value mixer, the sign of the relative velocity cannot be determined—the above method of resolving ambiguity by using the unknown sign of the received frequency is based on superimposed linear frequency modulation.

[0324] - If the radiation width is less than the radiation motion during the data acquisition pause between two pixels by using extremely fast scanning (e.g., outward scanning), objects with extremely narrow scanning directions may be missed.

[0325] - Continuous frequency scanning may be simpler and better to implement than hierarchical scanning.

[0326] - Continuous frequency scanning can slightly reduce speckle effect (i.e., statistical changes in received levels from the diffuser) because speckle effect is frequency-dependent; in addition, spatial scanning can also slightly reduce speckle effect because the illuminated area changes slightly during the data acquisition of one pixel.

[0327] To achieve frequency variation, the varying frequency can be modulated to a constant frequency of the laser source (which is difficult due to the large required frequency range), or a laser source with a directly controllable frequency can be used. Such lasers are typically controlled by a mechanical variable (e.g., using piezoelectric devices) or an electrical variable (voltage, current). The generation of the electrical control variable can be digitized on a computing unit, such as a processor, and converted to an analog range using a digital-to-analog converter (DAC), followed by low-pass filtering if necessary. Since the frequency curve and the required curve for the control variable are quite low-frequency from a signal theory perspective, an integral-differential modulation method can be used to determine the DAC input value, especially reducing the resolution requirements of the DAC. Besides direct control, a phase-locked loop (PLL) control loop can also be used. If the laser source cannot achieve a sufficiently large frequency tuning range (partly to accommodate the frequency variations required for scanning, such as 1 THz, and partly to cope with the effects of temperature, aging, and errors), multiple laser sources can be implemented and switched between them to select the appropriate source. As an alternative, the frequency variation required for scanning can be significantly reduced by using waveguides with slightly different frequencies for the parallel transmit and receive paths. The phase difference Δφ(f) of these waveguides is near the coupling point—at frequency f, the waveguides radiate in different directions, and their respective small scan regions are connected to each other (with some overlap if necessary); this method will be explained further later.

[0328] Finally, it is also worth mentioning that, Figure 16 The waveguide shown not only achieves scanning, but also focuses in its corresponding spatial direction. That is, at each cross section of the waveguide (more precisely, on the straight line where its coupling point is located), the wave has a flat wavefront portion.

[0329] Scanning in the second spatial direction

[0330] The following explains how to achieve focusing and scanning in a second spatial direction (perpendicular to the spatial direction defined by the waveguide). Figure 19 As shown, focusing can be achieved using a lens 19.2; Figure 19The arrangement is shown at the top, extending from waveguide 19.1. The waveguide is located on the focal plane of the lens, but offset from the focal line (this occurs because multiple adjacent waveguides will be considered later). Therefore, after being focused by the lens, the radiation constituting the plane wavefront is tilted relative to the optical axis of the lens by a radiation angle γ. 2、1 (exist Figure 19 Above, γ 2、1 The underscore is used as the separator. γ 2、1 (Because only the projection angle is indicated there). Since the waveguide itself is already focused in the first spatial direction, the lens only needs to be focused in the second spatial direction, which is perpendicular to the first spatial direction, so as to have a constant cross-section in the waveguide direction; such as Figure 19 As shown below. There, two radiation angles γ1 and γ2 can also be seen. 2、1 In a three-dimensional environment, these are defined—therefore, they are not angles as described in spherical coordinates; for simplicity, Figure 19 The lower part combines the refraction of the lens on the two surfaces into one.

[0331] It is important to note that a lens with a constant cross-section is easier to manufacture in one dimension than a lens without this property, especially a condenser lens used in two spatial directions. In addition to a single lens, a lens system can also be used, in which the cross-sectional properties must also be kept constant in the waveguide direction.

[0332] For scanning along this second spatial direction, a material that is transparent to the wavelength used can be used, whose dielectric constant can be changed by applying a voltage (which generates an electric field in the material) or by applying a current (especially to generate a magnetic field in the material); liquid crystals and ferroelectric materials are examples of materials with such properties. Figure 20 Above (and) Figure 19 (From the same viewpoint above) An example arrangement is shown, in which the body 20.3, made of this material, has a constant cross-section (triangular) along the extension direction of the waveguide 20.1, just like the lens 20.2, and thus has a prism shape; Figure 20 The image below shows a side view, taken from a 90° rotation, where the voltage U applied to both sides of the prism-shaped body 20.3 can also be seen. This voltage is applied directly to the body or through the current it generates, thereby changing its dielectric constant. By changing the applied voltage U, the incident angle γ of the prism 20.3 can be altered. 2,1 The difference γ between the emission angle γ2 and the emission angle γ2 2,2This alters the radiation angle γ2 itself, thereby enabling scanning in a second spatial direction orthogonal to the first spatial direction. It is also worth noting that the prism-shaped body is preferably larger than required by the actual radiation path to obtain the most uniform electric field possible in the relevant region, thus achieving the most constant spatial dielectric constant. In principle, it is also possible to consider forming a common body with a constant cross-section in the waveguide direction, consisting of the lens and the prism shape; however, the change in dielectric constant cannot remain constant across the entire cross-section, otherwise it will alter the focusing characteristics and thus the focal plane (therefore, a suitable field curve that is not constant across the cross-section is required). Besides this prism-shaped body, planar transmissive liquid crystal devices, especially one-dimensional grating structures, can be used to redirect radiation to the second spatial direction by applying a voltage between the upper and lower parts.

[0333] As an alternative to focusing and scanning, a liquid crystal array can be used. Figure 21 The illustration shows a possible arrangement, which includes a transparent one-dimensional array 21.2 and a rod-shaped device 21.3; Figure 21 The top view shows the array as seen from above, the middle view shows the arrangement and radiation path as seen from the direction of waveguide 21.1, and the bottom view, rotated 90°, shows the arrangement and radiation path as seen from the side. By applying appropriate voltages to a single rod-shaped device (between its upper and lower parts), the phase difference between the incident wave on one side and the emitted wave on the other side can be changed—the achieved phase difference depends on the value of the applied voltage. This allows the upper wave phase to have a linear curve, while achieving focusing (i.e., generating a plane wave) and a preset radiation angle γ2; by changing the applied voltage, the radiation angle γ2 can be achieved, thereby enabling scanning in the direction of the second ray.

[0334] Liquid crystals (LCDs) typically suffer from slow response to control changes, requiring a considerable amount of time to stabilize to a new control state. This is particularly critical when using a two-dimensional LCD array for two-dimensional scanning, as adjustments are needed after each pixel switch. Conversely, a one-dimensional LCD array only needs to complete the scan in the first spatial direction, approximately 4.5 milliseconds in the example above, which is not a problem at all. Another advantage compared to LCD arrays used for two-dimensional scanning is the significantly lower control voltage required (due to the single dimension).

[0335] By using an additional lens, focusing can be achieved essentially in a second spatial direction, and a constant cross-section can be achieved again in the waveguide direction, thereby reducing the number of rod-shaped liquid crystal devices required and thus reducing the required control voltage; therefore, the liquid crystal array basically only needs to achieve rotation in the radiation direction, especially when the rotation range is small, allowing a distance much greater than the outdoor spatial wavelength λ0 = 1550 nm.

[0336] One drawback of the first method, which uses lenses and prisms with controllable refraction, is that the lens geometry and its distance from the waveguide require high precision (otherwise, optical blurring may occur due to factors such as focal plane shift). This high precision increases the manufacturing cost of the sensor and / or raises component prices due to smaller mechanical errors. In contrast, in the second method based on a liquid crystal array, positional errors can be easily compensated for by correspondingly controlling the liquid crystal array devices to correct the corresponding phase errors; similarly, the positional and cross-sectional errors of the lenses, if applicable, can also be corrected.

[0337] Apart from Figure 21 In addition to the transparent liquid crystal arrays considered so far, a reflective liquid crystal array can also be used. For reflective liquid crystal arrays, the phase difference between the incident and emitted waves can also be effectively achieved by applying appropriate voltages, which can be considered as a local emission angle variation relative to the incident angle. The corresponding arrangement is as follows... Figure 22 As shown; the top view illustrates an arrangement with a waveguide 22.1, a polarization-dependent mirror 22.4 (reflecting one polarization and allowing projection of radiation perpendicular to that polarization), and a one-dimensional reflective array 22.2 with rod-shaped devices 22.3, rotated 90° from above; the middle view shows the arrangement and radiation path as viewed from the direction of the waveguide 22.1; and the bottom view, rotated 90°, shows the arrangement and radiation path as viewed from the side. The mirror 22.4 for radiation deflection allows the photonic chip 22.5 carrying the waveguide and the liquid crystal array 22.2, including its control device, to be placed on a circuit board 22.6, thereby reducing complexity and manufacturing costs. Of course, an arrangement without the mirror 22.4 can also be implemented, i.e., with a direct radiation path between the waveguide and the reflective liquid crystal array, in which case, for example, two circuit boards are required.

[0338] If the sensor's optical axis is, for example, perpendicular to the circuit board, one or more additional mirrors can be used to deflect the radiation.

[0339] All the other considerations mentioned above regarding the arrangement of transparent liquid crystal arrays also apply to the arrangement of reflective liquid crystal arrays—in particular, manufacturing errors of the relevant optical components and their relative placement can be compensated by the corresponding controllers of the liquid crystal array.

[0340] To date, liquid crystal arrays have been considered one-dimensional. Due to errors (e.g., the thickness or optical properties of the liquid crystal array itself are not constant within the extension of the rod-shaped device), it may not be possible to generate perfectly flat waves in the first spatial direction (i.e., the direction defined by the waveguide). To avoid this, such as Figure 23As shown, the rod-shaped device can be divided into multiple individual devices, and phase errors can be compensated by controlling these devices accordingly. However, such a two-dimensional liquid crystal array is more complex and requires more control voltages. Compared with two-dimensional liquid crystal arrays used for two-dimensional scanning, the number of devices required in this one dimension (where the rod-shaped device is divided into multiple devices to compensate for errors) is significantly less; if necessary, the grating structure can also include multiple regions driven by different voltages, especially to compensate for manufacturing tolerances in the arrangement of optically related components and their mutual arrangement.

[0341] Besides reflective liquid crystal arrays, reflective liquid crystal devices can also be used, especially a one-dimensional grating structure that deflects radiation to a second spatial direction by applying a voltage. If necessary, this grating structure can also have multiple regions, driven by different voltages, particularly for compensating for manufacturing tolerances between optically related components and their arrangement. For this purpose, the radiation must be pre-focused; and... Figure 22 Compared to the previous arrangement, it requires the use of focusing mirrors (approximately parabolic), reflecting lenses (i.e., one side with a reflective coating), or a combination of mirrors and lenses, in which these devices still have a constant shape in the waveguide direction.

[0342] Finally, it is worth mentioning that materials whose optical properties can be altered by applying an electrical control quantity can be used in other devices, besides the structures described above, to achieve scanning in a second spatial direction.

[0343] Waveguide arrangement and scanning mode

[0344] As mentioned several times above, there are, for example, 32 parallel transmit and receive paths, and therefore 32 waveguides; the following will explain the arrangement and design of these waveguides on the photonic chip. In principle, these waveguides can be used to achieve additional parallelism in detection in the first spatial direction (scanning through waveguides with varying frequencies) and / or the second spatial direction (scanning through materials with electro-controllable optical properties, such as liquid crystal arrays).

[0345] First, consider using all waveguides to achieve parallelism, i.e., a method of parallel detection in a second spatial direction. For example... Figure 24 As shown, 32 waveguides 24.2 are arranged side-by-side on a photonic chip 24.1 (winding paths are not shown if necessary for illustration); all waveguides employ the same design, thus each can achieve the same radiation angle γ1 at the same frequency. Depending on the distance between the waveguides, they can directly address adjacent pixels or pixel lines, or even more distant pixel lines with other pixel lines in between, for a second spatial direction; at least approximately, the geometric distance between the waveguides is proportional to the geometric distance between the pixel lines they achieve (if larger angles are ignored). It is initially assumed that the distance between the waveguides allows them to achieve 32 directly adjacent equidistant pixel lines. Figure 25This illustrates how to construct a two-dimensional pixel field (i.e., a scan pattern); where the terms "w," "f," and "u" used to assign to each pixel are defined as follows:

[0346] - w represents the number of each corresponding waveguide: w = 1, ..., 32,

[0347] - f refers to scanning via a waveguide using continuous frequency variations, i.e., continuous scanning in the first spatial direction, and thereby continuous scanning of the radiation component γ1; this continuous scanning will scan all 500 pixels through this first spatial direction, where f = 1, ..., 500 are defined as the corresponding frequency numbers (more precisely, the corresponding center frequencies, because the frequency changes continuously within a pixel).

[0348] - u refers to the stepwise scanning using a material with electro-optical properties, such as a voltage-controlled liquid crystal array, i.e., stepwise scanning in the second spatial direction, thereby scanning the radiation direction component γ2; here, u represents the number of stepwise scans (and thus also the gradually changing electrical control variable, such as the applied voltage) — after a continuous frequency scan in the first spatial direction that takes about 4.5 milliseconds, a new γ2 scan will be performed, i.e., the next continuous frequency scan will be performed on the new γ2; since each of the 32 waveguides itself realizes a different γ2, for the 320 pixels in the second spatial direction, i.e., 320 different γ2s, only 10 scan steps u = 1, ..., 10 are needed, and the radiation direction will jump 32 pixels from one step to the next.

[0349] This arrangement requires a small waveguide spacing, which may be difficult to achieve, especially when the waveguide is highly curved. As an alternative, the spacing can be increased by a factor of 32, enabling the waveguide to achieve pixel lines with a spacing of 32; for example... Figure 26 As shown. During scanning in the second spatial direction, only one pixel is skipped per step; this means that the required scanning range in the second spatial direction can be significantly reduced by using materials with electro-controllable optical properties, which is advantageous and expands the possibilities for simpler or more implementations. This arrangement also allows for the aforementioned method, i.e., determining the sign of the received frequency in a real-value mixer by selecting different frequency modulation bandwidths B in adjacent scanning planes, provided that the object is visible in at least two adjacent pixels (i.e., adjacent scanning planes, i.e., pixel lines); for this purpose, alternating directions of continuous frequency scanning for the first spatial direction must be applied over 10 scanning steps u = 1, ..., 10, thereby achieving alternating signs of the modulation bandwidth B.

[0350] So far, it has been assumed that waveguides are at least approximately equidistant from each other, which would also result in equidistant pixels in the second spatial direction. However, it is often only necessary to achieve maximum resolution within the central angular range, while resolution can gradually decrease outwards—and the gaps between pixels in this range can also be larger. This can be achieved through a non-equidistant arrangement of waveguides. As a simplified case, assume that the corresponding distance 'a' between the 16 intermediate waveguides is small and equal (so that there are no gaps between the pixel lines they generate), while the distance between the corresponding 8 outer waveguides is twice that, i.e., 2·a, and their distance from the group of 16 intermediate waveguides is larger, at 146·a. Figure 27 The diagram shows the resulting two-dimensional pixel field; during scanning in the second spatial direction, each step jumps 16 times the pixel distance (relative to the pixel distance in the middle range). At the upper and lower thirds of the pixel field, the detection density is only half that in the middle.

[0351] Unlike previous arrangements, waveguides can also be used for parallelism, i.e., for parallel detection in a first spatial direction. For this purpose, the waveguide design differs to achieve different radiation angles γ1 at the same frequency; the waveguides are still arranged parallel to each other (e.g., Figure 24 As shown, the distance between the connection points remains the same, but the length of the meandering waveguide between the two coupling points needs to be designed differently for each. Figure 28 A two-dimensional pixel field is shown when the waveguide spacing is small and constant; here, the spacing is chosen such that the realized pixel lines are each one pixel apart in the second spatial direction. Different designs of 32 waveguides enable them to realize radiation direction components γ1 (in the first spatial direction) that are 16 pixels apart at the same frequency; to detect the entire first spatial direction (now consisting of 16·32 = 512 pixels), only 16 pixels need to be scanned consecutively at each frequency, thus the number of frequencies extends only to the range f = 1, ..., 16. The stepwise scanning in the second spatial direction (using materials with electro-controllable optical properties) must now completely cover this spatial direction, i.e., all 320 pixels, thus requiring 320 steps. Figure 28 As shown, the two-dimensional pixel field is no longer a precise rectangle, but rather forms a parallelogram. A slightly rotated approximate rectangular shape can be achieved by slightly altering the frequency range used to scan the first spatial direction in each of the 320 steps. This slight rotation of the rectangle can be compensated for by rotating the photonic chip accordingly on the circuit board, or by rotating the waveguide arrangement of the chip itself.

[0352] The advantage of this method is that the frequency range required for scanning in the first spatial direction is much smaller (about 32 times smaller, which is beneficial for realizing a laser source), or, with the frequency range being the same as before, the waveguide’s frequency sensitivity to the phase difference Δφ is greatly reduced, thereby significantly shortening the waveguide length between the two coupling points and thus reducing waveguide loss; however, the frequency variation of each pixel will be much higher at this time, so continuous scanning will lead to an excessively high sampling frequency, thus requiring a gradual change in frequency.

[0353] One drawback of this method is that the increased number of steps leads to a slightly higher cost for reconfiguring control variables for scanning (for materials with electro-controllable optical properties and frequency modulation of the laser source).

[0354] By using continuous scanning in the second spatial direction, instead of stepwise scanning, the drawbacks of increased cost and excessively high scanning frequency associated with control variable reconfiguration can be overcome—that is, control variables for materials with electro-controllable optical properties are changed continuously, not stepwise. During continuous scanning of the entire detection range in the second spatial direction, continuous scanning at the frequency of the first spatial direction moves only one pixel; for example... Figure 29 As shown. Therefore, a total of 16 consecutive scans are required. With this method, the speed of continuous frequency scanning is much slower; therefore, using a simple waveguide with low frequency correlation sensitivity, the time-frequency change of the phase difference Δφ (i.e., the approximately linear frequency modulation superimposed on the phase modulation) is small enough to allow for the use of a moderate sampling frequency. During continuous scanning in the second spatial direction, frequency scanning in the first spatial direction can also be performed incrementally, i.e., keeping the frequency constant during each scan in the second spatial direction and then setting the frequency to a new value during the next scan in the second spatial direction; thus, 16 different frequencies will be used. In addition to a frequency-adjustable laser source, for example, 16 different constant-frequency laser sources can be used, with switching between these laser sources.

[0355] Continuous scanning in the second spatial direction can be performed at a non-constant scanning speed. For example, for a lidar sensor looking in the direction of travel, a higher angular resolution can be achieved in the direction of travel than at the edge of the detection range.

[0356] Of course, parallel detection through 32 waveguides can also be divided into two spatial directions to combine the final advantages (of course, the advantages will be reduced accordingly).

[0357] The method of using multiple different waveguides for different radiation directions in the first spatial direction can also be implemented in a simpler lidar system with fewer pixels, which in particular means that only one transmit / receive path can be switched to multiple waveguides; thus, the frequency tuning range required for the laser source is reduced in particular.

[0358] Alternative methods for scanning in the second spatial direction

[0359] To date, scanning in the second spatial direction has assumed the use of materials with electro-optical properties. Of course, other methods are also conceivable.

[0360] In the previous example of continuously scanning the second spatial direction (i.e.) Figure 29 In the scanning mode shown, a mechanical method (e.g., using a rotating mirror or prism) can also be used.

[0361] like Figures 25 to 27 The scanning mode shown is achieved by progressively switching in the second spatial direction. This can be accomplished by using 10 waveguides instead of one on each of the 32 parallel transmit / receive paths, which can be switched between each other. This results in a device with 320 adjacent waveguides, of which only 32 are used during the frequency scan in the first spatial direction. It is also worth noting that in this switching method, cascaded single-pole switches (i.e., with two inputs and one output) are typically used, so the input to the entire switching matrix is ​​a power of 2, for example, 8. In the example above, this means 256 waveguides, and therefore 256 pixels in the second spatial direction. If the pixels are sparsely distributed outwards in the second spatial direction, i.e., with a large angular spacing, the outward spacing of adjacent waveguides needs to be increased. If the pixels are sparsely distributed outwards, to avoid detection gaps, it is generally (not limited to this embodiment) preferable to increase the radiation width outwards in each corresponding spatial direction. This is often inherent to optical focusing errors, or in this example, it can be achieved by placing the waveguides outside the focal plane.

[0362] For a simple lidar system, especially when the number of pixels in the second spatial direction is small, it may be sufficient to use only a single transmit / receive path and switch it to multiple adjacent waveguides via a switching matrix to detect the second spatial direction.

[0363] Finally, it is worth mentioning that for all the methods described above, the two mutually perpendicular spatial directions are preferably set in the horizontal and vertical detection directions, corresponding to the azimuth and elevation angles, respectively. As mentioned above, the two radiation direction components γ1 and γ2 do not conform to the definitions of azimuth and elevation angles in spherical coordinates. In principle, both allocations are feasible, meaning both spatial directions can be used for either the vertical or horizontal detection direction; however, due to different requirements for detection range and / or resolution, there may be more advantageous allocations.

[0364] Determine the angle error

[0365] The radiation directions with components γ1 and γ2 depend on the control variables and hardware characteristics, especially on:

[0366] - The relationship between waveguide radiation angle γ1 and frequency.

[0367] - The relationship between laser frequency and its control variables

[0368] - The relationship between the radiation angle γ2 and the control variables of materials with electro-controllable optical properties.

[0369] These relationships vary significantly depending on the sensor, temperature, and degree of aging. Accurate understanding of these relationships is crucial for accurate environmental detection; otherwise, angular errors (inconsistencies between measured and actual angles) will occur, leading to distortion and shifts in the detected images. While differences between sensors can be preliminarily determined during production, temperature-related factors are often unresolved (otherwise, the measurement workload in production would be excessive), and aging effects can generally only be determined during the sensor's lifespan. Therefore, a method is needed to determine angular errors during operation; based on this, compensation for angular errors can be achieved by modifying the control, i.e., changing the selection of control variables.

[0370] In addition to sensor angle errors caused by changes in hardware characteristics, sensors can also produce errors due to orientation errors within the vehicle (especially due to mechanical errors, vehicle load, and aging effects). These orientation errors have a continuous impact on all detection directions in the relevant spatial orientation, while hardware-induced imaging errors may depend on the angle.

[0371] Now let's explain how to determine sensor imaging errors when viewed along the direction of travel. As is well known, radar sensors determine angular errors (especially errors due to orientation errors) by comparing the measured radial relative velocity of a stationary object with its own velocity. The radar sensor measures the radial component of the object's relative motion; relatively speaking, the stationary object moves at the vehicle's own velocity v. 本车速度 (v ego ) moves parallel to the direction of travel toward the sensor, therefore, as Figure 30 As shown, the motion v 本车速度 The measured radial component v m It still needs to be multiplied by the cosine of the azimuth angle α of the stationary object:

[0372] v m = v 本车速度 ·cos(α). (59)

[0373] However, this relationship only holds true when the elevation angle is zero; since radar sensors have only a small detection range at elevation angles close to zero, and the above relationship provides a good approximation for small elevation angles, it is typically used only in radar sensors. If the radar sensor has a orientation error, the azimuth angle α measured by the sensor... mThe actual azimuth angle α is inconsistent with the expected measured velocity v. 本车速度 ·cos(α m ) and the actual measured speed v 本车速度 • cos(α) does not conform. The actual angle α can be determined in principle using equation (59), and the difference between it and the measured angle represents the orientation error. It should be noted that radar sensors can usually only detect objects within a small azimuth range (at least with high quality), the cosine of small angles (i.e., close to zero) is very flat, and the vehicle's own velocity measured and transmitted to the sensor is usually of poor quality. Therefore, the relative velocity v measured for many stationary objects is not accurate. m (α m Perform a parabolic regression (the cosine of the smallest value is equivalent to a parabola), and then derive the angle of orientation error from the parabolic offset.

[0374] Like radar systems, coherent lidar systems can directly measure the Doppler effect, i.e., radial relative velocity (both systems operate coherently, differing only in the electromagnetic frequency range). Therefore, the above method can be applied to lidar systems. For lidar systems, in addition to orientation errors, angle-dependent imaging errors caused by changes in hardware characteristics can also be detected, including those in the horizontal and vertical spatial directions; in addition to equation (59), the radiation direction component α in the horizontal spatial direction and the radiation direction component β in the vertical spatial direction are also considered, such as Figure 31 As shown.

[0375]

[0376] If one of the two angles is zero, then use cos... 2 (γ) + sin 2 (γ) = 1 yields:

[0377]

[0378] This also corresponds to equation (59) above. When the values ​​of angles α and β are small, the following equation is used, and the approximation is quite good:

[0379]

[0380] According to equations (60) and (61c), the measured relative velocity depends on two angles α and β of each corresponding pixel; therefore, it is impossible to determine this from a single pixel without knowing the angular errors in the two spatial directions. However, the angular errors in the two spatial directions are independent of each other (they are determined only by hardware characteristics or control variables in each corresponding spatial direction); in a pixel field of 500x320, there are a total of 820 error variables (assuming the extreme case that the corresponding errors of adjacent pixels in the two spatial directions are independent of each other). If each pixel has a stationary object, then there are 500.320 measurements to determine 820 error variables, which far exceeds the required number (820 measurements would suffice). In reality, on the one hand, there are stationary objects in a small number of pixels, and on the other hand, within a spatial direction, the errors of adjacent pixels are not independent, but can be described with sufficient precision by an error curve having several parameters (such as polynomials or multiple linear or parabolic curves). Therefore, it is only necessary to determine these few unknown parameters. Typically, there are enough measurements for each detection cycle—and if necessary, the determination range can be extended to multiple cycles or many cycles, because temperature and aging effects are relatively slow in comparison.

[0381] Static reflections obtained from the short-range of the road surface are highly reliable; they cover most or all of the horizontal spatial direction. Since the sensors measure the reflection distance *r* from the road surface and the sensor mounting height *h* is known... sen Therefore, the actual angle β in the vertical spatial direction is also known, according to Figure 31 As shown, it satisfies the following conditions:

[0382]

[0383] The angle error curve for the horizontal spatial direction α can be determined using equation (60) or (61c), where only one unknown remains. In typical cases, given a small elevation angle and using equation (61c) with cos(β), the road reflections of different β (i.e., from different distances) can be simply averaged, and the angle error can be determined from the resulting curve for the horizontal spatial direction α. ​​If the angle error for the horizontal spatial direction α and the actual value of α for each pixel are known, the angle error β for the vertical spatial direction can be determined from each pixel (including those pixels that cannot be assigned to the road surface) (because only one unknown remains in equation (60) or (61c).

[0384] To determine the angular error using a fixed object, a high-precision vehicle speed v is required. 本车速度However, the vehicle's own velocity, measured and transmitted to the sensor, is usually not accurate enough. Therefore, the lidar sensor must determine its own velocity. The simplest approach is to determine the maximum value of the measured radial relative velocity, since, according to equation (60), the measured relative velocity can only reach the vehicle's own velocity at most—this is the case in the direction of travel (i.e., α = β = 0°). However, this approach relies on the presence of stationary reflections in or at least close to the direction of travel, and on the high accuracy of the relative velocity measurement (i.e., especially without noise due to a poor signal-to-noise ratio). As mentioned above, the angular error curve for each spatial direction can be described sufficiently accurately by an error curve with several parameters (e.g., polynomial or multiple linear or parabolic segments), so these unknown parameters only need to be determined from the measured values ​​(i.e., the relative velocity measured by a stationary object); in the case of an unknown velocity, this velocity will be another unknown and determined at least implicitly.

[0385] If the angle error is known, it can be corrected in the detection list determined by the sensor, thus specifying the actual angle for the detection result. Furthermore, the curves of the control variables used for scanning in subsequent detection cycles can be adjusted to eliminate the angle error (an iterative approach may be used if necessary)—especially for accurately achieving the expected detection range of the two-dimensional pixel field.

[0386] Determine the location of the road surface at a relatively long distance

[0387] As mentioned above, the road surface can be used to determine the angular error; here, the reflection of the road surface within a short range is used.

[0388] Reflection information from road surfaces at greater distances can reveal the precise location of the road surface at each corresponding distance (i.e., which pixel the road surface is at and which corresponding vertical angles, which can be further improved to sub-pixel accuracy through interpolation if necessary), which is very helpful in interpreting detection data. For example, the height of objects on the road can be accurately determined; if not only the measured reflection point of the object is known, but also the precise location of the object on the road surface, the height determination will be more accurate. It should be noted that the location of the road surface can only be known in advance at close range, based on the assumption that the sensor installation height is known and the road direction is not curved at short distances; at greater distances, this assumption no longer holds (for example, in a depression, a curved road direction will have a strong effect), and small changes in the vertical direction of the sensor (such as slight tilting of the vehicle during braking and acceleration) will also have a significant impact. However, lidar sensors generally cannot detect reflections from road surfaces at greater distances because the reflections are too weak; Figure 32The main reasons for this are shown (the vertical direction is not drawn to scale for ease of understanding). The figure shows a radiation with a width of 0.05°, whose center intersects a flat road at a distance of 100 meters, where the sensor is installed at a height h. sen = 60 cm; due to the very small angle of incidence, it constitutes radiation within a range of approximately 15 meters from the road. This means that in the two-dimensional correlation E m、k In this model, the road surface is visible at approximately 30 discrete distances m (not just at m0), so the received power is divided into a corresponding number of values ​​in the two-dimensional correlation. This results in a very poor signal-to-noise ratio, which may prevent detection. The reduction in received power at each distance value m in the two-dimensional correlation can also be explained by the fact that only about 1 / 30 of the radiation width and transmitted power are effective at each distance value.

[0389] To better detect the road surface, the fact that the road surface is observable on many pixels in the horizontal spatial direction α can be utilized; that is, at the same distance, the radiation centers of many pixels intersect the road surface; according to equation (62), these pixels have the same radiation angle β, i.e., the vertical spatial direction. A first approach is to accumulate the two-dimensional correlation E through these pixels (within the predicted driving direction) in an incoherent manner, specifically by using their magnitude or power values. m、k This involves accumulating at least all values ​​for the discrete distances *m* where the road might be located. When distance *m* represents the road's location, the accumulated value is significantly higher than the noise level; a single two-dimensional correlation E m、k The noise level is known (e.g., determined according to the calculation logic in Figures 8 and 10). The noise level after incoherent accumulation can then be calculated based on theoretical relationships—or estimated directly from the numerical values ​​of the incoherent accumulation, where a sufficiently large range of m must be evaluated, and different values ​​of the discrete frequency k must be evaluated if necessary. As mentioned above, the road can be observed over the entire discrete distance range m, which in the example covers approximately 30 values; the midpoint m0 ​​of this range represents the approximate actual distance of the road. In addition to incoherent integration over the values ​​of the discrete propagation time m, this can be extended to multiple adjacent m values, since the road is also observed over multiple or numerous m values.

[0390] It is not necessary to perform the above evaluation for every discrete frequency; in principle, only the frequency k0 containing road reflections needs to be evaluated. Since the vehicle's own speed is known (using high-quality detection data from lidar sensors), the discrete frequencies can usually be determined very accurately. Therefore, only one or a few discrete frequencies k need to be correlated. Using the calculation logic shown in Figure 8 or Figure 10, the two-dimensional correlation E of these discrete frequencies needs to be output. m、kNumerical values. If these discrete frequencies are variable, the implementation of the output becomes very complex; therefore, when using the computational logic shown in Figure 10, a rotation factor d, chosen depending on its own velocity, is used. n、m It is more beneficial to always set the required discrete frequency at the same output terminal of the computational logic.

[0391] Besides the two-dimensional correlation E m、k Besides performing incoherent integration on the numerical values, a second method can be chosen: divide the received sequence e(n) of the pixels into two half-length sequences, where the first sequence consists of even numbers n and the second sequence consists of odd numbers n. Determine the corresponding two-dimensional correlation for each of these two subsequences, and then, for each correlation distance m and frequency k, multiply the value of one correlation by the conjugate complex value of the other correlation. There is a phase difference between the two subsequences, which is determined by the received frequency f. e The time interval between each corresponding two sampled values, i.e., the sampling time T. s It is concluded that the product of the two correlations has a complex value at each distance m and frequency k0 of the road reflection, and its phase corresponds to the phase difference. This statement, of course, only applies to the signal part and not to the superimposed noise.

[0392] When m is a single value, this phase is at least approximately constant across all pixels with road reflections, because of the relative velocity and the receiving frequency f. e Within a small angular range of approximately 0°, the road surface can be considered constant (cos(α) in equation (61c) is approximately equal to 1). Therefore, this product can be coherently integrated, i.e., accumulated, between pixels, which improves the signal-to-noise ratio more effectively compared to incoherent integration. If integration is also performed over different m, the received frequency f... e The frequency component f in the distance depends on the distance. r Correction must be performed according to equation (12), preferably as the rotation factor d in the calculation logic shown in Figure 10. n、m Part of it. Using this computational logic, the two-dimensional correlations of odd and even n in the e(n) subsequence are also inherent—and appear in the output of the penultimate stage of the Fast Fourier Transform (FFT) of the alternating sequence. Further considerations and evaluations are equally applicable to the first method of incoherent integration.

[0393] This second method, consisting of the product of two correlations and summation across pixels, can also be generalized. Importantly, the two correlations must at least partially utilize received signals from the same reflection point on the road, i.e., the same area, to obtain the defined phase relationship. For example, the two correlations can be calculated using the first and second parts of the received sequence for each pixel, or, if they overlap, the correlation of two adjacent pixels; since the received values ​​belonging to the two correlations are temporally distant, they may become sensitive to changes in velocity (in which case the phase difference will change with the pixel sequence).

[0394] The sign of the received frequency in a real-value mixer is determined by non-binary phase modulation.

[0395] In the binary phase modulation with two phase values ​​of 0° and 180° considered so far, the two-dimensional correlation E is... m、k Two peak values ​​(m0, +k0) and (m0, -k0) will appear, thus only the amplitude of the received frequency can be determined. As mentioned above, this ambiguity can be at least partially resolved by using superimposed linear frequency modulation. An alternative method for determining the sign of the received frequency using a real-valued mixer will be introduced below.

[0396] Now, phase modulation φ TX (n) should not consist of two binary values, 0° and 180°, but rather a set of J phase values ​​φ. j The phase modulation sequence is composed of J phase values, where j = 0, ..., J-1, which can be assumed to have general phase values; furthermore, the phase modulation varies irregularly, especially pseudo-randomly, among these J phase values. The phase modulation sequence can be generated by a complex value...

[0397]

[0398] To describe this, using equation (3a), the received sequence e(n) of a single object is:

[0399]

[0400] Consistent with the above derivation, φ TX The negative sign of (n-m0) is because the complex-valued mixer is assumed to constitute a phase difference between the transmitted and received signals (in other cases, i.e., if the mixer constitutes a phase difference between the transmitted and received signals, it would be positive). The following real-valued components are obtained in a real-valued mixer:

[0401]

[0402] Two-dimensional correlation E in the form of equation (8) m、k From product sequence

[0403]

[0404] The Fast Fourier Transform (FFT) is constructed, where b(n) is calculated according to equation (63), and no conjugate complex value is constructed for b(nm), which leads to a correlation corresponding to the first part of e(n) at frequency +k0 in equation (65) (using the conjugate complex value of b(nm) would lead to a correlation corresponding to the second part at frequency -k0). According to equation (65), we can derive:

[0405]

[0406] In the discrete propagation time, when the object distance m = m0, the first part p1(n) represents the continuously rotating pointer of the object's discrete frequency +k0 (because φ TX (n-m0)-φ TX (n-m0) = 0), therefore, after the Fast Fourier Transform (FFT), i.e. in the two-dimensional correlation E m、k In the results, it constitutes the peak value in frequency +k0 and discrete propagation time m0.

[0407] The second part, p2(n), generates a peak value at frequency -k0 and discrete propagation time m0 under binary phase modulation (phase values ​​of 0° and 180°), because φ TX (n-m0)+φ TX (n-m0) = 2·φ TX (n-m0) can only take two practically identical values, 0° and 360° (the complex pointer formed by passing through them twice is 1). Choosing other phase values ​​will result in a phase jump in p2(n) because 2·φ TX (n-m0) assumes not only integer multiples of 360°. Now, let's assume J = 4 and a phase value φ. j Taking 0, 90°, 180°, and 270° as examples (the modulation sequence b(n) contains 4 values) Then, assume The values ​​are +1 and -1, and there are pseudo-random jumps between them. Since all four phase values ​​are used with the same probability or frequency, The average value is at least very close to zero; at the discrete frequency -k0 of the object, p2(n) has a two-dimensional correlation E m、k The amplitude disappears, so there are no more peak values. Since there is now only one peak value (at the positive frequency +k0), the set receiving frequency is clear and correct.

[0408] The disappearance of the peak value at error frequency -k0 can be observed through each group of phase values ​​that rotate once around point J, i.e., a 360° uniformly distributed phase value.

[0409]

[0410] To achieve this, at least J = 3 and phase value φ are required when the phase values ​​are uniformly distributed over 360°. j = 0, 120°, 240°. For phase values ​​that are not uniformly distributed across 360°, even J = 2 can be used, for example, the phase value φ. j =0, 90° can completely eliminate the peak value at the error frequency -k0. With other phase values, the peak value at frequency -k0 is usually not completely eliminated, but only reduced; when J = 2, the phase value φ j In the example with = 0, 45°, the amplitude decreases by 3 dB. To identify the correct peak value, the larger of the two peak values ​​at ±k0 should be selected; even small differences in the nominal values ​​are sufficient for correct identification, except for the signal-to-noise ratio difference. Generally, among the J phase values ​​φ used... j In a real-value mixer, at least two phase values ​​that are neither out of phase nor out of phase are required to determine the sign of the received frequency.

[0411] Arbitrarily precise phase values ​​are typically not achievable, especially when the phase value is not only out of phase. For example, with a equidistant phase value of J = 4, the considered nominal position is φ. j、nom = 0, 90°, 180°, 270°; the actual value should be φ. j、real = 15°, 75°, 195°, 255°, therefore the intervals have a relatively large error of ±30°. In this case, a phase jump of ±30° will occur in the first component p1(n), resulting in phase instability (i.e., phase jitter). This causes the peak value of the magnitude at the correct frequency +k0 to decrease by only 0.3 dB, thus the sensitivity loss is very small (the lost energy is transferred to small noise at other Doppler frequencies k, but this noise level is much lower than the noise level generated by the phase modulation itself in the two-dimensional correlation); at the erroneous frequency -k0, the peak value is not completely eliminated, but a small DC component in the p2(n) phase causes the peak value to be 11.4 dB smaller than the peak value at the correct frequency +k0, so it can still be correctly identified. From this example, it can be seen that even if there is a relatively large error in achieving the phase value, it does not constitute a serious problem.

[0412] Finally, we discuss possible cost-effective implementation methods. One approach is to switch between line segments of different lengths; Figure 33An example of the J = 3 equidistant phase value is shown, where switching is performed between three line segments of different lengths, where different phases of 120° or 240° are achieved by varying lengths. For the example of the J = 4 equidistant phase value above, such as... Figure 34 As shown, a combination of a switchable inverter 34.1 and a switching switch 34.2 can be used between two line segments with a phase difference of 90°.

[0413] Overall System

[0414] Based on the above design scheme, the lidar system preferably includes only the following three main components:

[0415] - A photonic chip, which includes:

[0416] ● A tunable frequency laser source (preferably only one; multiple sources are only needed when the frequency tuning range is too small).

[0417] ● Phase modulation unit, which includes a switchable inverter,

[0418] ● 32 parallel transmit / receive paths, including optical amplifiers, circulators, real-valued superposition mixers, and photodiodes.

[0419] ● 32 waveguides (or 320 waveguides if a switching matrix is ​​used for scanning in the second spatial direction).

[0420] ● The output has 32 high-MHz frequency band analog receiving signals;

[0421] - A digital chip, which includes:

[0422] ● 32 analog-to-digital converters used to receive the received signal output from the photonic chip.

[0423] ● Fixed wiring calculation logic used to determine two-dimensional correlations and perform subsequent evaluations.

[0424] ● Microcontrollers and / or digital signal processors (DSPs) used for further signal evaluation (especially for determining a detection list) and for calculating laser frequency and second spatial orientation scanner control variables.

[0425] ● For laser frequency control output and for scanning in a second spatial direction (can be analog or digital);

[0426] - A scanner in one spatial direction (for a second spatial direction) is implemented in the following way:

[0427] ● Materials with electro-controllable optical properties, especially liquid crystal devices or liquid crystal arrays, or

[0428] ● A switching matrix for each of the 32 parallel transmit / receive paths, or

[0429] ● For example, mechanical methods in the form of oscillations, rotating mirrors, or prisms.

[0430] ● Some of these methods also require a lens unit and / or a radiation deflection unit.

[0431] In the best-case scenario, all electronic components reside on a single circuit board; however, some arrangements may require two circuit boards. By unlocking the potential of semiconductor integration through the proposed method, costs and structural size can be significantly reduced.

[0432] To reduce the cost of required hardware, the number of parallel transmit / receive paths can be halved by halving the data acquisition time per pixel, thereby halving the size and current consumption of the fixed-route computational logic. The general guiding principle in this scheme is that all components are active (i.e., used) throughout the overall time, and all generated and radiated power is used for detection (i.e., radiating power only in the direction of simultaneous signal reception).

[0433] Alternative phase modulation methods

[0434] To date, Figure 1 In the coherent lidar system shown, the phase modulation has been considered to employ, such as Figure 2 The diagram shows a repeating pseudo-random binary sequence with a period N (i.e., consisting of two phase values, 0° and 180°). Besides pseudo-random sequences, irregular sequences with different definitions can also be used, such as sequences with small autocorrelation sidelobes (e.g., the gold code known in the literature). At least where high sensitivity is required, irregular sequences always need to be specifically adapted to the two-dimensional correlation E according to equation (8). m、k The determination is complex and therefore requires special computational logic, as shown in Figures 8 and 10. If such computational logic is unavailable, and only a digital signal processor (DSP) (especially one with parallel vector computation units) is available, a modulation sequence that can be more easily evaluated should be used (but this may also introduce other disadvantages).

[0435] As an example, Sebastian Banzhaf and Christian Waldschmidt proposed a sequence and evaluation method in their article "Phase-Coded-Based Modulation for Coherent Lidar," published in the October 2021 issue of the IEEE Transactions on Vehicle Technology, Volume 70, Issue 10. For this purpose, a binary modulation sequence b(n) of length N is synthesized from two parts: the first subsequence b1(n) has a length N1, during which the phase is constant, that is, for example...

[0436] b1(n) = 1, where n = 0, ..., N1-1, (69a)

[0437] The second subsequence b2(n) of length N2 = N-N1 is irregular, for example, pseudo-random:

[0438] b2(n) = ±1, where n = N1, ..., N-1. (69b)

[0439] The lengths N1 and N2 can be the same, that is...

[0440] N2 = N1 = N / 2, (69c)

[0441] The modulation sequence can be targeted at multiple detection directions, that is, the pixels are repeated periodically; Figure 35 This type of modulation sequence is shown, in which the sequence synthesized from the subsequence is repeated with a period of N = 4096.

[0442] For the received subsequence e generated by an object i 1、i The method described according to equation (3a) applies from (n) to the first constant modulation subsequence b1(n):

[0443]

[0444] In other words, it only carries the Doppler frequencies of the respective objects, which can be determined by the Discrete Fourier Transform (DFT) or the Fast Fourier Transform (FFT). However, due to the existence of an unknown time shift m... 0、i We do not know e 1、i (n) represents the exact time position of each corresponding element in the received sequence e(n); for simplicity, for example, we can assume that the time shift is zero, i.e., through the received subsequence.

[0445]

[0446] The constructed Fast Fourier Transform (FFT) is as follows:

[0447]

[0448] However, the zero-time-shift assumption leads to the fact that for other actual time shifts... Especially for distant targets, in receiving subsequences The initial m 0、i The -1 value does not contain the value of the constant modulation subsequence b1(n), but rather the value of the latter part of the previous period's modulation subsequence b2(n); therefore, the height of the peak values ​​of each corresponding quantity in the Fast Fourier Transform (FFT) and their distance from the noise are reduced, thus decreasing the sensitivity. Received subsequence Peak value of the Fast Fourier Transform (FFT) Detection occurs at a detection threshold. The frequency k of the peak value J above the detection threshold is... 0、j This will be used for further processing; typically, the peak of this magnitude will be observed in two adjacent Fast Fourier Transform (FFT) values ​​(because they are not at integer Doppler indices k0 as previously considered), and therefore their exact locations, i.e., non-integer frequencies k, can then be determined by interpolation. 0、j These frequencies k 0、j At least approximately corresponding to the Doppler frequency k of the object or a subset thereof 0、i (For objects with extremely low reflectivity, there may be no peak values ​​exceeding the detection threshold.)

[0449] According to equation (3a), the second modulation subsequence b2(n) results in approximately m in the received sequence e(n). 0。i The time shift, and the corresponding Doppler frequencies k 0。i Multiplication, i.e., multiplying by the modulation component.

[0450]

[0451] To eliminate the corresponding modulations at the Doppler frequencies, the received sequence e(n) is switched back to frequency k. 0.j :

[0452]

[0453] Here, it is necessary to consider the range of received sequences e(n) within which the received second modulated subsequence b2(n) can be located, n = N / 2, ..., N-1 + M-1, where M-1 corresponds to the assumed or maximum target distance of interest. It is important to note that the range of received signals n = N, ..., N-1 + M-1 can only be used for continuous scanning, not for switching scans (because the signals are incoherent due to different detection directions).

[0454] The sequence ẽ is corrected in this way 2,j (n) contains the modulated subsequence a shifted by the corresponding index j. i ·b2(nm 0、i ); has other Doppler frequencies k 0、i (i.e., k) 0、i ≠ k 0、j The object magnitude represents a modulation sequence unrelated to b2(n), because they can also be represented by the difference frequency k. 0、i -k 0、j Modulation is performed. Therefore, the sequence It can be correlated with the second modulated subsequence b2(n):

[0455]

[0456] In these one-dimensional correlations 2、j、m In the middle, the peak value appears at position m = m 0、i At that point, i.e., at the discrete propagation time of the object; at the non-integer discrete propagation time m 0、i In general, if appropriate measures are taken (e.g., correcting the modulation pulse to an approximately triangular shape), the magnitude peak will extend to two adjacent values ​​m, and its non-integer position can be determined by interpolation. The correlation of the magnitude peak occurrence... The index j and the Doppler frequency k belonging to j 0、j The Doppler frequency k of the object can be obtained. 0、i = k 0、j Therefore, correlations above a certain detection threshold can be observed. 2、j、m The peak values ​​are determined by the object distance and radial relative velocity in each corresponding detection direction. It is important to note that each corresponding direction has a different Doppler frequency k. 0、i (i.e., k) 0、i ≠ k 0、jThe object values, since they are uncorrelated with b2(n), will only cause noise in the corresponding correlations, that is, there will be no peak values ​​higher than a detection threshold. When the object reflection signal strength is roughly the same, this noise is significantly lower than the peak values ​​of interest, so it will not mask these peak values. Only when the object reflection signals are very different can objects with strong reflection signals mask weak reflection signals with different Doppler frequencies through their noise.

[0457] Use according to Figure 35 The modulation sequence shown and the analysis and evaluation explained above, in the typical case where there is only one object in a pixel, require the receive subsequence e1(n) of the first modulation subsequence b1(n), and the frequency-inverted receive subsequence of the second modulation subsequence b2(n). The correlation between them is subjected to a Fast Fourier Transform (FFT). This correlation in the time range can also be achieved by multiplying the Fast Fourier Transform (FFT) in the frequency range with a subsequent Inverse Fast Fourier Transform (FFT) (the computational workload required for the Inverse Fast Fourier Transform (FFT) is the same as that required for the Fast Fourier Transform (FFT) itself); the Fast Fourier Transform (FFT) of the modulation subsequence b2(n) can be determined a priori in one step, therefore only the received subsequence needs to be determined. The Fast Fourier Transform (FFT) is required. Therefore, a total of three Fast Fourier Transform (FFT) operations are needed, and according to equation (8), if a two-dimensional correlation E is used... m、k This would require calculating a total of M=500 Fast Fourier Transforms (FFTs), where, assuming the pixel duration is the same, the length of these FFTs would be approximately twice the original. This significantly reduces the required computational workload; modern digital signal processors (DSPs) with parallel vector computing units can reduce the computational workload by an order of magnitude.

[0458] and Figure 2 The modulation sequence shown and equation (6) or (8) are described by a two-dimensional correlation E m、k Compared to the analysis and evaluation conducted earlier, the following points are particularly unfavorable:

[0459] - If we assume the pixel duration is the same, the sensitivity will be more than 3 dB lower; half the length of the Fast Fourier Transform (FFT) of the received subsequence e1(n) and (from Half the time correlation length of the modulated subsequence b2(n) results in a 3 dB loss. Furthermore, as mentioned above, there are the following effects, especially for long-range targets, on the received subsequence. The first value does not originate from the constant modulation subsequence b1(n), but from the later value of the modulation subsequence b2(n) of the previous period. Therefore, these values ​​do not actually affect the peak values ​​of the corresponding quantities in the Fast Fourier Transform (FFT).

[0460] Due to the general length of the Fast Fourier Transform (FFT), Doppler measurements, i.e., radial relative velocity measurements, have twice the resolution and accuracy.

[0461] - Used to determine distance The temporal correlation with b2(n) is only half the length. As a result, when there are multiple objects in each pixel, the dynamic range is reduced by 3 dB. That is, the noise level of an object with a strong reflective signal is increased by more than 3 dB compared to an object with a weak reflective signal, so the probability of not detecting such a second object is higher.

[0462] In continuous scanning systems, pixels can overlap in principle, meaning that a portion of the received value e(n) is used for two adjacent pixels, especially to extend the data acquisition time for each pixel, thereby improving sensitivity. However, Figure 35 The modulation sequences shown can only overlap by 50% because each pixel requires two modulation sub-sequences. A smaller overlap, which is typically preferred, is not possible.

[0463] Figure 35 The drawbacks of the modulation sequence shown, and the related analysis and evaluation explained above, are all existing technologies. Figure 36 The novel modulation sequence shown largely eliminates these drawbacks without significantly increasing the computational performance required for analysis and evaluation. Figure 36 In the modulation sequence b(n) shown, the two modulation subsequences b1(n) and b2(n) so far are alternately interleaved:

[0464] b1(n) = 1, where n = 0, 2, 4, ..., N-2, (76a)

[0465] b2(n) = ±1, where n = 1, 3, 5, ..., N-1. (76b)

[0466] Therefore, the two modulation subsequences extend throughout the entire modulation duration, with each subsequence taking only the second grating value. (Index)

[0467] The received subsequence e generated by object i from the first constant modulation subsequence b1(n) 1、i (n) can be an even or odd value of n, depending on the discrete propagation time m of each corresponding object. 0、i Is it even or odd (let's assume m first)? 0、i(If it is an integer):

[0468]

[0469] Within this range of index n, the first value originates from the end of the previous modulation period (because the propagation time of each corresponding object is different) — compared to the previous Figure 35 Compared to the modulation sequence considered in the previous step, these values ​​also originate from the first constant modulation sequence, and therefore they also make coherent contributions to the subsequent Fast Fourier Transform (FFT). Due to the received subsequence e 1、i Since (n) has two possible positions, the Fast Fourier Transform (FFT) used to determine the Doppler frequency of the object must now be computed twice—once through the first sequence.

[0470]

[0471] Another calculation via the second sequence

[0472]

[0473] Two Fast Fourier Transforms (FFTs)

[0474]

[0475] This involves the output dimension, namely discrete frequency. k = 0, ..., N / 2-1, and the discrete frequency k considered so far due to the effective half-scan rate

[0476] k = mod N / 2 (k) (80)

[0477] They are related. Figure 37 In the equation (79), the two Fast Fourier Transforms (FFTs) represent the magnitudes of the two objects. The same received amplitude is present at each location; even discrete propagation time m 0、1 The first object at 300 is at the Doppler frequency. k 0、1 = k 0、1 A peak value is formed in the first Fast Fourier Transform (FFT) at -N / 2 = 1798. Odd discrete propagation time m 0、2 The second object at =101 is at the Doppler frequency. k 0、2 = k 0、2 = 1000, which constitutes a peak value in the second Fast Fourier Transform (FFT). .

[0478] Peak values ​​of two Fast Fourier Transform (FFT) values Perform a detection threshold check. The frequency of peak values ​​J above the detection threshold. k 0、j This is used for further processing; typically, the peak value will be seen in the values ​​of two adjacent Fast Fourier Transform (FFT) results (because, as shown in the example above, they are not in an integer Doppler index). k 0、j (in the middle), therefore, its corresponding accurate position, i.e., a non-integer frequency, can be determined by interpolation. k 0、j Apart from the potentially missing N / 2 part (due to the modulus function in equation (80), these frequencies k 0、j At least approximately corresponding to the Doppler frequency k of the object or a subset thereof 0、i (For objects with extremely low reflectivity, they may not cause peak values ​​exceeding the detection threshold.) Further processing must also take into account... k 0、j The peak value at a given point is detected in which of the two Fast Fourier Transforms (FFTs), i.e., whether the correlated discrete propagation time is even or odd; for this purpose, the following variables are introduced:

[0479]

[0480] Among them, if a peak value appears in the first Fast Fourier Transform (FFT) ,but m j = 0; if a peak value appears in the second Fast Fourier Transform (FFT) In the middle, then m j = 1.

[0481] According to equation (76b), the second modulation subsequence b2(n) results in approximately m in the received sequence e(n). 0、i The time shift is multiplied by the corresponding Doppler frequency k. 0、i That is, multiplied by the modulation component

[0482]

[0483] To eliminate the corresponding modulations at the Doppler frequency, the received sequence e(n) switches the frequency back to approximately k 0. j :

[0484]

[0485] Meanwhile, the variables introduced in equation (81)m j It is considered whether each corresponding received subsequence is in an odd-numbered grid or an even-numbered grid (i.e., has an even-numbered or odd-numbered discrete propagation time); it should also be noted that, as determined in the Fast Fourier Transform (FFT), within the index n range, the first received value (due to the propagation time of each corresponding object) originates from the end of the previous modulation period, but this does not violate the consistency of subsequent correlations, since the entire modulation sequence is periodic (period of N).

[0486] The sequence modified in this way ẽ 2、j (n) Contains a periodically shifting modulation subsequence a at the corresponding index j. i ·b2(mod N (nm 0、i It has other Doppler frequencies. k 0、i (Right now k 0、i ≠ k 0、j The object magnitudes represent modulation sequences unrelated to b2(n) because they are still expressed using Doppler frequencies. k 0、i - k 0、j Modulation; this also applies to those with other m j The magnitudes are because they are modulated by b1(n). Therefore, the sequence ẽ 2、j (n) can be periodically correlated with the second modulation subsequence b2(n):

[0487]

[0488] In these one-dimensional correlations The peak value of the medium value appears at At the location, i.e., the discrete propagation time m of the object. 0、i Subtract the displacement used in equation (83) m j At the location, the discrete propagation time is thus derived accordingly. For the example above with two objects, Figure 38 Two correlation values ​​are shown, one of which is the correlation coefficient. The value of, among which, k 0、1 = 1798 and m 1 = 0, the other is correlation. The value of, among which, k 0、2 = 1000 and m 2 = 1. The peak value is at the numerical value. At their location, they are related to discrete propagation time Correspondingly.

[0489] Correlation of the occurrence of peak values Index J and the Doppler frequencies belonging to index j k 0、j The Doppler frequency k of the object can be obtained. 0、i The maximum possible offset is one potential shift of N / 2, see equation (80), which takes into account that the Doppler frequency is determined by a sequence at half the sampling rate, meaning that additional half-cycles based on the full sampling rate are unrecognizable. However, such half-cycles cause the complex values ​​in the Fast Fourier Transform (FFT) to peak at their respective magnitudes. They are rotated 180° relative to each other, while in other cases (i.e., without an extra half-cycle in the Doppler frequencies), they have the same phase. This correlation allows us to resolve the relationship with respect to each corresponding Doppler frequency k. 0、i The ambiguity is addressed by adding two complex values ​​corrected for possible phase shifts (0° and 180°), i.e., the sum and difference of the two complex values—if the sum is greater than the difference, then the Doppler frequency k is determined. 0、i = k 0、j In other cases, then k 0、i = k 0、j +N / 2; In the example above, for the first object (i.e. for...) k = k 0、1 and (Because the two components are approximately out of phase), the Doppler frequency is k. 0、1 = k 0、1 +N / 2 = 3846, while for the second object (i.e. for...) (Because the two components are approximately in phase), the Doppler frequency is k. 0、2 = k 0、2 = 1000. Since the Doppler frequency is determined by the duration of the entire modulation period, therefore... Figure 35 The modulation sequence shown (according to the prior art) no longer suffers from the drawback that the resolution and accuracy of Doppler determination are reduced by a factor of 2 because the Fast Fourier Transform (FFT) is determined by half the duration of the modulation.

[0490] Using the relationships and operational steps described above, correlations above the detection threshold can be detected. 2、j、 m The peak value determines the distance and radial relative velocity of the object in each corresponding detection direction. It is important to note that each value also has corresponding other Doppler frequencies. k 0、i (Right now k 0、i ≠ k 0、j ) and / or other values m j The object's magnitude, due to its lack of correlation with b2(n), will only cause noise in the corresponding correlations, that is, it will not cause a magnitude peak higher than the detection threshold. When the intensity of the object's reflected signal is roughly the same, this noise is significantly lower than the magnitude peak of interest, and therefore will not be masked by it. Only when the object's reflected signal is very different can the object with the strong reflected signal mask the weak reflected signal with different Doppler frequencies through its noise.

[0491] Correlation from above a detection threshold An alternative to peak value generation and detection methods can be the Fast Fourier Transform (FFT) described above. The sum and difference, where, k = k 0、j and correlation The detection threshold is checked because the sum and difference of the two Fast Fourier Transform (FFT) components and the correlation are added together when the phase is correct, i.e., coherent addition, and their signal-to-noise ratio is about 3 dB better than that of the correlation itself; this is because the sum or difference is integrated over the entire modulation sequence b(n), but in the correlation, it is integrated over only half of the modulation sequence. Thus, the sum or difference also has the same signal-to-noise ratio, and therefore is comparable to the optimal two-dimensional correlation E according to equation (6). m、k They have the same sensitivity. However, this only applies at the respective locations, i.e., at... k = k 0、j A peak value has been detected in the Fast Fourier Transform (FFT) at the location; due to the halved integration duration, the signal-to-noise ratio of the FFT is 3 dB worse, therefore a correspondingly reduced detection threshold must be used. This reduced detection threshold increases the likelihood of detecting spurious noise peaks; however, considering the more clearly defined effective detection threshold for the sum and difference of the FFT and correlation, and the subsequent discarding of spurious noise, the only remaining drawback is a slight increase in computational workload (because the correlation must be calculated more frequently). It should also be noted that it is not necessary to specify each one. The sum and difference, but correlation can be checked first. The detection threshold decreased by approximately 3 dB, and the sum and difference were only calculated at locations exceeding that detection threshold. Because Figure 36 The modulation sequence shown and the new method constituted by the above analysis and evaluation have the best two-dimensional correlation E according to equation (6). m、k With the same sensitivity, therefore, it was eliminated. Figure 35 The modulation sequence shown (according to the prior art) suffers from a sensitivity variation of approximately 3 dB. A brief discussion should now be given regarding whether a new method, based on Fast Fourier Transform (FFT) and the sum and difference of correlations, can also be applied to [the modulation sequence]. Figure 35 The modulation sequence shown is only possible if the received phase remains stable and linear (i.e., the instantaneous frequency is constant) throughout the modulation sequence, but only to a limited extent, especially due to the speckle effect.

[0492] Fast Fourier Transform (FFT) and correlation summation and difference methods also avoid... Figure 35 (Prior art) shows that in the case of more than one object per pixel, the dynamic range of the modulation sequence is reduced by about 3 dB because it is now integrated through the entire modulation sequence.

[0493] Figure 36 Another advantage of the new modulation sequence shown is that, since each part of the periodic modulation sequence (period N) has the properties described in equation (76), arbitrary overlap between two adjacent pixels can be achieved.

[0494] use Figure 36 The new method and application of the modulation sequence shown Figure 35 The difference in computational workload between the modulation sequence methods shown (in the prior art) lies in the fact that the former requires an additional Fast Fourier Transform (FFT) of length N / 2 (independent of the number of objects in each corresponding pixel), while the correlation length is only half that. However, this is not advantageous when implementing correlations across the frequency range (i.e., via both FFT and inverse FFT). In the typical case of only one object in a pixel, two FFTs and one correlation must be determined. In principle, additional correlation calculations can also be performed in the aforementioned methods for lowering the FFT detection threshold. Thus, the required computational workload remains within the order of magnitude achievable by modern digital signal processors (DSPs) with parallel vector computing units, allowing for economical implementation without requiring special computational logic.

[0495] For the output dimension of the Fast Fourier Transform (FFT), namely the discrete frequencies, by usual convention, the above (in this section) takes into account... kThe asymmetric range of k = 0, ..., N / 2-1 is also the same for discrete frequencies after the ambiguity of the asymmetric range k = 0, ..., N-1 has been resolved; the actual relative velocity and Doppler frequency assumptions can have two signs, so the upper part, especially the upper half of k = 0, ..., N-1, is expressed as a negative value by subtracting N.

[0496] As of now, Figure 36 The alternative to the new modulation sequence shown can also be as follows: Figure 39 As shown, alternating values ​​(i.e., alternating +1 and -1) are used for the modulated subsequence b1(n). The only significant difference is that the received subsequence of b1(n) has an additional frequency with a period of 2 (related to the modulation rate of b1(n), i.e., the peak values ​​of the two correlated Fast Fourier Transform (FFT) values). 1、1、 k With Ẽ 1、2、 k The Fast Fourier Transform (FFT) length is shifted by half, N / 4, which must be corrected accordingly during further processing.

[0497] So far, the case considered in this section is the discrete propagation time m. 0、i It is an integer, the modulation duration T m Equivalent to sampling repetition time T s For a non-integer discrete propagation time m 0、i In the case of an ideal rectangular modulated signal where the received signal also maintains its ideal shape, sampling may occur precisely at the edges that do not contain meaningful information. To avoid this, the sampling repetition time T of the received sequence can be increased. s Set it to a smaller value beforehand, for example, set it to the modulation duration T. m Half of the modulated pulse shape is modified, and / or the shape is corrected directly during its generation or in the receiver to, for example, an approximate triangular shape (the latter is achieved through low-pass filtering in the receiver). This is done using modulation signal correction methods and non-integer discrete propagation time m. 0、i In general, peak values ​​will appear in both Fast Fourier Transform (FFT) results. Therefore, for two m j = 0, 1 determine correlation And the non-integer m is determined by numerical interpolation of the peak values ​​of the two correlation quantities. 0、i (non-integer m) 0、i It can also be determined by interpolating or combining the peak values ​​of the two Fast Fourier Transform (FFT) values, i.e., by combining the sum or difference of the FFT values ​​with the correlation. (In relation to the modulation duration T)m Compared to) sampling time T s For example, in the half-method, the interleaving of the two modulation subsequences has a period of 4 relative to the sampling time, and in the received signal, two consecutive sampled values ​​each originate from the first sequence b1(n), and the next two sampled values ​​each originate from the second sequence b2(n); therefore, on the one hand, for the received subsequence e 1、i For (n), four possible positions must be considered, meaning four Fast Fourier Transforms (FFTs) must be computed. On the other hand, the sampled values ​​corresponding to b2(n) are set to zero here (because they are no longer in the grid of period 2, and therefore can no longer be simply omitted), thus (in contrast to so far using half of the received values), the Fast Fourier Transform (FFT) can be calculated from all the received values. Then, in at least a portion of the Fast Fourier Transform (FFT), a single object will exhibit three magnitude peaks; the maximum value at the correct location, i.e., the Doppler frequency of the object (unlike above, there is no longer ambiguity, which is an advantage), and two other magnitude peaks of about 3 dB before and after a quarter of the length of the Fast Fourier Transform (FFT) (these should be ignored for further processing). Since four magnitude peaks appear in the Fast Fourier Transform (FFT). A non-integer discrete propagation time can be determined by interpolation (using Fast Fourier Transform (FFT) numerical values ​​or correlations or a combination thereof).

[0498] In addition to the two alternating cases of modulated subsequences b1(n) and b2(n) considered so far, namely the interleaving case with a period of 2, longer periods can also be used for interleaving, and alternatively, the number of b1(n) and b2(n) elements in each period can also be unequal.

[0499] The points considered in this section so far, such as Figure 1 In the lidar system shown, the mixer is complex-valued, which significantly increases the workload (almost double) compared to the real-value mixer in the receiving path. When using a real-valued mixer, only the magnitude of the relative velocity can be determined, not its sign, because in the Fast Fourier Transform (FFT), +... k 0 and - k There are two peak values ​​at position 0. To determine the sign, a confidence check and / or tracking method is required, i.e., tracking over multiple acquisition periods. Alternatively, as explained above, a complex-valued modulation sequence can be used, such as... (For 4 equidistant phase values ​​φ) jThe sequence consists of four values: 0, 90°, 180°, and 270°. For the first modulation subsequence b1(n), the four values ​​can be... The system rotates periodically, so that in objects with zero relative velocity, the peak value is located at one-quarter of the Fast Fourier Transform (FFT) length—negative relative velocities are below (i.e., lower frequencies), and positive relative velocities are above. This first modulation subsequence b1(n) with a period of 4 continues to alternate with the second modulation subsequence b2(n), which can be assumed, for example, to be four values ​​in a pseudo-random sequence. Although generating such complex-valued sequences requires increased circuitry work (see, for example...), Figure 34 However, this only needs to be done once, while the workload for a complex-valued receiver is many times greater for a parallel receiver. In a complex-valued modulation sequence, the conjugate complex value of the modulation sequence must be used for correlation. Alternatively, the conjugate complex value of the received sequence can also be used, as long as it is complex-valued.

[0500] So far, the Fast Fourier Transform (FFT) has been considered without using window functions, meaning the FFT input values ​​are not multiplied by a bell curve. This approach is only necessary or meaningful when two objects with similar relative velocities but significantly different reflectivity appear at the same distance within a pixel and need to be separated. In particular, if no window function is used at the FFT input, the sensitivity of the FFT output will decrease if the Doppler index corresponding to the relative velocity is not an integer—that is, the peak value is split by two adjacent FFT values. This reduces the ability to detect objects with weak reflectivity and at greater distances. This effect can be mitigated by choosing an FFT length longer than the input signal, i.e., adding zeros to the input signal, a process known as zero-padding.

[0501] Regarding the second modulator subsequence b2(n), it should also be noted that it can be composed not only of pseudo-random switching between discrete phase values, but also of an autocorrelation code with low sidelobes—for example, a gold code known in the literature. This results in a higher correlation dynamic range, provided that the correlated signal does not contain signal components representing noise (such as signal components of objects with different relative velocities).

[0502] The modulation sequence described in this section consists of two sub-sequences, which, according to the prior art, are arranged sequentially or nested in a new method. According to the present invention, they can also be used in conjunction with the linearly varying frequency in equation (9). Figure 40 This is an example of this. The only effect is, besides the Doppler frequency shift f... D In addition, the receiving frequency fe There is also a component related to the transmission time, f. r (See Equation (12)); therefore, all the above considerations still apply, the only thing to consider is that, when determining the relative velocity, the component dependent on the propagation time must be subtracted from the receiving frequency, where the propagation time can be obtained from the correlation. Superimposed frequency modulation makes possible the advantages of the above-described system method (especially with continuous spatial scanning of frequencies) and (especially in determining the sign of the receiving frequency in a real-value mixer). The methods described in the preceding sections are also applicable to this phase modulation, or can be applied to this phase modulation.

[0503] It should also be noted that for the identification of road surfaces at long distances, the advantage of a constant or periodic subsequence b1(n) is that it does not segment the received signal across discrete distances, thus the entire radiation width can be effectively utilized, and the signal-to-noise ratio is thus greatly improved, resulting in a high probability of detecting the road surface in a single pixel.

[0504] Functional check of the scan

[0505] If, due to hardware malfunctions or other reasons, scanning in one or two spatial directions fails to proceed normally, high energy densities will appear in certain radiation directions (because the system's dwell frequency in these areas is significantly higher than normal), potentially exceeding the permissible limits for eye safety. Therefore, scanning must be monitored, and the LiDAR sensor's radiation must be stopped immediately upon malfunction.

[0506] If the sensor stops scanning in one spatial direction, the received signal of all pixels in that spatial direction remains unchanged regardless of other spatial directions, except for system noise and moving targets. Therefore, it is necessary to examine the changes in the received signal in both spatial directions. A first method is to compare object reflections, i.e., the peak value exceeding the detection threshold; as mentioned above, all moving objects must be excluded here, which can be identified by the measured relative velocity and the known self-velocity. A second method uses the received signal components of internal reflections, coupling, and overlay reflections at approximately zero distance; as mentioned above, these components are used to compensate for correlation effects (via the correction value c1(n) in the fixed wiring calculation logic shown in Figures 8 and 10). What is utilized here is the characteristic that these received components typically change with scanning.

[0507] A radiating waveguide array achieves scanning in two spatial directions through frequency variation.

[0508] In the arrangements considered so far for scanning and focusing in two spatial directions, one spatial direction is achieved through one or more waveguides with coupling points, while the second spatial direction is achieved through other methods, requiring at least one optical element (e.g., a lens and / or a liquid crystal element) attached to the photonic chip. To achieve the simplest and smallest possible lidar system, a method eliminating the need for additional optical elements should be pursued, where scanning and focusing in both spatial directions are achieved on the chip itself, allowing the chip to emit and receive one or more focused and deflected radiations through a transparent area in the sensor housing. As a first example, a sensor for short and medium range applications will be considered, with a pixel size and radiation width of approximately 0.5° × 0.5°, using only a single transceiver channel.

[0509] Figure 41 One possible method is illustrated, employing an array of approximately 700 radiating waveguides 41.1 for both transmission and reception (as described in the examples above, a monostatic system is still considered here); for clarity, not all waveguides are shown here, as in all subsequent images. All strip-shaped array waveguides are straight and have the same shape—specifically, they have the same type and location of radiating coupling points, i.e., the same so-called gratings, with the distance between coupling points approximately 600 nanometers; the waveguides are arranged side-by-side in parallel and at equal intervals with a spacing of approximately 700 nm. The radiation of the waveguide is focused in a first spatial direction defined by it and scanned in that spatial direction by frequency variations—for example, -25°…+25°; for this purpose, the frequency f(t) needs to be varied by about ±10% relative to its average value f0 = c / λ0 = 194 THz, where the free space wavelength λ0 = 1550 nm (in addition to the effective refractive index of about 2.5, the dispersion characteristics of the waveguide, i.e., the frequency dependence of its effective refractive index, which has a small width due to the small spacing between waveguides, also need to be considered).

[0510] 700 radiating waveguides 41.1 are fed into a long, meandering waveguide 41.3 via connecting waveguides 41.2 of equal length, where the feed points 41.4 of every two adjacent waveguides maintain the same path distance of 160 micrometers. Therefore, frequency variations cause the phase of the feed signal along the waveguide to change linearly; this linear phase change along the waveguide also applies to the first coupling point and every subsequent coupling point, thus enabling scanning in a second spatial direction perpendicular to the first spatial direction. Since the path distance between the feed points of these waveguides is approximately 230 times longer than their spacing, the scanning effect produced by frequency variations in the second spatial direction is much stronger; a full-range scan from -90° to +90° is achieved in the second spatial direction when the radiation shifts by only about 0.5° of pixel width in the first spatial direction. Figure 42 The scanning pattern is illustrated, showing how radiation moves in two spatial directions by varying its frequency over time (decreasing in this case, from high to low frequency). As the radiation slowly moves from -25° to +25° in the first spatial direction, it rapidly and repeatedly traverses a complete scan from -90° to +90° in the second spatial direction (not all scan trajectories are shown for simplicity), jumping back in between. This requires a gradually increasing dead time as the frequency decreases—during which, because the waveguide spacing is less than the wavelength in half-free space, the radiation will be in a virtual angular region outside ±90°. This continuous frequency variation is achieved in such a way that a constant scanning speed is produced in the first spatial direction (the required frequency variation curve f(t) is not perfectly linear, but slightly curved due to dispersion effects, trigonometric relationships, and the relationship between relative frequency changes)—therefore, the angular axis in the first spatial direction also represents a linear time axis. Due to trigonometric relationships, the scanning speed in the second spatial direction increases sharply at outer angles close to ±90° (and the pixel width also increases accordingly), but this is irrelevant from a functional point of view because the required resolution is also lower at these angles (in addition, the radiation width at these angles also increases to the same extent due to the corresponding reduction in the effective aperture, i.e., the waveguide array extension size visible from these angles).

[0511] If the distance between waveguides is chosen to be larger—for example, equal to half the central free-space wavelength λ0 = 1550 nm—the resulting scanning mode is as follows: Figure 43 As shown, when the wavelength of the frequency is less than twice the waveguide spacing, scanning in the second spatial direction will overlap, resulting in ambiguity at large angles in the second spatial direction. However, simultaneously, the frequency and scanning range in the first spatial direction will decrease, and within this range, the dead-time effect that occurs during scan bounce (this effect only occurs when the wavelength is greater than twice the waveguide spacing) will occur. This can be a meaningful method, especially when the ambiguity at these large angles in the second spatial direction is functionally irrelevant. Figure 42 and Figure 43 The scanning mode shows that in the first spatial direction, the effective pixel width increases slightly from left to right, so it is not a completely constant 0.5°. This is because a complete scan in the second spatial direction (i.e., from -90° to +90°) always corresponds to a 0.5° change in radiation direction in the first spatial direction. However, due to the overlap of the scanning areas and dead time in the second spatial direction, the effective pixel width in the first spatial direction may be less than or greater than 0.5°. Furthermore, in reality, there are also... Figure 42 and Figure 43The effect not shown, namely, that the radiation angle and the phase difference between the two coupling points are sinusoidal and therefore non-constant (see the derivation of formula (55)), causes the pixel width in the first spatial direction to be stretched at angles with larger absolute values ​​(for a scan range of -25° to +25°, the stretching is about 10%).

[0512] The distance between the radiating waveguides in the array is only about half a wavelength; therefore, when implemented on a photonic chip, it is impossible to completely avoid a certain degree of coupling between them (even with appropriate countermeasures, such as silicon-based isolation strips). This will cause a slightly tilted wavefield to form in the waveguide array (i.e., across multiple waveguides) when the radiation angle in the second spatial direction is not 0°, meaning it tends to shift slightly laterally, resulting in interference reflection effects at the outer waveguides, and consequently, especially in the second spatial direction, leading to radiation broadening and sidelobes. Therefore, this invention proposes adding additional waveguides 41.5 to the sides of the waveguide array. These waveguides are not connected to the feed waveguide and are called blind waveguides; this allows the wavefront to shift laterally in the waveguide array—the only, and non-critical, effect is a slight change in the two-dimensional radiation shape (because the radiating surface is no longer a precise square, but a slight trapezoid). In addition to using blind waveguides, the radiation from the edge waveguides in the array can, in principle, be shielded, i.e., absorbed; however, this is difficult to implement (it is necessary to avoid disrupting the consistency of waveguide characteristics) and also results in power loss. It should also be noted that if coupling causes frequency- and angle-related distortions in radiation in one or two spatial directions, the design should ensure that these distortions have minimal impact in the main radiation direction (e.g., 0°) or compensate for them through appropriate measures.

[0513] Figure 41 The meandering extension of the feed waveguide 41.3 shown is difficult to achieve because the width of the waveguide and the gap between adjacent segments must be very small, requiring very tight bends; furthermore, this would lead to high coupling between adjacent waveguide segments. In addition, the small waveguide and tight bends would result in considerably high transmission loss. This problem can be addressed by… Figure 44 The arrangement shown is improved; the feed waveguide 44.3 is extended, thus increasing its width and making the bends less compact, and the coupling effect is correspondingly weakened due to the increased distance between adjacent waveguide segments. Since the feed waveguide now has a larger extension range than the waveguide array composed of the radiating waveguides 44.1, it is spaced apart from the waveguide array to allow sufficient space for the connecting waveguides 44.2. The meandering feed waveguide 44.3 is also bent to ensure all connecting waveguides have the same length.

[0514] However, compared to Figure 41 , Figure 44The arrangement loses its inherent geometric consistency—in Figure 41 In the system, all connecting waveguides automatically have the same length through the system's geometric arrangement, while... Figure 44 This requires precise design to achieve; the same applies to meandering feed waveguides, as they are now curved (not precisely located on a circular arc). Figure 45 In the alternative arrangement shown, inherent geometric consistency is restored by stretching the feed waveguide: the meandering feed waveguide becomes straight again and maintains a constant shape throughout its extension; all connecting waveguides have the same 90° bend, and their lengths now increase linearly (i.e., no longer constant) for geometric reasons, because the two straight waveguide segments have linear length increases due to the arrangement. The linear length increase of the connecting waveguides is superimposed on the linear length increase of the feed waveguide to its feed point, thus maintaining the linear length change required to reach the radiating waveguide (and the length of the meandering feed waveguide is correspondingly shortened to achieve the same total length change). It is worth mentioning that this arrangement also has advantages in terms of coupling between waveguides 45.2; the straight waveguide section 45.6 between the feed waveguide and the 90° bend has a large distance and therefore almost no coupling, while the straight waveguide section 45.7 after the 90° bend has coupling due to the smaller distance, but since its structure is basically the same and there are blind waveguides 45.5 (now more are needed above than below), it will basically only cause a slight lateral shift in the wave field.

[0515] Figure 45 This illustrates one possible implementation of connecting feed waveguide 45.3 to connecting waveguide 45.2 via coupling point 45.4; the two waveguides extend in parallel for a short distance with a small spacing—the coupling can be further enhanced by perturbation points. Figure 45 (Not shown). If the lengths of these coupling points and the waveguide spacing remain constant across all feed sections 45.4, the coupling strength will decrease exponentially along the feed waveguide. To counteract this effect and achieve amplitude arrangement (see subsequent paragraphs), the coupling points are designed in different forms—if the waveguide spacing of the coupling remains constant, the length of the coupling point will vary (e.g., ...). Figure 45 (As shown).

[0516] To ensure the generated radiation has the sharpest possible shape and lowest possible sidelobes in both spatial directions, appropriate amplitude and / or phase arrangement (referred to as "tapering") is required. For the first spatial direction, not all coupling points of the arrayed waveguides radiate with the same intensity; rather, the outer coupling points tend to radiate less intensely than the inner ones (achieved through appropriate design of the coupling points), and, if necessary, slightly deviate from strictly linear phase characteristics between the radiators (achieved through non-perfectly equidistant radiator arrangement). The same applies to the second spatial direction, achieved through appropriate design or the position of the coupling points between the feed waveguide and the connecting waveguide, where the distance between the radiating array waveguides can also be slightly adjusted for phase. Undesirable sidelobes may appear in the radiation characteristics when amplitude and / or phase errors exist in the distribution (e.g., failure to achieve the expected linear phase characteristics due to geometric tolerances); if the sidelobes are known, they can be partially compensated for through digital signal processing (e.g., using so-called relaxation algorithms). Regarding sidelobes, it is advantageous to use a monostatic system (i.e., using the same array for both transmission and reception), as the sidelobe effect can be attenuated quadratically (i.e., on a logarithmic scale, this is equivalent to doubling the sidelobe level difference).

[0517] exist Figure 45 In the meandering feed waveguide 45.3, the axis extends in the same direction as the radiating waveguide 45.1, while... Figure 46 This shows the axis of the tilt direction of the feed waveguide 46.3, which eliminates the need to connect the first straight section of waveguide 46.4.

[0518] Furthermore, another drawback of these arrangements is the presence of numerous 180° bends, which lead to relatively high losses. By adopting... Figure 47 The arrangement shown reduces the number of bends the waveguides need to traverse. The feed waveguide 47.3 has only a quarter of the original number of bends because within one bend cycle, it branches off to four connecting waveguides 47.2, which in turn feed the corresponding radiating waveguides 47.1; the positions of the coupling points 47.4 are chosen such that the path length for wave propagation to each radiating waveguide 47.1 in the array increases linearly. In principle, all waveguides (radiating waveguides 47.1, connecting waveguides 47.2, and the common feed waveguide 47.3 in the array) can have the same width (and the same cross-section), thus having the same effective refractive index and dispersion characteristics. Figure 47 As shown, to reduce transmission loss, a wider feed waveguide can be advantageously used (which typically reduces loss). Therefore, as... Figure 47As shown, this connecting waveguide segment 47.6 should also have this greater width, so that the wave path in these wider waveguides increases linearly with the radiating waveguides in the array. The wave path in the corresponding, narrower connecting waveguide segment 47.7 also increases linearly with the radiating waveguides in the array. A transition region 47.8 is provided in all connecting waveguides from the wider to the narrower waveguide width (designed so that the wave mode can be preserved). In this design, despite the different effective refractive indices and dispersion characteristics of the two waveguide cross-sections, the signal phase is ensured to change linearly along the radiating waveguide at all frequencies, thus ensuring correct focusing. The different dispersion characteristics of the two waveguide cross-sections (more significant when the waveguide width is smaller) also slightly alter the curve of the scan rate ratio in the two spatial directions (this ratio is typically not constant relative to the same angle in the second spatial direction); in this way, the slightly different pixel distances in the first spatial direction mentioned above (according to...) can be offset. Figure 42 and 43 (The scanning mode shown) - where possible, even overcompensation can be achieved.

[0519] It is worth mentioning that the feed waveguide can also be used in other locations or orientations; for example, it can be as follows: Figure 44 The waveguide array is arranged in a curved shape in front of it, but due to the lack of inherent geometric consistency, it is difficult to design precisely.

[0520] Figure 47 In this context, the connecting waveguide, after coupling from the common feed waveguide, directly forms a region with a relatively wide waveguide width of 47.6; these regions can also be located away from the coupling point, and thus away from the region of the meandering feed waveguide, particularly to provide more space for the feed waveguide, thereby further increasing its width. Figure 47 It was not drawn to scale; in reality, the geometry of the feed waveguide region is more... Figure 47 (The density shown is even higher).

[0521] exist Figure 47 In a meandering feed waveguide, each cycle is coupled to only four radiating waveguides; more couplings are typically implemented to provide more space for the bends and width of the common feed waveguide.

[0522] The required radius of the bend increases with the width of the waveguide; otherwise, it would result in higher losses at the bend. Therefore, it is advantageous to use a narrower waveguide width for the bend of the feed waveguide compared to the straight section; however, the transition between the narrower and wider waveguide widths still needs to be designed to have only low losses (and without partial mode conversion). These transitions and bends are advantageously designed to be thermally insulating.

[0523] In each bend cycle of the feed waveguide, the number of traversing bends (approaching 180°) increases by two, while the number of transition sections between the narrower and wider waveguide widths increases by four. Using... Figure 47 In the arrangement shown, the propagation time of the wave in these bends and transitions must be precisely known so that when designing the coupling points between the feed waveguide and the connecting waveguide at the junctions of the bend periods (i.e., the locations of every two additional bends and four transitions), phase discontinuities (i.e., phase errors in the radiating waveguide signal) can be avoided. This must hold true at all frequencies; it cannot be achieved if the transitions between different waveguide widths have different dispersion characteristics than the connecting waveguide with a constant width. Therefore, it is also meaningful to achieve the inherent geometric consistency of these bends and transitions. To this end, such bends and transitions need to be inserted into the connecting waveguide, where their number decreases by two and four respectively in each bend period. Thus, the connecting waveguide in the nth bend period has 2·(Nn) bends and 4·(Nn) transitions (where the index of the first bend period from the feed end is n=1, and there are a total of N bend periods). Since the propagation time of the bends is known with considerable precision, and realizing 2n bends in the connecting waveguide requires significant space, it may be necessary to abandon the implementation of these 2n bends in the connecting waveguide and instead use a straight waveguide segment of corresponding length and narrow width.

[0524] Since the connecting waveguide from the front bending period requires the most additional structure (bending and / or transition sections), it is advantageous for it to have a longer length than the rear section. This can be achieved by... Figure 47 This is achieved by feeding the waveguide from another direction (i.e., from...). Figure 47 (Feeding is performed from above the view shown). This also has the advantage that the long connecting waveguides with higher attenuation (especially in sections with narrow waveguide widths) will no longer have additional attenuation due to the long path in the feed waveguide, which helps to achieve proper amplitude taper and reduce total loss.

[0525] exist Figure 47 In this approach, a meandering feed waveguide period directly feeds multiple connecting waveguides to the radiating waveguides. Alternatively, each period can be coupled to only one waveguide at a single location, which itself has a meandering path, and thus feeds from, for example, four connecting waveguides to the radiating waveguides; in this case, there is one primary feed waveguide and multiple secondary feed waveguides. However, such approaches may be more demanding in terms of design and robustness; particularly at the coupling points of the secondary waveguides, where much stronger percentage-based coupling is required.

[0526] The common feed waveguide is approximately 11 cm long in all arrangements. Provided the waveguide cross-section is not too small, it can be implemented in a photonic chip, for example using a silicon oxide substrate or silicon material coated with silicon oxide, with a transmission loss of <0.1 dB / cm (this is pure waveguide loss, excluding power coupling loss). Therefore, this waveguide length only introduces extremely low and acceptable losses (≤0.5 dB per half-waveguide length, and bends may introduce similar orders of magnitude of additional losses).

[0527] As mentioned above, the frequency must be changed by approximately ±10% (i.e., a total change of 20%) to achieve the desired scanning range. A laser with a 20% tuning range is difficult to achieve. Therefore, multiple lasers with different frequency ranges can be used to feed an array (for simplicity, "array" in this and the following two paragraphs refers to the overall arrangement including the radiating array waveguide, connecting waveguides, and common feed waveguide, also referred to as the total array below). Alternatively, multiple arrays with different structures fed by the same laser can be implemented; these structures are designed such that they correspond to different scanning ranges in the first spatial direction within the available frequency range, and preferably are seamlessly connected to each other. In the example of the ±25° detection range and two arrays above, the first array covers the range of -25°…0°, and the second array covers the range of 0°…25°, which reduces the laser's tuning range to only approximately ±5%. Due to the reduced frequency range, the array design is simplified (e.g., in terms of frequency dependence at the coupling point and effects caused by refractive index), and the pixel width variation in the first spatial direction is smaller. Multiple arrays can operate in serial or parallel modes. In serial operation, only one transmit / receive channel is required, connected to each array sequentially via a multiplexer or switchable amplifier. In parallel operation, each array has an independent transmit / receive path (all fed by the same modulated laser signal), thereby enabling higher pixel counts, longer single-pixel integration times, or shorter cycle times. The arrays are preferably arranged relative to each other to minimize the total radiation cone angle at the sensor surface or its vehicle side panel; for example, an upward-radiating array is positioned below in this arrangement.

[0528] Using multiple arrays can also shorten the feed waveguide length, again assuming the frequency can be tuned across the entire range of approximately ±10%. For example, if the feed waveguide is only one-third of its original length (and thus the distance between the feed points of adjacent radiating waveguides is also proportionally reduced), the scan range in the first spatial direction will be expanded to three times, or approximately 1.5°, during a full scan in the second spatial direction. If three arrays are used, with slightly different distances between their radiating waveguide coupling points, resulting in each array having approximately 0.5° different radiation directions in the first spatial direction at the same frequency, then the following pattern is formed: Figure 48The scanning patterns shown involve three single arrays nested together to achieve the same result as... Figure 42 The same scan density as the original scan mode. It should also be noted that if a sufficient number of arrays are used in this method, a completely straight feed waveguide can be achieved; in Figure 45 and 46 Taking the original configuration as an example, the feed waveguide will no longer be meandering, but straight.

[0529] So far, the consideration has been a single static operation mode, where a single array is used for both transmission and reception. This requires a transmit / receive duplexer, which, to avoid expensive and non-photonically integrated circulators, is typically implemented using a ring coupler, incurring a 3 dB loss in both transmission and reception. To mitigate this loss, a bistatic scheme can be used, employing two independent, identical arrays for transmission and reception, respectively. These two arrays should be placed as close as possible to achieve near-field detection (a larger radiation bandwidth is also beneficial here).

[0530] In the arrangements to date, radiation is achieved through an array of multiple parallel strip-shaped waveguides with coupling points. As an alternative, such as Figure 49 As shown, a single wide waveguide, i.e., waveguide surface 49.1 (shaded), can be used. When the radiation angle in the second spatial direction is not 0°, the phase at the input end of this waveguide surface (i.e., the interface side connecting the waveguide) changes linearly, thereby propagating a tilted wavefront within the waveguide surface, and thus generating a corresponding tilted wavefield 49.5. Figure 49 (Marked by dashed lines). Therefore, the waveguide surface needs to gradually widen as its distance from the connecting waveguide interface increases, i.e., it needs to unfold (although it can also maintain the widened width over its entire length). At this point, it is no longer a single coupling point, but rather parallel and equidistant coupling lines 49.6, which run through the entire surface and are arranged perpendicular to the central axis of the waveguide surface. It is worth mentioning that, due to the presence of the tilted trapezoidal wavefield, the two-dimensional radiation shape will change slightly with the change of the radiation angle in the second spatial direction.

[0531] Scanning is achieved by varying the effective refractive index of the waveguide.

[0532] In all the methods and arrangements considered so far, the radiation direction of the waveguide has been achieved through frequency variation. Besides adjusting the radiation direction by changing the frequency, it can also be achieved by changing the effective refractive index—since this also changes the number of wavelengths (usually a non-integer), thus altering the phase relationship between the two coupling points. Several methods exist for changing the effective refractive index. For example, a slight effect on the refractive index can be achieved by applying a lateral voltage to a silicon-based waveguide (as used in photonic chips); however, since this effect is very weak, very long waveguides are required, which is disadvantageous in terms of space occupation and high transmission loss. Another approach is to use liquid crystal materials to realize the waveguide, as they can produce a stronger refractive index change by applying a voltage. A third approach will be considered next: liquid crystal materials will be used around the silicon-based waveguide, such as… Figure 50 The example is shown and can be implemented on a photonic chip. Figure 50 The left side shows a cross-section of the waveguide perpendicular to its extension direction: the waveguide 50.1 is surrounded by silicon dioxide 50.2 on three sides (bottom, right and left); liquid crystal material 50.3 is arranged above the waveguide, and both sides of the liquid crystal material are covered with a thin conductive layer 50.4 that is transparent to the laser frequency used. Figure 50 The right side shows a cross-section along the waveguide's extension direction: Perturbation points 50.5 are set at regular intervals at the bottom of the waveguide as coupling points, meaning that during transmission, a portion of the wave 50.6 propagating in the waveguide is coupled outwards through these points and coupled inwards during reception; these coupled input / output waves 50.7 pass through the optically transparent liquid crystal material 50.3 and the transparent conductive layer 50.4 into the surrounding environment. The waveguide not only contains a field within itself, but this field also permeates into the surrounding environment to some extent—this field component is called the evanescent field. Therefore, the evanescent field also exists in the liquid crystal material above the waveguide. The optical properties of the liquid crystal material can be changed by applying a voltage u(t), thereby altering its effect on the evanescent field within it. The effective refractive index of the waveguide is related to the evanescent field, and consequently to the material properties through which the evanescent field propagates. Therefore, for Figure 50The waveguide shown can have its effective refractive index altered by applying a voltage u(t) to the liquid crystal layer; this is also described in the paper "Liquid Crystal Waveguides: New Devices Enabled by >1000 Waves of Optical Phase Control" by Scott R. Davis et al., Proc. of SPIE Vol. 7618. The effect of the effective refractive index is proportional to the amount of evanescent wave propagation in the liquid crystal layer, which can be achieved by using a smaller waveguide cross-section and / or selecting the wave mode. It is not necessary to cover both sides of the liquid crystal material with a conductive layer (which must be transparent to the laser frequency); instead, two conductive strips (distributed along the waveguide's extension direction) can be arranged only on the left and right sides of the lower waveguide, with a voltage applied between them—this conductive layer does not need to be optically transparent. Alternatively, it can be envisioned that a continuous conductive layer or conductive stripe (which does not need to be optically transparent) is placed only on the top side of the liquid crystal material, while conductive layers at different potentials are placed on a plane below the waveguide. The liquid crystal material can also be placed on the left and right sides of the waveguide instead of on top of it; if multiple parallel waveguides exist, voltages can be applied to these waveguides alternately (in which case the silicon material needs to be appropriately doped to achieve sufficient conductivity). Of course, there are many other conceivable configurations.

[0533] Now consider Figure 47 The settings shown are for example only. Figure 51 As shown, a liquid crystal layer 51.6 is now deposited above the radiating waveguide array 51.1, and a voltage u1(t) is applied between its upper and lower sides. By changing this voltage u1(t), radiative rotation in the first spatial direction can be achieved. For the second spatial direction, frequency variation is further utilized, i.e., the radiating waveguide 51.1 is fed from the common waveguide 51.3 with different phases. Figure 47The method shown differs from the one that achieves scanning in two spatial directions through frequency variation. In this case, the required frequency variation range in the second spatial direction does not need to cover multiple (approximately 100) 180° scans, but only needs to be completed once (potentially even less than 180°, for example, only covering the range of -50°…+50°). Therefore, the required frequency variation is reduced by at least 100 times, i.e., only on the order of ±0.1%. This same frequency variation will be applied continuously 100 times (corresponding to 100 scans in the second spatial direction), while a slow, continuous scan (range of -25…25°) is performed in the first spatial direction via a voltage u1(t) on the liquid crystal layer. The frequency variation in each scan in the second spatial direction also causes a slight scan of approximately 0.5° pixel width in the first spatial direction, which is superimposed on a scan of the same size caused by the voltage variation; if the signs of these two 0.5° scans are reversed (achieved by correspondingly selecting the signs of the frequency variation and voltage variation), approximately vertical lines will be formed in the scanning pattern—see [link to relevant documentation]. Figure 52 In principle, rapid repetitive scanning can be achieved in the first spatial direction by repeatedly applying the same type of voltage change to the liquid crystal layer; while continuous frequency scanning in the second spatial direction only needs to be performed once. This requires the liquid crystal switching to have sufficiently low inertia (both in the liquid crystal material itself and in the resistive and capacitive characteristics of its voltage drive), and the extremely slow continuous frequency changes can significantly reduce the distance-dependent frequency shift effect in the received signal. It should also be noted that the slow scanning changes can be performed gradually, i.e., the voltage u1(t) or frequency f(t) can be gradually changed before each rapid scan in another spatial direction. Furthermore, it should be noted that the inertia of the liquid crystal material may differ between voltage on / up and off / down, which can be beneficially considered when selecting the sign of the voltage change used for scanning and scan bounce (especially for rapid bounce, a direction of change with lower inertia should be used).

[0534] As mentioned above, in Figure 51 The method shown requires only a tiny frequency change on the order of ±0.1%. Since the frequency tuning range of lasers is typically much larger, the common feed waveguide can be significantly shortened, and its meandering structure can even be eliminated entirely if possible; however, this would significantly increase the frequency variation, thereby enhancing the frequency-induced scanning effect in the first spatial direction, and if necessary, compensation would be required by using different voltage variations u1(t) for different layers in the second spatial direction (in which case a rapid scan in the first spatial direction is required).

[0535] Up to this point, it has been assumed that the effective refractive index variation of the waveguide, achieved by voltage control of the surrounding liquid crystal material, is large enough to cover the required scan range in the first spatial direction. This is often difficult or impractical, especially for less sluggish liquid crystal materials, as only an effective refractive index variation of, for example, 0.05 can be achieved. To cover the entire scan range, different frequencies (approximately 17 in the example above) can be used to coarsely cover the scan area and finely scan within it by changing the effective refractive index. Therefore, a large range of frequency variations close to ±10% is still required; advantageously, the laser used can only operate in a narrow frequency range in a single mode (because continuous tuning across the entire frequency range is no longer necessary).

[0536] Replaces voltage-controlled liquid crystal layers used in radiating waveguides (such as...) Figure 53 As shown and labeled 53.7, it can also be superimposed on the common feed waveguide 53.3, and by changing its effective refractive index, the linear phase characteristics of the feed signal of the radiation waveguide can be altered, thereby achieving a change in the radiation direction in the second spatial direction. In this case, the common feed waveguide 53.3 can have a significantly shortened length (approximately 1.5 cm when a relative refractive index change of 0.05 is achievable), thus eliminating the need for a meandering structure and allowing for a completely straight extension. Scanning in the first spatial direction is achieved through frequency variation. If the feed waveguide extends along the direction of the radiation waveguide (rather than as...), Figure 53 As shown in the image, frequency changes also affect the radiation direction in the second spatial direction (because the path length of the wave propagation to each radiation waveguide increases linearly); therefore, in order to achieve the desired radiation direction, it is necessary to select the voltage control u2(t) on the liquid crystal layer for reverse compensation accordingly. Figure 53 The appropriate tilt angle shown (which can be obtained through geometric calculation) can approximately avoid the influence on the frequency in the second spatial direction (because the path length to all radiating waveguides is constant, and there are [missing information] above the connecting waveguides depending on the purpose of use). Figure 53 A liquid crystal layer (not explicitly shown, to which the average value of u2(t) as a constant voltage) is applied, while maintaining inherent geometric consistency; thus, it is not necessary to cover the full scan range of -90° to +90° in the second spatial direction through effective refractive index variations, but only the scan area required for functionality needs to be covered. Instead of using a 90° bend in the connecting waveguide, a bend with the same tilt angle as the feed waveguide can be used directly after each coupling point (where the connecting waveguide extends parallel to the feed waveguide for a short distance), and a similar, typically smaller, reverse bend of the same type can be provided in a further extension of the connecting waveguide (after the feed region of the feed waveguide) to maintain a constant path length to all radiating waveguides.

[0537] Figure 54 A method is shown in which voltage-controlled liquid crystal layers 54.6 and 54.7 are applied over both the radiating array waveguide 54.1 and the common feed waveguide 54.3, thereby changing their effective refractive indices to achieve scanning in two spatial directions; therefore, two control voltages u1(t) and u2(t) are required for scanning. Multiple (thus fast) scans are performed in one spatial direction, while only one (thus slow) scan is performed in the other spatial direction, which can be implemented continuously or in steps. Since the liquid crystal layer above the feed waveguide has a very small dimension in one dimension and thus requires a smaller area or length of conductive layer to be driven, the reduced electrical inertia effect makes it more advantageous to perform fast scanning in the second spatial direction. As mentioned above, the achievable effective refractive index variation is usually insufficient to cover the entire scan range in the first spatial direction; therefore, different frequencies are needed to cover the scan area with a coarse grid and fine scanning is performed in between by changing the effective refractive index. Since frequency scanning is no longer used in the second spatial direction, only individual constant frequencies are required, which is generally easier to achieve.

[0538] Instead of using refractive index variation in the feed waveguide, such as Figure 55 As shown, it can also be used on connecting waveguides. The liquid crystal layer 55.7 covering the connecting waveguide 55.2 is triangular, causing the length of the connecting waveguide covered by the liquid crystal layer to increase linearly, thus increasing the waveguide length affected by the refractive index linearly as well; this generates a phase shift that varies linearly along the connecting waveguide, which can be changed by a control voltage u2(t) applied to the liquid crystal layer, thereby achieving scanning in the second spatial direction. Since the feed waveguide itself already generates a phase that varies linearly along the connecting waveguide (given the linearly increasing length to the feed point) and the effective refractive index of the connecting waveguide is usually affected (even at intermediate operating voltages), to compensate, the feed waveguide 55.3 can be... Figure 55 The tilted arrangement is shown. Due to the relatively limited ability to change the relative refractive index, a triangular liquid crystal layer 55.7 of appropriate width needs to be selected and arranged above the connecting waveguide. Figure 55 (Not drawn to scale).

[0539] Figure 56 It shows relative to Figure 55 Three feasible improvement schemes are proposed. First, wave distribution is no longer achieved through a common feed waveguide, but instead uses a cascaded beam splitter 56.9—the beam splitter is designed to be asymmetric accordingly to achieve amplitude and / or phase tapering. Second, in the region covered by the liquid crystal layer, the connecting waveguides 56.2 are arranged separately to maintain low coupling. Third, in addition to using a control voltage u... 21 In addition to the triangular liquid crystal layer 56.7 of (t), there is also a layer using a control voltage u22 (t) are complementary triangular liquid crystal layers 56.8, and the two control voltages are relative to the intermediate voltage u. 20 They are complementary (i.e., u) 21 (t) = u 20 + u2(t) and u 22 (t) = u 20 - u2(t), thus only a single variable voltage u2(t) needs to be implemented, which acts on the two liquid crystal layers with different polarities. By employing two complementary triangular liquid crystal layers, compared to Figure 55 The single-triangular liquid crystal layer shown (assuming they are of the same size) can double the scanning range. The connecting waveguide section after the liquid crystal layer is designed such that its length increases linearly along the connected radiating waveguide 54.1 due to the geometry. If the length increase of two adjacent connecting waveguides is exactly an integer multiple of their inner wavelength, then in u... 21 (t) = u 22 (t) = u 20 At this time, vertical radiation is generated in the second spatial direction. When multiple frequencies are used to cover the scanning range in the first spatial direction, this condition can only be satisfied for one frequency, and compensation is required for other frequencies by adjusting the control voltage. If the path of the connecting waveguide before the liquid crystal layer is geometrically complementary to its path after the liquid crystal layer, so that the total length of the connecting waveguide remains constant, the frequency dependence in the second spatial direction can be eliminated. To achieve this, the cascaded beam splitter network needs to be rotated by 90° so that the waveguide distance at its output end is the same as that of the waveguide array, and arranged at the lower left of the two triangular liquid crystal layers. Then, the connecting waveguide in front of the liquid crystal layer will also include two straight sections and a 90° bend. Therefore, the liquid crystal layer no longer needs to cover the complete scanning range from -90° to +90° in the second spatial direction, but only needs to cover the scanning area required for the function.

[0540] As an alternative, one can also use, such as Figure 57 The arrangement shown is to prevent frequency correlation of the radiation direction in the second spatial direction; wherein the connecting waveguide is again fed through a straight and appropriately tilted waveguide 57.3 (the required angle can be obtained through geometric calculation).

[0541] Of course, such improvements can also be achieved through other combinations. For example, in Figure 57 In the arrangement shown, scanning in the first spatial direction is achieved by a voltage-controlled liquid crystal layer, which can be reverted to pure frequency scanning. Unlike the previous example, this no longer affects scanning in the second spatial direction (therefore no need for further cancellation compensation).

[0542] If a liquid crystal layer is also placed above the connecting waveguide, and the voltage applied to it is complementary to the voltage of the liquid crystal layer above the feed waveguide, the method of using two liquid crystal layers with complementary voltages can be extended to... Figure 53 and 54 The arrangement within the space allows for a doubling of the change in radiation direction in the second spatial direction.

[0543] In the example above where scanning in two spatial directions is achieved through a voltage-controlled liquid crystal layer, multiple different frequencies are required because the effective refractive index variation of the radiating waveguide alone is insufficient to achieve scanning in the first spatial direction. However, in addition to using multiple frequencies, multiple arrays (consisting of radiating waveguides, connecting waveguides, and a common feed waveguide) can be used, where the coupling points of the radiating waveguides need to have correspondingly different spacings. In this case, the laser does not need to have frequency tuning capability, and all components can be frequency-optimized (without needing to operate within a specific frequency range and possess the required optical characteristics). Of course, besides the effective refractive index variation of the waveguide, slight frequency variations can also be used—especially for phase modulation superimposed with linear frequency variations, thereby determining the sign of the received frequency via a real-valued mixer; the change in radiation direction caused by such slight frequency variations (usually small) can be considered in the control of the liquid crystal layer.

[0544] It should also be noted that using multiple total arrays has the advantage of eye safety when the total output power is the same, because the power is distributed to different locations.

[0545] Radiation redirection using waveguide arrays with prisms

[0546] The method of using multiple arrays (including radiating waveguides, connecting waveguides, and common feed waveguides) has been described above. The following will focus on... Figure 47 Consider the three arrays shown as examples, which scan in two spatial directions by frequency. In the second spatial direction, each array scans the entire detection range. In the first spatial direction, the radiation direction of each array is deflected by a prism (or a prism-shaped portion of a larger common element) above each radiation array; the shape and position of the prism, and the direct scanning range of the array (i.e., without prism deflection), are chosen such that the three scanning ranges obtained by prism deflection are interconnected, thereby achieving, for example, a scanning range of 180° wide and maintaining high resolution even in edge regions up to ±90° (where resolution is poor due to the continuously shrinking effective aperture without prisms). In principle, a detection range wider than 180° can even be achieved using this method.

[0547] The scanning range can be enhanced by using a prism above the radiation array, especially when the radiation angle in the prism approaches the boundary of total internal reflection. Furthermore, if a frequency scanning method is used, a prism made of dispersive material can also achieve scanning enhancement.

[0548] Finally, it should be noted that if the sensor only requires a relatively small scanning range and it is desired to achieve a change in radiation direction through a small change in frequency or effective refractive index, this can be achieved without a prism by using a non-centered, tilted array scanning area, i.e., the array is tilted relative to the central radiation direction rather than perpendicular to it. This creates an enhancement effect through geometry (however, to achieve the same radiation direction width, the array size needs to be increased because the tilted angle reduces the effective aperture, which is the reason for the enhancement effect; if this tilted arrangement is used in the first spatial direction, the radiation waveguide needs to be extended).

[0549] Waveguide array fed by a planar lens via parallel transmit / receive paths

[0550] In the arrangements considered so far that directly radiate using waveguide arrays (i.e., without additional optical elements), the primary consideration has been single-channel sensors, i.e., sensors with only a single transmit / receive path. However, for sensors with long detection ranges and high resolution, multiple transmit / receive paths are required because the product of the number of pixels and the necessary data acquisition time per pixel far exceeds the typically used 50 ms cycle—as mentioned above, approximately 16…32 parallel transmit / receive paths are needed. A basic approach is to configure independent waveguide arrays with different designs for each transmit / receive path (so that different scan ranges can be covered when using the same frequency and / or the same or similar liquid crystal layer control voltage); however, this requires a large amount of space, especially since, due to the high resolution, the individual waveguide arrays must be significantly larger than the arrangement shown above (in the range of 0.2…2 square centimeters).

[0551] Therefore, a solution is needed that allows the same waveguide array to be preferentially used for all transmit and receive channels. One possible implementation is to use a planar lens, similar to the Rotman lens or a circular Luneburg lens with a radially dependent refractive index known in the microwave field. To realize such a planar lens on a photonic chip, a constant refractive index different from the surrounding environment can be used within a planar region of a corresponding shape (the classic lens approach), or a spatially varying refractive index can be used (especially a continuously gradient approach similar to a Luneburg lens). Different refractive indices can be achieved using different silicon-based materials, or by changing the effective refractive index by covering a uniform or non-uniform silicon substrate as a wide waveguide, where different control voltages can be applied to different regions of the liquid crystal layer if necessary to compensate for tolerances or frequency dependence.

[0552] Figure 58 One possible arrangement is shown, in which such a planar lens 58.3 is used, symbolically represented only by a module having 32 uniformly distributed inputs (left) and approximately 10,000 equally spaced outputs (right) (the specific structure and implementation on the photonic chip are not shown; the terms "input" and "output" here refer to transmission, and vice versa for reception). When fed from a single input port, the wavefronts arriving at each output port are typically oblique plane waves, i.e., the phase varies linearly along the output. The slope of this linear phase arrangement depends on the input selected for feeding; it varies at least approximately linearly from negative values ​​to positive values ​​of equal absolute value with respect to the input n=1…32 (due to the symmetrical arrangement and design). From the output of the planar lens, an array of connecting waveguides 58.2 extends to radiating waveguides 58.1, wherein the connecting waveguides are controlled by two complementary triangular liquid crystal layers 58.4 and 58.5, whose variable voltages are u 21 (t) = u 20 + u2(t) and u 22 (t) = u 20 - u2(t). When the control voltage u2(t) = u 20 At this time, the liquid crystal layer does not change the phase relationship between the connecting waveguide signals, that is, the radiation waveguide maintains the linear phase arrangement from the output end of the planar lens, thereby producing focused and typically tilted radiation in the second spatial direction; the planar lens should be designed such that the radiation angle varies from -7.75° to +7.75° in 0.5° steps at the input end. In order to achieve pixels with a spacing of 0.05° in the second spatial direction (i.e., 10 pixels per input end), the control voltage u2(t) needs to be changed accordingly; thus, a 16° wide scan range containing 320 pixels can be covered in the second spatial direction. Since only a 0.45° change in radiation direction needs to be achieved through the liquid crystal layer, the length scale and / or voltage change of the liquid crystal layer can be very small, which can provide support for fast scanning and fast bounce for changes in radiation direction.

[0553] The scanning in the first spatial direction, i.e. the deflection of the radiation direction of each waveguide 58.1 in the array, is achieved by frequency variation.

[0554] For parallel operation, i.e., parallel detection of 32 directions and 32 pixels, the same modulated laser signal is applied to all 32 inputs; these 32 parallel-acquired pixels have different angles in the second spatial direction (due to different lens inputs) but the same angle in the first spatial direction (because the input signal and frequency are the same). In principle, the specific spatial directions for implementing fast scanning (i.e., repeated scanning) and slow scanning (i.e., single scanning) can be freely chosen; similarly, a hybrid mode is also conceivable.

[0555] Figure 58 The parallel-extending connecting waveguide 58.2 and the radiating array waveguide 58.1 in the arrangement shown can be as follows: Figure 59 The images shown are replaced by single wide waveguides, namely waveguide surfaces 59.2 and 59.1 (marked with crosshairs). Since the propagation of a tilted wavefront in the waveguide surface results in a corresponding tilted wavefield, the waveguide surface must gradually widen, i.e., expand, with increasing distance from the planar lens (although the widened width can be maintained over the entire length). In the radiating waveguide surface 59.1, instead of a single coupling point, there are parallel and equidistant coupling lines 59.7, which run the entire surface and are arranged perpendicular to the central axis of the waveguide surface. It is worth noting that, due to the presence of the tilted trapezoidal wavefield, the two-dimensional radiation shape changes slightly with the frequency of the scan in the first spatial direction as the radiation angle changes in the second spatial direction.

[0556] Instead of scanning in the first spatial direction by frequency variation, such as Figure 60 As shown, a voltage-controlled liquid crystal layer 60.6 can be used instead on a radiating waveguide array consisting of waveguides 60.1. All 32 parallel-acquired pixels still have the same angle in the first spatial direction, which is defined by the control voltage u1(t) of the liquid crystal layer above the waveguide array. Since the radiating waveguide can only achieve relatively small effective refractive index changes, multiple different discrete frequencies are usually required to cover the required scanning range. In principle, the specific spatial direction for implementing fast or slow scanning can be freely chosen; since only 10 pixels need to be scanned in the second spatial direction, far fewer than the number of pixels in the first spatial direction, considering the inertia during bounce and the associated time loss, it may be better to choose to implement slow scanning in the second spatial direction.

[0557] To perform scanning in the second spatial direction, i.e., to achieve 10 pixels in a 0.05° grid at each input, instead of the aforementioned method of covering the waveguide with two triangular liquid crystal layers, one can use... Figure 61Frequency scanning is performed using connecting waveguides 61.2 of varying lengths (this applies only when frequency scanning is not used in the first spatial direction, but instead a voltage-controlled liquid crystal layer 61.6 covering the radiating waveguide 61.1 is used). The parallel and equally spaced connecting waveguides 61.2 have, for example, a 90° arc-shaped bend, causing their length to increase linearly. The constant length increase between adjacent connecting waveguides precisely accommodates the two waveguide wavelengths at the center frequency f0, thus ensuring that each input signal from lens 61.3 maintains the linear phase arrangement received from the planar lens output at the radiating waveguide 61.1. A change in frequency results in an additional linear phase shift in the signal reaching the radiating waveguide (because the length increase between adjacent connecting waveguides is no longer exactly equal to the two waveguide wavelengths), thereby altering the radiation angle in the second spatial direction; to achieve the required 0.45° change, only a frequency change of approximately ±0.07% is needed (thus, the small change in radiation direction in the first spatial direction can be compensated for when controlling the liquid crystal layer 61.6). It's also important to note that the bends connecting waveguides are typically not 90° because the required condition is that the increase in length between two adjacent waveguides must be sufficient to accommodate an integer multiple of the waveguide wavelength at the center frequency f0. If multiple discrete frequencies are needed to cover the required scan range in the first spatial direction (because the achievable effective refractive index variation is insufficient), then each frequency must satisfy the above condition, i.e., the increase in length between two adjacent waveguides must be sufficient to accommodate an integer multiple of the waveguide wavelength, which is generally difficult to achieve.

[0558] The 32 input ports of the lens can also be arranged in a compact design, such that their corresponding radiation directions differ by only one pixel width (0.05°). In this case, a 14.4° scan can be achieved using the liquid crystal layer or through frequency variation, representing a 32-fold improvement over the first method. This compact arrangement of the planar lens input ports simplifies its design and improves its robustness at different frequencies. If a rapid scan is performed in the second spatial direction (while scanning slowly and continuously in the first spatial direction), this method reduces the number of bounces in the second spatial direction by a factor of 32, which is particularly helpful when using voltage-controlled liquid crystal layers due to their inertia. Alternatively, a hybrid arrangement of sparse and dense input ports can be used—for example, grouping closely adjacent waveguides with larger gaps between groups (this also reduces the number of bounces required for rapid scanning in the second spatial direction).

[0559] For closely adjacent input terminals, coupling may occur between their leads, leading to fuzzy environmental recognition. To avoid this, signals with different pseudo-random phase modulations can be used for different input terminals, making the signals coupled to other input terminals, especially adjacent pixels, incoherent, i.e., appearing only as noise at the receiver and not generating power peaks in the two-dimensional correlation (thus preventing detection). This method according to the invention, employing different modulation signals for parallel transceiver paths, can of course be applied to all the aforementioned methods and arrangements with multiple parallel transceiver paths. If the evaluation of the received sequence is implemented using a fixed wiring circuit as shown in Figure 8 or 10, the value of the modulation sequence b(n) needs to be changed for each transceiver path.

[0560] So far, the consideration has been that the number of inputs to the planar lens matches the number of parallel transmit / receive paths, and that 10 pixels per transmit / receive path are achieved in the second spatial direction by influencing the wavelength in the connecting waveguide (by changing the frequency or effective refractive index). The 10 pixels per transmit / receive path can also be achieved by having the lens have 320 uniformly distributed inputs, and by using a switching matrix, each of the 32 transmit / receive paths can be switched between 10 different lens inputs.

[0561] Finally, it's important to note that the planar lens method can also be applied to single-channel sensors, i.e., sensors with only one transmit / receive path, used in conjunction with a switch matrix at their input. The number of inputs to the planar lens can be equal to the number of radiation directions in the second spatial direction, or fewer inputs can be used, with further alterations to the radiation direction achieved through relatively small changes in the frequency or effective refractive index of the connecting waveguide. Scanning in the first spatial direction, as described above, is achieved by changing the effective refractive index or frequency of the radiation array waveguide.

[0562] A waveguide array fed with signals of multiple frequencies to achieve parallel transmission and reception.

[0563] In the foregoing section, parallel transmission and reception were achieved by simultaneously driving multiple inputs of the planar lens, thereby achieving parallelism in the second spatial direction (i.e., simultaneously acquiring multiple pixels with different angles in the second spatial direction but the same angle in the first spatial direction). Alternatively, parallelism can be achieved in the first spatial direction; for this purpose, as... Figure 62 As shown, multiple different frequencies f can be used. n Signals (n = 1, …, N, e.g., N = 32 different frequencies) are transmitted, causing the waveguide array 62.1 to radiate simultaneously in N different directions, i.e., to transmit and receive frequencies of f in each direction. nThe received signal is divided into N different frequencies. To separate these N frequencies from the received signal, the received signal is distributed to N mixers, each with its own mixing frequency f. n (n = 1, …, N)(The mixer is not in Figure 62 (as shown in the diagram); thus, a low-frequency received signal corresponding to its respective transmission frequency fn is obtained at the output of the mixer, that is, corresponding to its respective radiation direction. The N radiation directions acquired in parallel in the first spatial direction are scanned by a voltage-controlled liquid crystal layer 62.6 covering the radiating waveguide array 62.1 to cover the entire detection range; if these N different frequencies fn... n If the radiation direction is uniformly distributed throughout the detection range, then the effective refractive index change of the radiation waveguide only needs to be within a small range, which can be achieved through a voltage-controlled liquid crystal layer.

[0564] Scanning in the second spatial direction is achieved through two complementary triangular liquid crystal layers 62.4 and 62.5 covering the connecting waveguides 62.2, which are fed through a common waveguide 62.3. The N radiation directions corresponding to different frequencies, acquired in parallel in the first spatial direction, have the same radiation direction in the second spatial direction because, from a top-down perspective, adjacent connecting waveguides 62.2 compensate for their spacing difference with the feed point of the common waveguide 62.3 precisely through their 360° / (2π) = 57.3° arc-shaped bends, thus making the second radiation direction independent of frequency. Alternatively, a combination of tilted feed waveguides and connecting waveguides with identical bends (i.e., constant bend radii) can be used, similar to... Figure 53 The setup shown is illustrated. Alternatively, a cascaded beam splitter network can be used instead of the feed waveguide to feed the subsequent straight connecting waveguide; however, this would make the amplitude and / or phase arrangement more difficult to achieve.

[0565] To generate multiple frequencies f n This requires using different lasers, either a laser capable of generating the entire frequency comb, or modulating different frequencies on a single laser frequency. Of course, this method of incorporating signals of multiple frequencies can also be used in other setups to achieve parallel transmission and reception.

[0566] Scanning is performed in the first spatial direction by utilizing the effective refractive index change of the waveguide, and focusing and scanning are performed in the second spatial direction using a non-waveguide-based method.

[0567] For scanning and focusing in two spatial directions, the invention, as previously considered, particularly... Figure 20-29In the arrangement mentioned, frequency scanning in the first spatial direction is achieved through one or more waveguides with coupling points, while in the second spatial direction it is achieved through other methods, which requires at least one optical element (e.g., a lens and / or a liquid crystal element) in addition to the photonic chip. Instead of changing the radiation of the waveguide by frequency, it can also be achieved by changing its effective refractive index, for example, by changing the voltage applied to the liquid crystal layer surrounding the waveguide. In an arrangement with multiple waveguides, the same voltage applied to the liquid crystal material can be used for all waveguides. Since the effective refractive index variation achievable by voltage-controlled liquid crystal layers is limited, an arrangement with multiple waveguides is advantageous, with each waveguide corresponding to a different and narrower scanning range in the first spatial direction—similar to... Figure 28 and 29 The example shown is a scanning pattern in which multiple waveguides achieve different radiation angles at their respective frequencies.

[0568] The advantage is that the laser frequency can remain constant when scanning by changing the effective refractive index of the waveguide with coupling points; therefore, the laser does not need to have frequency tuning capability, and all components, especially those required for changing the radiation direction in the second spatial direction, can be frequency-optimized (without operating within a certain frequency range and possessing the required optical characteristics). Of course, in addition to changing the effective refractive index of the waveguide, the frequency can also be changed—the goal of which is to enhance the scanning effect, or to superimpose linear frequency changes on phase modulation, especially in real-valued mixers to determine the sign of the received frequency.

[0569] Accurately determine distance and / or relative velocity under high modulation bandwidth

[0570] Near-field lidar systems require high-precision distance measurement capabilities. Distance measurement via phase modulation necessitates a very short modulation time T. m This requires a very high sampling frequency, which also leads to a high computational load. To avoid this, a high modulation bandwidth can be used in the linear frequency variation superimposed on the phase modulation; in this way, as long as the frequency offset caused by the Doppler effect, i.e., relative velocity, is known, a very accurate distance measurement can be derived from the receiving frequency (according to formula (12), the receiving frequency is composed of the distance-related component generated by the linear frequency modulation and the relative velocity-related component).

[0571] One important application scenario for near-field systems is autonomous parking, which primarily focuses on static environments (infrastructure and other stationary vehicles) and requires high-precision distance detection. For reflections from infrastructure, the relative velocity is known a priori—not only when the vehicle is stationary but also when it is moving, because the relative velocity of static objects can be calculated from the vehicle's own velocity and the angles in the two spatial directions.

[0572] If the relative velocity is unknown, two methods of inverse linear frequency variation can be used (e.g., through two consecutive periods or two adjacent scanning planes; these have been explained in detail above). By using the sum and difference of the received frequencies obtained through the two inverse modulation bandwidths, the Doppler shift and distance-related frequency components can be separated, thus enabling very accurate determination of relative velocity and distance; this also avoids the problem of severe inaccuracy in relative velocity measurement caused by insufficient distance determination accuracy due to the combination of phase modulation and high modulation bandwidth.

[0573] Final Note

[0574] Based on the above application examples, the ideas and designs described in this invention can be applied in a simple manner to general detection and parameter interpretation; that is, they can also be used for other numerical values. Therefore, general parameters are often given in the equations and figures.

[0575] Some of the proposed new methods are not only novel in combination with other methods, but are also novel in themselves compared to existing technologies. Examples of this include:

[0576] - The method for determining and implementing correction values ​​is designed to compensate for coupling and reflection effects within or in the immediate vicinity of a lidar system, particularly in the overlay layer. The method can also be used without superimposed frequency modulation.

[0577] - The method includes a transceiver unit having one or more waveguides for scanning in a first spatial direction by frequency, and a scanner for a second spatial direction. It can also be used in combination with other modulation forms, especially when the frequency is gradually changed, and can even be used in incoherent lidar systems.

[0578] The method of determining the angle error (e.g., by changing hardware characteristics or sensor orientation error) can also be used for other modulation forms, as it is essentially based solely on the inherently given Doppler measurement capabilities of the coherent lidar system.

[0579] - The method described is used to determine the location of a road surface at a distance. It can also be used in other modulation forms, such as in purely linear frequency modulation (typically consisting of two frequency ramps with opposite signs of their slopes); the correlation values ​​used are calculated based on corresponding other correlations (in purely linear frequency modulation, calculated in the form of a Fast Fourier Transform (FFT) for each frequency ramp).

[0580] - A method for determining the sign of a real-value mixer receiver frequency using non-binary phase modulation can be used even without superimposed frequency modulation.

Claims

1. A lidar system for detecting the surrounding environment, wherein - for transmitting and / or receiving light beams, the lidar system has an array of waveguides, which consists of a plurality of radiating, preferably identical and parallel arranged waveguides (58.1, 60.1, 61.1), which each have a plurality of coupling points, or the lidar system has a radiating waveguide facet (59.1) with a plurality of, in particular strip-shaped, coupling structures (59.7), wherein preferably the coupling points or coupling structures (59.7) for radiation are located in an approximately equidistant grid, - the radiation direction is made to wobble along a first spatial direction by changing the frequency and / or the effective refractive index of the waves in the radiating waveguides (58.1, 60.1, 61.1) or in the radiating waveguide facet (59.1), - the array of radiating waveguides or the radiating waveguide facet (59.1) is fed from the output of a planar lens (58.3, 59.3, 60.3, 61.3) by connecting waveguides (58.2, 60.2, 61.2) or connecting waveguide facets (59.2), - the radiation direction is made to wobble along a second spatial direction, preferably perpendicular to the first spatial direction, by changing the frequency and / or the effective refractive index of the waves in these connecting waveguides (58.2, 60.2, 61.2) or these connecting waveguide facets (59.2), - the planar lens (58.3, 59.3, 60.3, 61.3) has a plurality of inputs, which realize different radiation directions in the second spatial direction and which are coupled to transceiving channels running in parallel, so that the array of radiating waveguides or the radiating waveguide facet (59.1) can be used for simultaneous detection of different radiation directions, - wherein preferably one of the two spatial directions is horizontal and the other is vertical, - preferably the array of radiating waveguides or the radiating waveguide facet (59.1) and the connecting waveguides (58.2, 60.2, 61.2) or connecting waveguide facets (59.2) together with the planar lens (58.3, 59.3, 60.3, 61.3) are realized on a photonic chip.

2. The lidar system of claim 1, wherein, The lidar system works in a coherent manner and preferably superimposes continuous frequency changes, in particular phase modulations, wherein the continuous, in particular linear, frequency changes are implemented within a pixel and across pixels.

3. The lidar system of claim 1 or 2, wherein, The different radiation directions realized by the plurality of inputs of the planar lens (58.3, 59.3, 60.3, 61.3) lie in a coarse grid in the second spatial direction, while by the change in frequency or effective refractive index in the connecting waveguides (58.2, 60.2, 61.2) or in the connecting waveguide facets (59.2) the radiation direction is only scanned over a small range, so that the range required for the change in frequency and / or effective refractive index is advantageously reduced.

4. The lidar system of any one of the preceding claims, wherein, In the immediate vicinity of the radiating waveguides (60.1, 61.1) or the radiating waveguide faces and / or the connecting waveguides (58.2, 60.2) or the connecting waveguide faces (59.2) of the waveguide array, in particular above them in a planar manner, there is a liquid crystal material (58.4, 58.5, 59.4, 59.5, 60.4, 60.5) whose optical properties are influenced by the application of a voltage, respectively, so that the effective refractive index of the radiating waveguides (60.1, 61.1) or the radiating waveguide faces and / or the connecting waveguides (58.2, 60.2) or the connecting waveguide faces (59.2) changes, which in turn enables a change in the radiation direction in the first spatial direction and / or the second spatial direction.

5. The lidar system of claim 4, wherein, In the immediate vicinity of the connecting waveguides (58.2, 60.2) or the connecting waveguide faces (59.2), in particular above them in a planar manner, there are two triangular areas (58.4, 58.5, 59.4, 59.5, 60.4, 60.5) with liquid crystal material which complement each other, which are manipulated by two complementary voltages consisting of a direct current component and a variable component of opposite polarity.

6. The lidar system of claim 4 or 5, wherein, In the immediate vicinity of the radiating waveguides (60.1, 61.1) or the radiating waveguide faces of the waveguide array, in particular above them in a planar manner, there is a liquid crystal material (60.6, 61.6) for changing the effective refractive index, wherein a change in the radiation direction in the first spatial direction is effected in a coarse raster with different frequencies, while by changing the effective refractive index only a small area is scanned, respectively, so that a reduced variability of the effective refractive index is sufficient.

7. The lidar system of any one of the preceding claims, wherein, The length of the connecting waveguides (61.2) changes at least approximately linearly on the input end of the connected radiating waveguides (61.1) or the radiating waveguide faces of the waveguide array, so that a scan in the second spatial direction is achieved by a change in the frequency, for which the connecting waveguides (61.2) preferably consist of one or more straight and respectively parallel sections and one or more bends with exactly the same shape or at least the same angle of curvature.

8. The lidar system of any one of the preceding claims, wherein, On the radiating waveguide array or the radiating waveguide face (59.1), preferably in both spatial directions, there is an amplitude arrangement and / or a phase arrangement, so-called amplitude tapering and / or phase tapering, in order to reduce or suppress side lobes in the radiation as far as possible, wherein the amplitude tapering is achieved by different intensities of the coupling, while the phase tapering is preferably achieved by an inexact equidistance arrangement of the coupling structures and / or an inexact equidistance arrangement of the radiating waveguides (58.1, 60.1, 61.1) or an inexact equidistance arrangement of the input ends in the radiating waveguide face (59.1) and / or the output ends of the planar lenses (58.3, 59.3, 60.3, 61.3).

9. The lidar system of any one of the preceding claims, wherein, Further waveguides of the same type are provided parallel to the radiating waveguides (58.1, 60.1, 61.1) and parallel to the connecting waveguides (58.2, 60.2, 61.2), which are not coupled to the output of the lens (58.3, 59.3, 60.3, 61.3), thereby avoiding or at least reducing effects due to inter-waveguide coupling.

10. The lidar system of any one of the preceding claims, wherein, The planar lens (58.3, 59.3, 60.3, 61.3) is implemented by a liquid crystal layer above the waveguide plane, if necessary composed of a plurality of regions, wherein tolerances or frequency dependencies can preferably be compensated by one or more control voltages.

11. The lidar system of any one of the preceding claims, wherein, The transceiver channels are switchable in turn between a plurality of lens inputs, respectively, thereby enabling different radiation directions in the second spatial direction.

12. The lidar system of any one of the preceding claims, wherein, Different phase modulations are used for different transceiver channels, in particular to enable robustness to coupling in the circuit section before the lens input.

13. The lidar system of any one of the preceding claims, wherein, A plurality of radiating waveguide arrays or radiating waveguide planes and associated connecting waveguides or connecting waveguide planes and planar lenses are provided, which are preferably coupled in series or in parallel to the same modulated laser source and have different structures, such that they enable different radiation directions in at least one spatial direction at the same frequency, thereby advantageously reducing the required range of frequency and / or effective refractive index variations.

14. The lidar system of any one of the preceding claims, wherein, A plurality of radiating waveguide arrays or radiating waveguide planes and associated connecting waveguides or connecting waveguide planes and planar lenses are provided, and for radiation deflection there are prisms or larger, preferably common, prismatic sub-regions of the main body above the radiating waveguide arrays or radiating waveguide planes, in particular to enable a large detection range and / or a high resolution even in the edge region.

15. The lidar system of any one of the preceding claims, wherein, Deviation in the radiation direction can occur in particular due to orientation errors, due to frequency inaccurately known and / or due to the relationship of the effective refractive index to the respective control variable not being accurately known, characterized in that these deviations are determined from a measured radial relative speed of the stationary object, in order to subsequently take these deviations into account and / or to correct them.

16. The lidar system of claim 15, wherein, The road surface is used as a stationary object, the angle thereof in the vertical direction being determined preferably from the measured distance and the sensor mounting height.

17. The lidar system of any one of the preceding claims, wherein, Means for changing the radiation direction are provided, which are used to compensate for orientation errors and / or to adapt the detection range in particular adaptively to the traffic situation.

18. The lidar system of any one of the preceding claims, wherein, In particular to ensure eye safety, the means for changing the radiation direction are monitored and / or the radiation direction change is monitored, and the monitoring is effected by checking the change in the received signal in each respective spatial direction with respect to object reflections or internal reflections and coupling and cover layer reflections.

Images