Coherent lidar system for environmental perception with binary power modulation and low processing overhead.

Abstract

The present invention relates to a coherent lidar system for environmental sensing, which emits a power-modulated signal realized by irregular switching on and off, the switching times forming a subset of an equidistant time grid. The reflected signals are received and converted into a low-frequency signal by mixing and digitized into a received sequence, wherein the variable dimensions of time shift and frequency shift of signals reflected from objects are determined from the received sequence in digital signal processing means.This involves using a two-dimensional correlation filter for the dimensions of time shift and frequency shift, or, to reduce the required computational effort, a discrete Fourier transform is calculated. The respective frequencies are determined from the values ​​of this Fourier transform, and the received signal sequence is then tuned back in frequency to each of these frequencies. A correlation is determined between the sequence thus generated and the modulation sequence, or the modulation sequence adjusted for its mean value. The respective time shift, and thus the object distance, is determined from the values ​​of this correlation, and the radial relative velocity of the respective object is determined from the respective frequency.

Description

[0001] The invention relates to a coherent lidar system, particularly for environmental detection in automotive applications. The coherent lidar system operates with power modulation, which is implemented by switching on and off, and features a cost-optimized signal evaluation according to the invention. State of the art

[0002] Motor vehicles are increasingly being equipped with driver assistance systems that use sensor systems to monitor the surroundings and, based on the detected traffic situation, derive automatic vehicle reactions and / or instruct the driver, particularly by issuing warnings. A distinction is made between comfort and safety functions.

[0003] Now, however, developments are moving in an even more far-reaching direction. The driver is no longer merely assisted; instead, the driver's task is increasingly performed autonomously by the vehicle, meaning the driver is increasingly being replaced. This is referred to as autonomous driving.

[0004] Autonomous driving in particular requires sensors with highly accurate environmental information that is easy to evaluate by machines. Radar systems are limited in their angular accuracy and separation capability and, at least not yet, cannot satisfactorily meet these high detection requirements, either alone or in combination with camera systems. Therefore, lidar systems are also being used in parallel. These systems have a similarly high angular resolution (horizontal and vertical) as a camera, but also provide distance information and separation capability in each pixel. Today, so-called time-of-flight lidar systems are commonly used, which treat electromagnetic radiation as particles and can therefore only directly measure distance but not relative speed.However, coherent lidar systems are increasingly coming into focus. These treat electromagnetic radiation as waves (like radar systems) and can thus directly measure the relative velocity of objects via the Doppler effect. Further advantages of coherent lidar systems include their robustness to interference from other sources (e.g., from other lidar systems or sunlight) and their increased sensitivity at greater distances, thus allowing for longer ranges. Furthermore, coherent lidar systems are considered to have a higher potential for high-level semiconductor integration, which promises lower manufacturing costs.

[0005] In coherent lidar systems, the emitted electromagnetic wave is modulated, i.e. it changes in at least one of the parameters amplitude, frequency or phase over time - otherwise, distance measurement would not be possible. The most commonly used modulation in coherent lidar systems is linear frequency modulation (FMCW = frequency modulated continuous wave), which usually consists of two frequency ramps whose gradients have opposite signs. However, this modulation has ambiguity problems, particularly with multiple reflections in the same beam direction, and generating a highly linear frequency change is complex. These disadvantages do not occur, or are less so, with phase modulation (e.g., with pseudo-random changes across discrete phase values). However, the digital evaluation of the received signals is more complex, and the approaches proposed in the state of the art (e.g.,in the article “Phase-Coded-Based Modulation for Coherent Lidar” by Sebastian Banzhaf and Christian Waldschmidt, published in IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 70, NO. 10, OCTOBER 2021) are associated with disadvantages, particularly with regard to sensitivity and thus range. Task, solution and advantages of the invention

[0006] The object of the invention is to provide a simple modulation and an associated low-cost signal evaluation for a coherent lidar system without any disadvantages, in particular with regard to sensitivity and thus range.

[0007] This problem is fundamentally solved by a coherent lidar system according to claim 1. Expedient embodiments of the invention are claimed in the subclaims. The core idea is that power modulation is realized using a switch or changeover switch, which represents a very simple and cost-effective approach, and that the signal analysis for determining the relative speed and distance of objects is implemented in a cost-effective manner, in particular by using a comparatively small number of fast Fourier transforms.

[0008] The advantages of the invention arise from the fact that a lidar system with high performance and low price can be realized.

[0009] The coherent lidar system for environmental detection emits a power-modulated signal, which is realized by irregular, particularly pseudo-random, switching on and off. The switching off need not be perfect, but can only be characterized by a complete power shutdown or only by a significant power reduction, and the switching on and off times form a subset of an equidistant time grid. Furthermore, it receives the signals reflected back from objects, which are delayed compared to the transmitted signal by the distance-dependent propagation time and advanced in frequency by the relative speed-dependent Doppler effect. It converts these signals into a low-frequency signal by mixing and digitizes them into a reception sequence.to determine the variable dimensions of time shift and frequency shift of signals reflected from objects from this reception sequence in digital signal processing means. According to the invention, a two-dimensional correlation filtering is used for the variable dimensions of time shift and frequency shift of signals reflected from objects, preferably implemented using a hard-wired digital circuit, or to reduce the required computational effort, firstly, a discrete Fourier transform is calculated, preferably using a fast Fourier transform, over the values ​​of the reception sequence, optionally extended with zeros. Secondly, the respective frequencies are determined from values ​​of this discrete Fourier transform, in particular from magnitude peaks lying above a first detection threshold. Thirdly, the reception sequence is frequency-reversed for the respective frequencies.fourthly, a correlation is determined between the sequence thus generated and the modulation sequence, in particular formed from the switching values ​​1 and 0, or the modulation sequence adjusted by its mean value, and fifthly, the respective time shift and thus object distance are determined from values ​​of this respective correlation, in particular from magnitude peaks lying above a second detection threshold, and the radial relative velocity of the respective object is determined from the respective frequency.

[0010] In an advantageous embodiment of the invention, the power modulation is realized by a switch between two transmission paths, which thus have inverse power modulation to each other.

[0011] Preferably, the first detection threshold for magnitude peaks of the discrete Fourier transform is less above the noise than the second detection threshold for magnitude peaks of the respective correlation.

[0012] Advantageously, the power modulation sequence can be repeated periodically, whereby, in particular during continuous scanning of the laser beam, the cyclical property of the modulation and reception sequence can be exploited and detection in different directions can be realized.

[0013] Furthermore, the laser beam can be scanned continuously and overlapping sections of a long, possibly periodic power modulation sequence can be used for the successive detection directions.

[0014] According to an advantageous embodiment of the invention, the laser frequency changes continuously at least in sections, preferably with an at least approximately linear progression, so that the entire modulation is composed of a line modulation and a frequency modulation.

[0015] Preferably, it is taken into account that the frequency is shifted both by the relative speed-dependent Doppler effect and by the distance-dependent transit time caused by the linear frequency change, in particular by determining the object distance from the determined time shift and the radial relative speed of the object from the determined frequency shift less its contribution caused by the time shift.

[0016] Furthermore, a sign-correct determination of the reception frequency and thus the unambiguous determination of the relative speed of objects when using a real-valued mixer can be realized by taking into account that, especially for more distant objects, only a relative speed hypothesis is possible or at least more plausible due to the known propagation time-dependent component of the frequency shift.

[0017] Advantageously, a determination of the reception frequency with the correct sign and thus the unambiguous determination of the relative speed of objects when using a real-valued mixer can be realized by varying the steepness of the frequency change and recording and evaluating data for an object at different steepnesses and using the fact that the two frequency shift effects of relative speed and transit time are then in different relations to one another, in particular characterized in that the sign, but not the amount of the steepness of the frequency change is varied, so that the two frequency shift effects of relative speed and transit time add up with different signs and thus the sign of the reception frequency can be determined.

[0018] Furthermore, the frequency change can be used to change the beam direction, in particular in a continuous manner at least in sections, in order to be able to acquire data for several pixels of different directions. Brief description of the drawings Fig. Figure 1 shows the coherent lidar system with binary power modulation. Fig. 2 shows the course of a pseudorandom binary power modulation. In Fig. Figure 3 shows the real part of the low-frequency analog received signal and the corresponding sample values ​​for an object. Fig. Figure 4 shows the magnitude of the two-dimensional correlation of the reception sequence for two objects. Fig. 5 shows the FFT of the reception sequence for the example of two objects. Fig. Figure 6 shows the two one-dimensional correlations for the example of two objects. In Fig. Figure 7 shows a lidar system with two transmit / receive paths, where the power modulation is realized by a switch between the two paths. Fig. Figure 8 shows the linear change of the transmit frequency during binary power modulation. Fig. 9 shows the received frequency shifted in time to the transmitted frequency in the case of no Doppler shift (i.e. at zero relative speed). Examples of implementation

[0019] Fig. Figure 1 schematically shows a coherent lidar system 1.1. A coherent signal in the wavelength range of approximately λ=1550nm is generated by the laser source 1.2; the coherence length is at least several microseconds and the frequency is constant. The signal then reaches a switch 1.3, via which it can be switched on and off for the further signal path. Switching (i.e., switching on or off) only takes place in a fixed grid of, for example, 6.67ns and occurs, for example, pseudorandomly, i.e., only with a frequency or probability of 50% is T m = 6.67ns the switch state is changed. This binary power modulation can also be considered as an amplitude modulation with a modulation sequence b(n) consisting of the values ​​+1 and 0; its course is shown in Fig. 2. The modulation repeats with the period N m= 4096, i.e., every 27.3µs (especially for data acquisition for multiple detection directions, i.e., pixels). The modulated signal passes through an amplifier 1.4, a circulator 1.5 (i.e., a transmit-receive switch), is radiated via a transmit-receive unit 1.6, and partially reflected back by an object 1.7; with a delay dependent on the object distance r and thus variable. t0=2r / c, where c = 3·10 8 m / s is the speed of light, and with a frequency shift dependent on the radial relative speed v and thus variable, generated by the Doppler effect fD=2v / λ This signal is then picked up by the transmit-receive unit 1.6 and passed through the circulator 1.5 into the further receive path. In a complex-valued mixer 1.8 (also called an IQ mixer), the modulated received signal is superimposed with the unmodulated laser signal and converted into a complex-valued low-frequency signal with the help of the photodiode unit 1.9; the frequency f e this signal corresponds to the Doppler shift f D according to reference (1b): fe=fD=2v / λ, The modulation of this signal is delayed by the signal propagation time compared to that of the transmitted signal. Fig. 3 is the real part of this low-frequency analogue received signal e a (t) in the case of an object. The signal is then converted into an analog-to-digital converter unit 1.10 with the sampling frequency f s = 150MHz, ie all T s= 6.67ns sampled and digitized - the resulting values ​​of the real part are in Fig. 3 marked as points; because here sampling time T s = 6.67ns and modulation time T m = 6.67ns are equal, the number N s of the samples per modulation period equal to their raster length N m , so N s = N m = 4096, so that N = 4096 without an index is used for both. The complex-valued sampling signal e(n), also referred to as the receive sequence, can be described as follows: e(n)=a⋅b(n−m0)⋅exp(2πj^⋅n / N⋅k0), where it is assumed that the running time t0 is an integer multiple m0 of the modulation time T m = 6.67ns and thus also the sampling time T s = 6.67ns is: m0=t0 / Tm=2r / c / Tm and the Doppler shift corresponding k0=N⋅Ts⋅fD=N⋅Ts⋅2v / λ is also an integer; a is the complex-valued amplitude of the received sequence, “exp” denotes the exponential function and ĵ is the imaginary unit.

[0020] The complex-valued reception sequence e(n) according to (3) refers to a single object without longitudinal extension and to an ideal receiver (with a virtually infinite reception bandwidth; real receivers blend the pulses, especially for a good signal-to-noise ratio, which will be disregarded here for the time being). In fact, there can be multiple and / or longitudinally extended objects, and an additional noise r e (n) generated, in particular by thermal noise; then the reception sequence results e(n)=sumi=1,…,I[ai⋅b(n−m0,i)⋅exp(2πj^⋅n / N⋅k0,i)]+re(n), where “sum i=1,...,I “ represents the sum function over the index i = 1,...,I ​​of the I non-extended individual objects.

[0021] From the reception sequence e(n) of the period n = 0,1,...,N-1, the discrete propagation times m 0,i and the discrete Doppler shifts k 0,i of the I objects. For the most accurate determination, i.e. the best possible separation of signal and noise and thus maximum sensitivity and range of the lidar system, the so-called optimal filtering must be used, i.e. the filtering by correlation between the reception sequence e(n) and the two-dimensional space ê m,k (n) of the possible ideal amplitude-normalized reception sequences of a single object: e^m,k(n)=b(n−m)⋅exp(j^2π⋅n / N⋅k) with m=0,…,M−1 and k=0,…,N−1; where M-1 corresponds to the largest assumed or interesting object distance, and the Doppler shift k is assumed to take any value. This results in the two-dimensional correlation E m,k to Em,k=sumn=0,…,N−1(e(n)⋅conj(e^m,k(n))=sumn=0,…,N−1(e(n)⋅b(n−m)⋅exp(−j^2π⋅n / N⋅k)), m=0,…,M−1 and k=0,…,N−1, where “conj” denotes the conjugate complex formation and the modulation sequence b(n) does not change due to its real value. The correlation E m,k has magnitude peaks (often referred to as power peaks) at the points (m,k) = (m 0,i ,k 0,i ) of objects; Fig. 4 shows the magnitude of the two-dimensional correlation for two objects of the same reception amplitude at (m 0,1 ,k 0,1 ) = (150, 3000) and (m 0,2 ,k 0,2) = (51, 500), which corresponds to their distances r1 = 150m and r2 = 51m and the radial relative velocities v1 = -31.1 m / s and v2 = 14.2m / s (in the range k = 0,...,N-1 chosen asymmetrically above, k > N / 2 is assigned negative Doppler frequencies and thus negative relative velocities; a negative relative velocity means that the object is moving away relatively). As in Fig. As can be seen in Figure 4, the two-dimensional correlation in the Doppler shifts k 0,i of the objects exhibit an approximately constant base level, from which the magnitude peaks differ by about 6 dB; this will be explained in more detail later, and a possible approach to suppressing this base level will also be presented.

[0022] So to find the distance r i and the radial relative velocity v i of objects, the magnitude peaks of the two-dimensional correlation E m,k, using only magnitude peaks that are above a respective detection threshold to distinguish them from the system noise and the above-mentioned background level of the Doppler shift of objects. According to equations (3b) and (3c), r i and v i from the positions of the amount peaks, i.e. the discrete maturities m 0,i and the discrete frequency shifts k 0,i be calculated as follows: ri=m0,i⋅c⋅Tm / 2, vi=k0,i⋅λ / (2N⋅Ts).

[0023] From a single modulation sequence, the distance and relative velocity of multiple objects can be determined directly and unambiguously. This is a significant advantage over the linear frequency modulation commonly used in coherent lidar systems with two frequency ramps whose slopes have opposite signs—where ambiguities are unavoidable when multiple objects are involved.

[0024] The calculation of this two-dimensional correlation and its subsequent evaluation take place in the digital signal processing unit 1.11. It represents a high complexity with an order of N·M·N. However, the above relationship (6) can also be considered as a discrete Fourier transform via the product e(n)·b(nm), n = 0,...,N-1, which is to be determined for each m = 0,...,M-1; the discrete Fourier transform (DFT) is calculated via the fast Fourier transform (FFT): Em,k=FFTk(e(n)⋅b(n−m)) with m=0,…,M−1, where k = 0,...,N-1 is the output dimension of the FFT, i.e. the discrete frequency, the computational effort is reduced to the order M·N·log2(N).

[0025] The reception sequence e(n) of the period n = 0,1,...,N-1 considered so far and the corresponding correlation E m,krefer to a single detection direction, i.e., the horizontal and vertical directions of a pixel. In fact, in each 50ms detection cycle, approximately 160,000 detection directions, i.e., pixels, are covered; this is typically achieved by a combination of parallel transmitter and receiver, i.e., parallel detection of pixels, and scanning, i.e., sequential detection of pixels. Parallel transmitter and receiver means that all elements 1.4-1.10 of the lidar system 1.1 are Fig. 1 (possibly apart from a common focusing and beam steering device in the transmit-receive unit 1.6) exists multiple times, e.g., 64 times. Scanning can be done, for example, by sequential switching or by continuous mechanical movement (e.g., of a mirror); with continuous scanning, it is also possible for pixels to partially overlap, i.e., the last of the N values ​​of the receive sequence e(n) of a pixel are also used as the first values ​​of the next pixel. Assuming that M = 250 (corresponding to a maximum distance of 249 m in the above design), then the FFT of length 4096 from equation (8) must be calculated 800 million times per second. Implementing this many FFT calculations is not possible using microprocessors or DSPs (Digital Signal Processors); the clock speed of such processors is typically in the range of 1 GHz, i.e., almost a complete FFT of length 4096 would have to be calculated per clock cycle. However, modern processors, even when using parallel vector processing units, can generally only perform up to 100 multiplications and additions per clock cycle, which is orders of magnitude less than the requirement for an FFT of length 4096. Therefore, these many FFTs can only be implemented using special computing logic implemented in hardware (i.e., a largely hard-wired digital computing circuit); but even this computing logic is generally associated with a large implementation effort (large chip area) and high electrical power consumption.

[0026] Therefore, an approach is proposed here that requires less computational effort. According to equation (3a), the reception sequence e generated by an object i is i (n) in the period n = 0,...,N-1: ei(n)=ai⋅b(n−m0,i)⋅exp(2πj^⋅n / N⋅k0,i) with n=0,…,N−1, ie it corresponds to a complex oscillation with the Doppler frequency k 0,i , which is set to half and pseudorandomly to zero. The Doppler frequency can be determined using a DFT or FFT.

[0027] For this purpose, the reception sequence e˜1(n)=e(n) with n=0,…,N−1 an FFT is formed: E˜1,k=FFTk(e˜1(n))=FFTk(e(n)) with n=0,…,N−1, where k = 0,...,N-1 is the output dimension of the FFT, i.e. the discrete frequency. Fig. 5, this FFT is absolute for two objects of the same reception amplitude at (m 0,1 ,k 0,1 ) = (150, 3000) and (m 0,2 ,k 0,2 ) = (51, 500), ie peaks occur at the discrete Doppler frequencies k 0,1 = 3000 and k 0,2 = 500. Since the modulation b(nm 0,i) half of the respective oscillation is set to zero in a pseudo-random manner, a noise component is generated for each object which is 10·log 10 (N / 2) = 33dB below the respective magnitude peak; due to the two objects considered with the same reception amplitude, the noise generated by them is then 30dB below the magnitude peaks. Fig. 5, it is assumed that the signals reflected by the objects are so strong that the noise generated by them due to their modulation dominates over the system noise, thus the signal-to-noise ratio is 30 dB. For weaker reflections, the signal-to-noise ratio is limited by the system noise. It should be noted here that the signal-to-noise ratio with respect to the system noise is reduced by 6 dB compared to an unmodulated received signal of the same amplitude, since only half of the signal duration is effective for the coherent integration of the FFT (in the other half, the useful signal is zero). When referenced to a received signal of the same average received power (i.e., also related to the same average transmitted power), the reduction in the signal-to-system noise ratio due to the modulation is still 3 dB.

[0028] The magnitude peaks of the FFT Ẽ 1,k are checked for a detection threshold. The frequencies k0,j of the J magnitude peaks above the detection threshold are used for further processing; typically, these magnitude peaks are seen in two adjacent FFT values ​​(since they are not - as considered in the example above - at an integer Doppler index k 0,j lie), so that by interpolation their exact position, i.e. a non-integer frequency k 0,j can be determined. These frequencies k 0,j correspond at least approximately to the Doppler frequencies k 0,i of the objects or a subset of them (for objects with very low reflectivity they may not lead to a magnitude peak above the detection threshold).

[0029] For the output dimension of the FFT, i.e., the discrete frequency, the non-symmetrical range k = 0,...,N-1 was considered above, as is generally the case; the actual relative velocities and thus Doppler frequencies can assume either sign, so that the upper range, e.g., the upper half of k = 0,...N-1, can be mapped to negative values ​​by subtracting N.

[0030] So far, we have considered the case where no window function is used for the FFT, i.e. no multiplication of the FFT input values ​​with a type of bell curve; this would only be necessary or useful if two objects with similar relative speed and significantly different reflectivity can occur at the same distance in a pixel and are to be separated. In particular, if no window function is used at the FFT input, the sensitivity at the FFT output is reduced (i.e. the detection ability of objects with weak reflectivity and / or great distance) if the Doppler index corresponding to the relative speed is not an integer, i.e. the magnitude peak is split between two neighboring FFT values. This effect can be reduced by choosing a longer FFT length than that of its input signal, i.e., zeros are appended to the input signal, a process known as zero padding.

[0031] From the FFT Ẽ 1,k the Doppler frequencies k 0,i of the objects; to determine their travel times, a second calculation step is necessary. According to equation (9), the reception sequence e generated by an object i is i (n) a term m 0,i shifted modulation sequence b(nm 0,i ), which determines the Doppler frequency k 0,i is superimposed (of course multiplied by the complex signal amplitude a i ). In order to eliminate the superimposed Doppler frequency, the reception sequence e(n) is incremented by the frequency k 0,j turned back: e˜2,j(n)=e(n)⋅exp(−2πj^⋅n / N⋅k0,j) with n=0,…,N−1 and j=1,…J.

[0032] The modified sequences ẽ 2,j (n) contain shifted modulation sequences a in the corresponding index j i b(n-mo 0,i ); contributions from objects with other Doppler frequencies k 0,i (also k 0,i ≠ k 0,j) represent a modulation sequence that is uncorrelated to b(n), since they are still correlated with the difference frequency k 0,i -k 0,j are modulated. Therefore, the sequences ẽ 2,j (n) can be correlated with the modulation sequence b(n): E˜2,j,m=sumn=0,…,N−1(e˜2,j(n)⋅b(n−m)) with m=0,…,M−1 and j=1,…J; Since the modulation sequence b(n) has the period N, this can also be described as a cyclic correlation using the cyclically shifted modulation sequence b(mod N (nm) can be understood (“mod N “ represents the modulo function to the module N): E˜2,j,m=sumn=0,…,N−1(e˜2,j(n)⋅b(modN(n−m))) with m=0,…,M−1 and j=1,…J.

[0033] In these one-dimensional correlations Ẽ 2,j,m peak amounts occur at the corresponding discrete maturities m 0,i of the objects. For the above example with the two objects at (m 0,1 ,k 0,1 ) = (150, 3000) and (m 0,2 ,k 0,2) = (51, 500) are in Fig. 6 the magnitudes of the two correlations Ẽ 2,1,m to ẽ 2,1 (n) for k 0,1 = 3000 and Ẽ 2,2,m to ẽ 2,2 (n) for k 0,2 = 500; the amount peaks are at the corresponding discrete maturities m 0,1 = 150 and m 0,2 = 51. In the general case of a non-integer discrete running time m 0,i With suitable measures (e.g., smoothing the modulation pulses in the receiver to an approximately triangular shape by optimal filtering), the magnitude peak extends over two adjacent values ​​of m, and its non-integer position can be determined by interpolation. From the index j of the correlation Ẽ 2,j,m , at which the magnitude peak occurs, and the Doppler frequency k corresponding to this j 0,j the Doppler frequency k 0,i = k 0,jof the object. In this way, peaks in the correlations Ẽ that are above a detection threshold can be used to 2,j,m the distances and radial relative velocities of objects in the respective detection direction are determined. It should be noted that contributions from objects with different Doppler frequencies k 0,i (also k 0,i ≠ k 0,j ) due to their uncorrelatedness to b(n) only lead to noise in the respective correlations, i.e. to no magnitude peaks above a detection threshold; if the reflection signals from objects are of approximately the same strength, this noise lies significantly below the magnitude peaks of interest and thus does not mask them - only if reflection signals from objects are very different, objects with a strong reflection signal can mask weak reflection signals with a different Doppler frequency via their noise.

[0034] As in Fig. As can be seen in Figure 6, the cyclic correlations Ẽ 2,j,man approximately constant base level, from which the magnitude peaks stand out by about 6 dB. This is due to the fact that the modulation sequence b(n) and thus also the reception sequences belonging to the individual objects are not free of mean values ​​after being rotated back by their Doppler frequency; their mean value is half the amplitude in the switched-on, i.e. active, state. If, at a radial relative velocity, i.e. a Doppler frequency, there are several objects in a pixel with different distances and with different reception levels, the magnitude peaks of the weaker objects are generally masked by the base level of the strongest object and are no longer detectable. To avoid this, the base level can be subtracted or eliminated; one approach is to adjust the modulation sequence used for the cyclic correlation by the mean value 0.5, i.e. the modulation sequence b_(n)=b(n)−0.5 used. The disadvantage of this is that the signal-to-noise ratio is then 3 dB worse (the correlation then no longer represents optimal filtering). Therefore, a two-stage procedure can also be used, where first the correlation according to equation (13) is applied with the actual modulation sequence b(n), and then - at least with a good signal-to-noise ratio - an approach with the eliminated base level (from the mean value of the modulation sequence) is carried out.It should also be noted that when the base level is eliminated in the correlation, noise still occurs outside the magnitude peaks, since the two factors of the correlation product (received sequence and shifted modulation sequences) are uncorrelated to each other and thus add up randomly, i.e. with a random sign; if - as in the case considered above - the period of the modulation sequence and the length of the evaluated received sequence have the same value N, then this signal-related noise is 10 log. 10 (N) = 36dB below the magnitude peak, for a non-periodic modulation sequence it would be only 10·log 10 (N / 2) = 33dB, i.e. a design with the same period of the modulation sequence and length of the evaluated reception sequence (and thus length of the correlation) is advantageous in this regard.

[0035] However, since in most cases there is only one object in a pixel with a radial relative velocity, a generally acceptable simplification is to only calculate the maximum of the respective cyclic correlations Ẽ 2,j,m to determine and observe; if this maximum lies above a detection threshold, which is preferably derived from the system noise, then an object detection occurs. The signal-to-system noise ratio for the cyclic correlations Ẽ 2,j,m 3dB better than the FFT Ẽ 1,kfor determining Doppler frequencies, since at the magnitude peaks of the correlation, the values ​​that contain only system noise but no signal are suppressed by multiplying with the shifted modulation sequence, which assumes the value zero there. These values ​​(50% of all values) are included in the FFT and thus increase the noise, but not the magnitude peak. Therefore, it is advisable to use a reduced detection threshold for the FFT (with the consequence that more magnitude peaks generated by noise are detected) and only check for a stricter detection threshold (for the desired or permissible false positive rate) during the correlation.

[0036] The only disadvantage may be that the correlation has to be calculated a little more frequently.

[0037] In the typical case of an object in a pixel, the evaluation explained above requires an FFT for the reception sequence e(n) to determine the Doppler frequency and a cyclic correlation to determine the propagation time (the reversal of the Doppler frequency in relation (12) is negligible compared to these two operations and is therefore not considered). The cyclic correlation in the time domain can also be realized by multiplying the FFTs in the frequency domain followed by an inverse FFT (inverse FFT requires the same computational effort as the FFT itself); the FFT of the modulation sequence b(n) can be determined once a priori (and then used for all pixels, provided the distance between the pixels and the period of the modulation sequence are identical), so that only the FFT of the reception sequence ẽ reversed in the Doppler frequency 2,1(n) must be determined (it should be noted that instead of turning back the Doppler frequency in the time domain, a shift in the frequency domain, i.e. the FFT of ẽ 2,1 (n) could be used, although then only integer discrete Doppler frequencies would be possible, which generally has disadvantages. Overall, therefore, the effort of three FFTs is incurred, while with the two-dimensional correlation E m,k According to equation (8), a total of M = 250 FFTs are to be expected. This results in a very strong reduction in the required computational effort to an order of magnitude that modern DSPs with parallel vector processing units can achieve. It should be noted that, despite the strong reduction in computing power, the presented approach achieves the same sensitivity as the optimal two-dimensional correlation E m,kaccording to ref. (8), since the full emitted signal energy (per pixel) is coherently effective (without integration over noise from sections without received signal).

[0038] A similar reduction in the required computational effort can also be achieved by using phase modulation and evaluation, as described in the article "Phase-Coded-Based Modulation for Coherent Lidar" by Sebastian Banzhaf and Christian Waldschmidt, published in IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 70, NO. 10, OCTOBER 2021. However, only half of the emitted signal energy (per pixel) is used for the signal-to-noise ratio, whereas the inventive approach presented here utilizes the full signal energy in the correlation, resulting in a 3 dB improvement in the signal-to-noise ratio. And, in contrast to the approach described in the above article, arbitrary overlaps can be realized for two neighboring pixels, since the modulation sequence has N uniform, pseudorandom structures over its entire period (and is not divided into two separate sections).

[0039] If the peak power in the lidar system is limited (e.g., by amplifier 1.4) or if significant electrical power is consumed (e.g., by amplifier 1.4) even when switch 1.3 is open, the approach presented here of power modulation, i.e., switching a signal on and off, can at least partially lose the above-described advantage with regard to the signal-to-system noise ratio. However, a lidar system usually has several parallel transmit / receive paths in order to realize the required number of pixels per cycle. Therefore, the power modulation can be implemented by having a switch between two transmit / receive paths, i.e., the signal is applied pseudorandomly to one of the two transmit / receive paths. This is Fig. 7; the first transmit-receive path 7.5 receives the modulation signal b(n), the second transmit-receive path 7.6 receives the inverse modulation signal inv(b(n)), where the inverse function "inv" exchanges the values ​​0 and +1. The amplifier 7.4 is now located in front of the switch 7.3 and thus has a constant output power over time. It should be noted that the unmodulated laser signal also goes into the transmit-receive paths for downconversion of the received signal, which for the sake of simplicity is Fig. 7 is not explicitly shown.

[0040] So far, a perfect switch or changeover switch has been assumed, meaning that when switched off, zero, or absolutely no signal, is present at the output. Real switches are not perfect, so that in the actually switched off state, a residual signal can still be present (which, however, is significantly lower than the full signal when switched on). For the first calculation step, i.e., determining the FFT Ẽ 1,k , this is even a slight advantage (provided the phase is maintained in the downshifted state), since the magnitude peaks then increase slightly. In the second calculation step, i.e. the determination of the correlations Ẽ 2,j,m , the dynamic range is slightly reduced, but this is not critical. Therefore, less efficient switches or toggles can also be used. In general, it should be emphasized that generating modulation with switches or toggles is a very simple and cost-effective approach.

[0041] Regarding the modulation sequence b(n), it should be noted that it can be formed not only from a pseudorandom sequence, but also from a code whose autocorrelates exhibit low side lobes—e.g., a Gold code known from the literature. This can lead to a higher dynamic range of the correlation, but only if there are no signal components (e.g., from objects with a different relative velocity) that represent noise in the signal to be correlated, and if appropriate countermeasures are taken against the basic correlation level of objects explained above.

[0042] So far, it has been assumed that the frequency to which the power modulation sequence b(n) is applied is constant, i.e., does not change. In the following, an inventive approach is considered that the frequency changes continuously in a linear manner; this is Fig. 8, where the frequency of the laser source 1.2 and thus the transmission frequency f TX (t) within a modulation period of duration T pm = N T m = 27.3µs around B = 800MHz (i.e. with slope B / T pm = 29.3MHz / µs) increases linearly. In general, the transmit frequency f TX (t): fTX(t)=fTX0+t⋅B / Tpm with Tpm=N⋅Tm.

[0043] Without Doppler shift (i.e. with initially assumed relative velocity zero) and as in Fig. As shown in Figure 9, the frequency f RX (t) of the received signal is delayed accordingly due to the propagation time t0: fRX(t)=fTX(t−t0)=fTX0+(t−t0)⋅B / Tpm, and thus compared to the transmission frequency f TX (t) shifted downwards. This transit time-related frequency shift f r amounts fr=−t0⋅B / Tpm and with running time t0 according to (1a): fr=−2r / c⋅B / Tpm; For the example of an object distance r = 150m and thus a transit time t0 = 1µs as well as the above modulation values, the transit time-related frequency shift results in f r = -29.3MHz. According to equation (17b), the propagation-dependent frequency shift f r proportional to the object distance r.

[0044] The total frequency shift and thus the frequency f e of the received signal after mixing is now composed of the Doppler shift f D according to (1b) and the above term-related share f r according to reference (17b) together: fe=fD+fr=2v / λ−2r / c⋅B / Tpm.

[0045] This is the only difference from the initially considered case of a constant transmit frequency. The modulation sequence, shifted by the propagation time, is still applied to this receive frequency, so that for the receive sequence e(n) after sampling and digitization, equation (3a) still applies, with equation (3c) now applying to the discrete receive frequency k0 instead of equation (3c): k0=N⋅Ts⋅(fD+fr)=N⋅Ts⋅2⋅v / λ−N⋅Ts⋅2⋅r / c⋅B / Tpm, and using the discrete running time m0 according to (3b) and T pm = N T m : vi=λ / (2N⋅Ts)⋅(k0,i+m0,i⋅Ts⋅B);

[0046] Thus, the evaluation approaches presented above can still be used - both the two-dimensional correlation E m,k according to (8), thus also the cost-effective two-step method via FFT Ẽ 1,k according to (11) and one-dimensional correlation Ẽ 2,j,maccording to (13), which is preferably realized with the help of two FFTs (per object). The only difference is that for the determination of the radial relative velocity v i of the respective object, the propagation time-related portion of the discrete frequency shifts k 0,i must be taken into account - with the help of the relation (19b) we get: vi=λ / (2N⋅Ts)⋅(k0,i+m0,i⋅Ts⋅B); In contrast to the original expression (7b), the discrete frequency shift k 0,i the term-related contribution -m 0,i ·T s ·B which is proportional to the discrete running time m 0,i is.

[0047] Reasons for a superimposed, preferably at least approximately linear frequency modulation can be, firstly, a continuous scanning of the detection direction via frequency change, secondly, a more accurate distance and / or relative speed measurement (particularly when using different signs for the frequency change) and thirdly, when using a real-valued mixer, the at least partially unambiguous determination of the sign of the received frequency.

[0048] Finally, the following should be noted: Based on the above application examples, the inventive considerations and designs presented can be easily transferred to general measurements and parameter interpretations, i.e., they can also be applied to other numerical values. Therefore, general parameters are often given in formulas and figures. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited non-patent literature

[0000] Phase-Coded-Based Modulation for Coherent Lidar” by Sebastian Banzhaf and Christian Waldschmidt, published in IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 70, NO. 10, OCTOBER 2021 [0005, 0038]

Claims

[1] Coherent lidar system for environmental detection, which - transmits a power-modulated signal which is realized by irregular, in particular pseudo-random, switching on and off, whereby the switching off is characterized by a complete switching off of the power or only by a significant power reduction, and the times of switching on and off form a subset of an equidistant time grid, - receives the signals reflected back from objects, which are delayed compared to the transmitted signal by the distance-dependent propagation time and advanced in frequency by the relative speed-dependent Doppler effect, and converts them into a low-frequency signal by mixing and digitizes them into a reception sequence, and - from this reception sequence, the variable dimensions of time shift and frequency shift of signals reflected from objects are determined in digital signal processing means, whereby a two-dimensional correlation filtering is used for the variable dimensions of time shift and frequency shift of signals reflected from objects, preferably implemented using a hard-wired digital circuit, or to reduce the required computational effort - a discrete Fourier transform is calculated, preferably using a fast Fourier transform, over the values ​​of the reception sequence, possibly extended with zeros, - the respective frequencies are determined from values ​​of these discrete Fourier transforms, in particular from magnitude peaks lying above a first detection threshold, - the reception sequence is turned back in frequency for the respective frequencies, - a correlation is determined between the sequence thus generated and the modulation sequence, in particular formed from the switching values ​​1 and 0, or the modulation sequence adjusted by its mean value, and - the respective time shift and thus object distance are determined from values ​​of this respective correlation, in particular from magnitude peaks lying above a second detection threshold, and the radial relative speed of the respective object is determined from the respective frequency. [2] Lidar system according to claim 1, wherein the power modulation is realized by a switch (7.3) between two transmission paths, which thus have inverse power modulation to each other. [3] Lidar system according to one of the above claims, in which the first detection threshold for magnitude peaks of the discrete Fourier transform is less above the noise than the second detection threshold for magnitude peaks of the respective correlation. [4] Lidar system according to one of the above claims, in which the power modulation sequence is repeated periodically, wherein in particular during continuous scanning of the laser beam the cyclical property of the modulation and reception sequence is exploited and detection in respectively different directions is realized. [5] Lidar system according to one of the above claims, in which the laser beam scans continuously and overlapping sections of a long, possibly periodic power modulation sequence are used for the successive detection directions. [6] Lidar system according to one of the above claims, in which the laser frequency changes continuously at least in sections, preferably with an at least approximately linear course, so that the entire modulation is composed of a line modulation and a frequency modulation. [7] Lidar system according to claim 6, in which it is taken into account that the frequency is shifted both by the relative speed-dependent Doppler effect and by the distance-dependent transit time due to the linear frequency change, in particular by determining the object distance from the determined time shift and the radial relative speed of the object from the determined frequency shift less its contribution caused by the time shift. [8] Lidar system according to claim 6 or 7, in which a sign-correct determination of the reception frequency and thus the unambiguous determination of the relative speed of objects when using a real-valued mixer is realized by taking into account that, in particular for more distant objects, only a relative speed hypothesis is possible or at least more plausible due to the known propagation time-dependent component of the frequency shift. [9] Lidar system according to claim 6 or 7, in which a sign-correct determination of the reception frequency and thus the unambiguous determination of the relative speed of objects when using a real-valued mixer is realized by varying the steepness of the frequency change and recording and evaluating data for an object at different steepnesses and using the fact that the two frequency shift effects of relative speed and transit time are then in different relations to each other, in particular characterized by that the sign, but not the amount of the steepness of the frequency change is varied, so that the two frequency shift effects of relative speed and transit time add up with different signs and thus the sign of the reception frequency can be determined. [10] Lidar system according to one of claims 6-9, in which the frequency change is used to change the beam direction, in particular in at least partially continuous manner, in order to be able to acquire data for several pixels of different directions.

Images