Method for three-dimensional positioning and height control of screed based on coupling of GNSS-RTK and domain laser

Abstract

The application discloses a GNSS-RTK and domain laser coupling-based three-dimensional positioning and elevation control method for a screed, and relates to the technical field of automatic control.The method is executed in a periodic sequence by the following steps: establishing a coding reference, using a plurality of transmitting ends to use a non-overlapping subcarrier number set;installing a linear array laser receiver on both sides of the screed, receiving and decoding to obtain a reference plane number and a phase step count value, and constructing a particle set containing multiple modes, each mode corresponding to a GNSS-RTK solution state;switching between different modes according to the solution state label of the GNSS-RTK to generate a three-dimensional positioning stream;generating a control target according to the three-dimensional positioning stream and a digital design surface model to complete closed-loop control.Compared with a traditional single GNSS or single-plane laser control system, the application has higher stability and precision under multipath interference, complex terrain, dynamic construction and slope transition conditions.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology, and in particular to a method for three-dimensional positioning and elevation control of a screed based on GNSS-RTK and domain laser coupling. Background Technology

[0002] In road paving, airport runway construction, and large-scale flooring construction, the elevation and attitude control of the screed are crucial factors determining the smoothness, thickness uniformity, and structural density of the surface. Current elevation control primarily relies on GNSS-RTK positioning systems, single-laser plane control systems, or mechanical tracking leveling devices. While these systems have advanced construction automation at different stages, they still have significant shortcomings in areas such as handling multi-source environmental interference, short-term obstruction, dynamic response, and adaptability to complex terrain.

[0003] In traditional GNSS-RTK positioning and control systems, the elevation of the screed is primarily calculated from real-time coordinates provided by the GNSS receiving antenna. Such systems offer high absolute accuracy in open areas, reaching centimeter-level precision. However, when construction sites are obstructed by steel structures, experience multipath reflections from construction vehicles, or have limited antenna fields of view, the multipath error of the GNSS signal increases significantly. The calculation state frequently switches between fixed, floating, and unavailable states, leading to elevation signal jumps or short-term loss. Existing systems typically smooth these jumps using delay filtering or low-pass filtering, but this reduces the system's dynamic response, causing lag when the screed enters uphill or concave areas, directly affecting the consistency of the pavement thickness. Furthermore, GNSS positioning errors are primarily unstable in the vertical direction, while the accuracy requirements for elevation in construction control are usually higher than those for horizontal position. Therefore, relying solely on GNSS signals is insufficient to meet the requirements of high-grade roads or highly flat sites. Summary of the Invention

[0004] In view of this, the present invention provides a 3D positioning and elevation control method for a screed based on GNSS-RTK and domain laser coupling. The method utilizes an RBPF line-of-sight mode switching structure and a particle Kalman joint tracking mechanism to dynamically switch between direct line-of-sight, weak multipath, strong multipath, and occlusion modes according to the GNSS solution status, ensuring continuous output even under signal interference, occlusion, or reflection conditions. Simultaneously, the domain laser employs a phase- and frequency-encoded multi-plane reference to avoid interference between multiple transmitters, maintaining high accuracy in elevation calculations in complex cross-sections and variable slope scenarios. The control layer achieves independent adjustment of average elevation and cross slope through adaptive feedforward and PID dual-channel decoupling, and drives the electro-hydraulic servo closed loop with millisecond-level cycles, resulting in rapid and smooth screed lifting and lowering responses.

[0005] The technical solution adopted in this invention is as follows:

[0006] A method for three-dimensional positioning and elevation control of a screed based on GNSS-RTK and domain laser coupling is proposed. This method comprises the following steps executed sequentially: establishing a coded reference; deploying at least two domain laser transmitters in the construction domain, each transmitter configured with a multi-plane reference; assigning a unique phase sequence number and a unique frequency subcarrier number to each reference plane; using non-overlapping subcarrier number sets for multiple transmitters; installing linear array laser receivers on both sides of the screed to receive and decode the reference plane number and phase step count; combining this with a pre-set reference plane model; and generating elevation readings for the left and right cutting edges of the screed through geometric solving; simultaneously, acquiring the aircraft's attitude and displacement using an airborne GNSS-RTK module and attitude and displacement measurement unit. Motion information; Construct a particle set containing multiple modes, each mode corresponding to a GNSS-RTK solution state. Each particle has a built-in Kalman filter to track the continuous state including the three-dimensional position and attitude of the ironing plate; Switch between different modes according to the GNSS-RTK solution state label, and route the GNSS-RTK planar observation and the elevation readings of the left and right edges of the ironing plate to the Kalman filter of the corresponding particle according to the current mode, perform condition prediction and condition update, and generate a three-dimensional positioning flow; Generate a control target based on the three-dimensional positioning flow and the digital design surface model, calculate the elevation error and cross slope error, generate a total current command through adaptive feedforward and PID controller, drive the electro-hydraulic servo mechanism, and complete the closed-loop control.

[0007] Furthermore, in the step of establishing the coding reference, the phase sequence allocated to each reference plane is composed of uniform steps and has a length of not less than 8; the interval between frequency subcarriers is fixed and the subcarrier table length is not less than 8; the domain laser transmitter rotates the reference plane with a fixed-length frame structure and outputs the data in the order of the phase sequence number.

[0008] Furthermore, the steps for generating the elevation readings of the left and right cutting edges of the ironing plate through geometric solution include: a) establishing a beam direction table and beam starting point for each linear laser receiver, generated by a combination of installation parameters and real-time attitude; b) using a numerical stepping line to search for and locate the intersection point between the beam and the reference plane, stepping outward from the starting point along the beam direction with a step size of no more than 5 mm. When the vertical coordinate of the sampling point is greater than or equal to the elevation value of the reference plane for the first time, the enclosed interval formed by the current sampling point and the previous sampling point is recorded; c) performing bisection refinement on the enclosed interval, successively halving the step size until the step size is no more than 1 mm, taking the sampling point of the last refinement as the intersection point, and using the vertical coordinate of the intersection point as the elevation reading of the cutting edge of the ironing plate on the corresponding side.

[0009] Furthermore, the constructed modes are four types: direct-view mode, weak multipath mode, strong multipath mode, and occlusion mode.

[0010] Furthermore, according to the rules for switching modes of GNSS-RTK's solution status label and multipath indicator label: a) if the solution status label is fixed, enter line-of-sight mode; b) if the solution status label is floating and the multipath indicator label is weak, enter weak multipath mode; c) if the solution status label is floating and the multipath indicator label is strong, enter strong multipath mode; d) if the solution status label is unavailable, enter occlusion mode.

[0011] Furthermore, according to the current mode routing rules for GNSS-RTK planar observations and the elevation readings of the left and right cutting edges of the screed, the following rules apply: a) In line-of-sight mode, both the planar observations and the elevation readings of the left and right cutting edges of the screed are simultaneously fed into the Kalman filter of the corresponding particle; b) In weak multipath mode, both the planar observations and the elevation readings of the left and right cutting edges of the screed are simultaneously fed into the Kalman filter of the corresponding particle; c) In strong multipath mode, only the elevation readings of the left and right cutting edges of the screed are fed into the Kalman filter of the corresponding particle, and the hydraulic cylinder stroke change and vehicle speed are added for displacement extrapolation; d) In obstructed mode, only the elevation readings of the left and right cutting edges of the screed are fed into the Kalman filter of the corresponding particle, and the hydraulic cylinder stroke change and heading angle change are added for attitude extrapolation.

[0012] Furthermore, in the step of generating the total current command, the adaptive feedforward is implemented as follows: read the vehicle speed and the slope of the design surface in the current travel direction, call the feedforward function to obtain the feedforward current command; the feedforward function is pre-set with at least 3 sets of curves according to the equipment model and material properties at the factory, and selects one of them according to the vehicle speed range.

[0013] Furthermore, the steps for calculating errors and generating total current commands also include: subtracting the estimated elevations of the left and right blade edges within the three-dimensional positioning flow from the control target to obtain the left channel error and the right channel error; calculating the sum and difference of the two channel errors to generate the average elevation channel error and the cross slope channel error, and establishing a sum-difference mapping matrix between the two channels; performing PID regulation on the average elevation channel error and the cross slope channel error respectively, and adding them to the feedforward current command, then converting them into current commands for the left and right valves through a sum-difference backcompilation matrix to achieve decoupled control of the average elevation and cross slope.

[0014] Furthermore, a spatiotemporal alignment step is included before execution: using the 1PPS signal of GNSS-RTK as the system time reference, synchronization pulses are allocated to each device; each device records a 64-bit cycle count along with the data during sampling; after receiving all the data, the industrial controller aligns each data packet to the same control cycle according to the cycle count.

[0015] Furthermore, the preset reference plane model is generated in the following way: at least three calibration points with known elevations are set on the construction site, and calibration prisms are used to point to the reference plane in sequence. The industrial controller reads the east and north directions and their elevation values ​​of the calibration points, and at the same time reads the phase step count value of the linear array laser receiver for the reference plane. The built-in plane fitting process is called to generate a set of reference plane model coefficients, which are then bound and stored with the reference plane number.

[0016] By adopting the above technical solution, the present invention achieves the following beneficial effects: The present invention couples the GNSS-RTK plane solution with the domain laser multiplane elevation solution to realize three-dimensional real-time positioning and elevation closed-loop control of the screed. Firstly, regarding positioning accuracy and robustness, an RBPF line-of-sight mode switching structure is adopted. GNSS visibility and multipath states are represented by particle sets, and a Kalman filter is built into each particle to track continuous states, achieving smooth switching under conditions of signal loss, reflection interference, and obstruction. This enables the system to maintain centimeter-level horizontal and millimeter-level vertical stability in complex construction environments. Secondly, in the elevation calculation stage, the domain laser transmitter adopts a phase and frequency encoded multiplane structure to avoid mutual interference and signal confusion between different planes. The receiver achieves high-precision elevation reading recovery under dynamic conditions through linear laser fine-step search and bisection refinement intersection calculation, thereby eliminating the systematic deviations of single-plane laser systems in the control of curved surfaces, slopes, and intersections. Furthermore, the system's spatiotemporal alignment mechanism adopts a unified sampling reference of 1PPS signals from GNSS-RTK, enabling synchronization of all observation sources within milliseconds and fundamentally solving the servo lag caused by multi-source delays. The control layer achieves independent control of average elevation and cross slope through adaptive feedforward based on vehicle speed and slope along the route, and decoupled regulation using dual-channel PID control. Current output is distributed to the left and right electro-hydraulic servo valves via a sum-difference back-compilation matrix, resulting in smooth and stable lifting and lowering response of the screed. Compared to existing systems relying on a single GNSS or single-plane laser reference, this invention maintains continuous elevation output under conditions of signal distortion, multipath interference, slope transition, and high-speed dynamics, avoiding the problems of traditional filtering delays and mechanical lag. The system can operate directly under different material and temperature conditions without relying on field learning, exhibiting strong versatility and field feasibility. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the layout and reference plane geometry of the domain laser emitting end in an embodiment of the present invention;

[0018] Figure 2 This is a schematic diagram illustrating the geometric solution principle of the intersection point of the beam and the reference plane in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the experimental curves of horizontal positioning accuracy under different GNSS-RTK modes in the embodiments of the present invention;

[0020] Figure 4 This is a schematic diagram of the step response characteristic curve of the elevation PID controller in an embodiment of the present invention. Detailed Implementation

[0021] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0022] Any feature disclosed in this specification, unless otherwise stated, may be replaced by other equivalent or similar features. That is, unless otherwise stated, each feature is merely one example of a series of equivalent or similar features.

[0023] A method for three-dimensional positioning and elevation control of a screed based on GNSS-RTK and domain laser coupling is proposed. This method comprises the following steps executed sequentially: establishing a coded reference; deploying at least two domain laser transmitters in the construction domain, each transmitter configured with a multi-plane reference; assigning a unique phase sequence number and a unique frequency subcarrier number to each reference plane; using non-overlapping subcarrier number sets for multiple transmitters; installing linear array laser receivers on both sides of the screed to receive and decode the reference plane number and phase step count; combining this with a pre-set reference plane model; and generating elevation readings for the left and right cutting edges of the screed through geometric solving; simultaneously, acquiring the aircraft's attitude and displacement using an airborne GNSS-RTK module and attitude and displacement measurement unit. Motion information; Construct a particle set containing multiple modes, each mode corresponding to a GNSS-RTK solution state. Each particle has a built-in Kalman filter to track the continuous state including the three-dimensional position and attitude of the ironing plate; Switch between different modes according to the GNSS-RTK solution state label, and route the GNSS-RTK planar observation and the elevation readings of the left and right edges of the ironing plate to the Kalman filter of the corresponding particle according to the current mode, perform condition prediction and condition update, and generate a three-dimensional positioning flow; Generate a control target based on the three-dimensional positioning flow and the digital design surface model, calculate the elevation error and cross slope error, generate a total current command through adaptive feedforward and PID controller, drive the electro-hydraulic servo mechanism, and complete the closed-loop control.

[0024] In one specific implementation, the construction unit first demarcates a working corridor along the paving direction. The width is determined based on a combination of the paver's nominal width and safety distance, typically ranging from 8 to 14 meters. Stable base locations are selected on both sides of the working corridor to install domain laser emitter supports. The installation height is preferably 3 to 6 meters to create continuous reference coverage without obstruction by vehicles or material piles. The spacing between two domain laser emitters along the line direction is preferably 40 to 90 meters. The advantages of this layout are: when the height is sufficient, the reference plane formed by the rotating laser passes over the working corridor, and the linear laser receiver falls into a stable energy zone; the spacing along the line, controlled at 40 to 90 meters, ensures that at least one domain laser emitter is visible during turns or short-term obstruction, thus maintaining reference continuity.

[0025] Each domain laser transmitter has a built-in set of reference planes, generated by a rotation mechanism and a pitch mechanism. The rotation mechanism drives the beam scanning at a constant speed, while the pitch mechanism positions the optical plane at a fixed angle, forming an optical plane with a preset slope or horizontal orientation. To achieve identification with minimal computation at the receiving end, each reference plane is bound to a phase sequence number and a frequency subcarrier number, ensuring uniqueness across the entire construction domain. The practical advantage of this combination is that the receiving end only needs to perform phase step counting on the corresponding subcarrier channel to complete reference plane identification within a single cycle, avoiding the uncertainties caused by cross-channel comparisons.

[0026] Several phase step levels are set for each reference plane, with consistent intervals between step levels. The sequence length is preferably 16, 24, or 32. When the sequence length is 16, 16 consecutive phase steps are output within a single frame; when 24 or 32 is selected, the number of steps within the frame is extended accordingly. The value of the uniform stepping method lies in the fact that the receiver can recover the phase step count value by counting on the rising edge, without relying on complex interpolation or matching processes, and has stronger resistance to slight mechanical jitter and short-term blockage. When assigning sequence numbers, different reference planes are assigned to non-repeating numbering intervals. For example, when the number of multi-plane references of a domain laser transmitter is 4, they can be assigned as 1, 2, 3, and 4 in sequence. When another domain laser transmitter is operating in the same site, it uses 5, 6, 7, and 8. The purpose of this is to allow the receiver to directly map the phase step count value to the reference plane number after parsing it, without the need for cross-device fusion.

[0027] refer to Figure 1 ,like Figure 1As shown in the diagram, the geometric principle diagram of the domain laser transmitter layout and reference plane illustrates the overall system layout of this invention. In the diagram, the ground baseline is located at the bottom of the coordinate system, serving as a reference benchmark for elevation measurement. Domain laser transmitter 1 and domain laser transmitter 2 are respectively installed on the left and right sides of the work corridor. Specifically, domain laser transmitter 1 is fixedly installed by a bracket, with the bottom of the bracket fixed to the ground baseline and the transmitter body installed on top. The preferred installation height H is 4.0 meters. This installation height is chosen based on the following considerations: first, to ensure that the reference plane formed by the rotating laser passes over the work corridor without being obstructed by vehicles and material piles; second, to ensure that the linear array laser receiver falls within a stable energy coverage area. Two reference planes radiate from the domain laser transmitter 1 to the right. Reference plane 1 is horizontal, bound to phase sequence number 1 and frequency subcarrier f1; reference plane 2 has a 2% slope, bound to phase sequence number 2 and frequency subcarrier f2. The reference planes form optical planes through rotating laser scanning, and the propagation direction is indicated by dashed arrows in the diagram. The domain laser transmitter 2 is deployed similarly to transmitter 1, located on the right side of the work corridor, and also uses an installation height of 4.0 meters. The distance L between the two transmitters is 67.0 meters, preferably between 40 and 90 meters, to ensure that at least one transmitter is visible during turns or short-term obstruction. Domain laser transmitter 2 radiates to the left onto reference planes 3 and 4, respectively, and binds phase sequence numbers 5 and 6 and frequency subcarriers f9 and f10. It should be noted that the two transmitters use non-overlapping subcarrier number sets: transmitter 1 uses f1 to f4, transmitter 2 uses f9 to f12, leaving f5 to f8 in between as an isolation zone to improve channel separation. Figure 1 The paver body and screed are shown in the middle. The paver body has a rectangular structure with a GNSS-RTK antenna mounted on top for receiving satellite positioning signals. The screed is located at the bottom of the paver, maintaining a certain gap with the ground. Linear laser receivers are installed on the left and right sides of the screed, their installation positions being at the same longitudinal section as the screed's cutting edge. The left linear laser receiver receives laser signals from transmitter 1, with the light path indicated by a red dashed line in the diagram; the right linear laser receiver receives laser signals from transmitter 2. The working corridor width W is marked on... Figure 1 At the bottom, the example value is 10.0 meters. This width is preferably between 8 and 14 meters, determined by a combination of the paver's nominal width and safety distance. The diagram also indicates how the distance L between the two emitters is measured, marked by connecting the centers of the two emitters with a horizontal dashed line.

[0028] The frequency subcarrier numbering is planned using an equally spaced division strategy. The available RF or optical modulation frequency range is divided into several subcarrier channels, preferably 16 or 32, with a fixed channel spacing. Each domain laser transmitter occupies one subcarrier number set, the size of which is consistent with the number of reference planes for that transmitter. For example, the set for the first domain laser transmitter can be 1 to 4, the set for the second domain laser transmitter can be 9 to 12, leaving 5 to 8 as a gap. The reason for setting the gap is to improve the channel separation under near-far power difference conditions and to avoid the receiver's influence on the far-far channel from the high-energy near-far channel. With a fixed-interval subcarrier number set, the receiver can configure a constant bandwidth bandpass processing flow for each channel, reducing implementation complexity and improving stability. The frame structure and timing adopt a fixed frame length and fixed time slot method. Each domain laser transmitter outputs a phase step sequence according to the reference plane numbering order, with the frame length selected within the range of 20 milliseconds to 40 milliseconds, and the duration of a single phase step remains consistent. The advantage of a fixed frame length is that the receiver can use a stable sampling window, avoiding boundary effects caused by frame length drift. Consider one feasible configuration: a frame length of 32 milliseconds, a sequence length of 16, and a single-step duration of 2 milliseconds. In this configuration, any reference plane outputs at the exact same rhythm within any frame, thus ensuring that the phase step counter on the receiver keeps rolling continuously between frames.

[0029] The domain laser transmitter uses a fixed power and beam divergence angle configuration. This fixed configuration ensures a stable intensity curve for the receiver across different construction sections, avoiding phase abrupt changes caused by automatic gain variations. The power level is selected on-site via DIP switches or a human-machine interface, typically set to 3 or 5 levels. The divergence angle selection follows a coverage priority principle, ensuring that, given a fixed installation height and corridor width, the reference plane covers the effective incident range of the linear array laser receivers on both sides of the ironing plate. For example, with an installation height of 4 meters and a corridor width of 10 meters, a medium divergence angle setting provides full coverage and redundancy. The implementation process of using non-overlapping subcarrier number sets for multiple domain laser transmitters includes three stages: set division, equipment binding, and on-site verification. The subcarrier numbering is completed before deployment, assigning each device a consecutive number based on the number of devices on site. Device binding is performed after power-on, by selecting the subcarrier number set in the menu through the local HMI and writing it to non-volatile memory. On-site verification is performed after device installation. The receiver switches to observation mode and checks channel by channel to confirm whether the energy peak falls within the corresponding numbered set. If energy leakage between adjacent sets exceeds a preset threshold, the set interval between the two devices is increased, for example, changing 1 to 4 and 9 to 12 to 1 to 4 and 13 to 16. This process remains simple and repeatable, allowing the construction team to quickly complete the configuration within the actual project timeline.

[0030] The mapping relationship between a unique phase sequence number and a unique frequency subcarrier number assigned to each reference plane is written into the configuration table once. The fields of the configuration table include the reference plane number, phase sequence number, frequency subcarrier number, rotation speed setting, and elevation angle setting. After writing, the domain laser transmitter outputs each reference plane sequentially within each frame period according to the configuration table. The receiver only needs to read the same version information from the configuration table to directly query the reference plane model in the subsequent geometric solution stage using the reference plane number as an index, avoiding delays caused by temporary matching. The combination of phase sequence number and frequency subcarrier number adopts a unidirectional mapping, meaning that the same combination is always used for the same reference plane on the same device, without switching during operation. This arrangement is to avoid discontinuity in recognition caused by changes in the combination on the receiver side in a short period of time. During construction, ambient light, dust, and mechanical vibration can cause short-term energy fluctuations. A fixed combination allows the receiver's decoder to achieve steady-state operation with minimal internal states, thereby improving the stability of the screed elevation readings.

[0031] In another implementation, where the site is relatively long and narrow with a small turning radius, more reference planes can be configured on each domain laser transmitter to reduce the number of devices, for example, increasing from 4 to 8, and the frame length can be increased from 32 milliseconds to 64 milliseconds to maintain the same single-step duration. When the number of reference planes increases, the allocation of phase sequence numbers and frequency subcarrier numbers follows an interleaved principle; that is, adjacent reference planes are adjacent in sequence number but separated by at least one channel in subcarrier number. The advantage of the interleaved principle is that when the receiver occasionally observes a decrease in the intensity of a certain channel, the reference plane can be quickly recovered in neighboring channels, reducing the impact of short-term fluctuations on the geometric solution entry point.

[0032] In another implementation, the construction domain has numerous obstructions and varying elevations. In this case, the installation height of the domain laser transmitter is increased to 5 to 6 meters, and the interval of the frequency subcarrier number sets is expanded; for example, the first device uses numbers 1 to 4, and the second device uses numbers 13 to 16. With the increased installation height, the effective coverage area of ​​the reference plane expands. The expanded frequency number set interval further reduces the sidelobe effect generated by strong near-end signals at the receiving side. The combination of these two factors can maintain clear separation of the reference plane in complex obstructed environments.

[0033] In another implementation, to reduce reliance on the long-term operation of the rotating mechanism, some reference planes can be converted to fixed-attitude output. Fixed-attitude output maintains the output by locking the pitch mechanism at a specified angle and rotating it at a low speed, while still using the binding relationship between phase sequence numbers and frequency subcarrier numbers. The advantage of fixed-attitude output is higher long-term stability and lower maintenance workload. For construction sections requiring cross slope settings, one of the reference planes can be set to an attitude with a fixed slope. The receiving side directly obtains the reference plane number after decoding, thereby reducing conversion steps when the control layer generates cross slope targets.

[0034] In all implementations, the configuration and operation process provides a clear sequence for on-site engineers. First, a site survey is conducted to determine the installation height and spacing. Then, after powering on the equipment, a subcarrier number set is selected and written. Next, the phase sequence number is entered and saved for each reference plane on the transmitter. Subsequently, the receiver enters the observation interface and confirms whether the energy peak value falls into the corresponding set for each channel. After confirmation, continuous output is initiated, and the receiver can count the phase step count value and recover the reference plane number on the designated channel within a single control cycle. Since the combination of phase sequence number and frequency subcarrier number remains fixed throughout the operation, the receiver repeats the same analytical process in subsequent cycles, and the geometric solution entry stably obtains the identification and time sequence of the corresponding reference plane. Using the above-described implementation of establishing the coding reference, physical isolation is achieved between domain laser transmitters through non-overlapping subcarrier number sets, and clear differentiation of reference planes in the time dimension is achieved within a single device through uniformly stepped phase sequence numbering. Therefore, the receiver can stably recover the reference plane number and phase step count value with a fixed bandpass processing flow and fixed counting logic. The entire process does not rely on complex threshold learning or data-driven weight adjustment; construction units only need to allocate and select templates according to the configuration table to implement the project. With at least two domain laser transmitters deployed at an installation height range of 3 to 6 meters, the working corridor achieves coverage redundancy within a width of 8 to 14 meters. Through a combination of phase sequence numbers of length 16 or 32 and equally spaced frequency subcarrier numbers, the receiver-side identification is completed within a single frame period, thus providing a stable coding reference for subsequent generation of left and right blade elevation readings for the ironing plate, three-dimensional positioning stream output, and electro-hydraulic servo closed-loop control.

[0035] In one embodiment, a linear array laser receiver is installed on both the left and right sides of the ironing plate. Each linear array laser receiver is arranged laterally perpendicular to the direction of travel, with its installation height from the ground roughly equal to the height of the ironing plate's cutting edge. Its front-to-back position is maintained at the same longitudinal cross-section as the cutting edge, ensuring a simple geometric relationship between the incident light path and the cutting edge, and clearly identifying the source of error. The effective photosensitive length of the linear array laser receiver is preferably between 120 mm and 240 mm, the pixel pitch is preferably between 0.5 mm and 1.0 mm, and the internal sampling frequency is preferably between 10 kHz and 50 kHz. These lengths and pitches are chosen because, under conditions of construction dust and slight tilt, a longer photosensitive length and denser pixel arrangement can maintain a stable energy distribution, avoiding phase distortion caused by critical incidence. Each linear array laser receiver has a pre-set set of parallel channels corresponding one-to-one with the frequency subcarrier numbers. The number of channels is consistent with the number of subcarrier numbers used on-site, commonly 16 or 32. Each channel undergoes bandpass processing, envelope extraction, quadrature demodulation, and phase unrolling sequentially to obtain a continuous phase curve. Then, step edges are identified on the time axis of the phase curve using a fixed threshold; the edge count value is the phase step count value. The fixed threshold is used because the transmitter operates with constant power and a constant phase step rhythm, ensuring stable and repeatable edge positions at the receiver under a fixed threshold. Stable counting results can be obtained without online learning during construction.

[0036] During decoding, each channel generates only one data record per frame period. This record includes the reference plane number, channel number, phase step count value, and the channel's average intensity. The reference plane number is recovered using a direct mapping between the channel number and the phase sequence number. This mapping table is written during equipment deployment, with the format being channel number corresponding to reference plane number; for example, channel 1 corresponds to reference plane 1, channel 2 to reference plane 2, and so on. Within the same laser transmitter domain, the reference plane numbers are output sequentially over time. The receiver performs separation in the channel dimension and step counting in the time dimension, thus determining the reference plane number and phase step count value within a single control cycle. This direct reference plane number-to-channel number approach eliminates the need for cross-channel comparisons and complex pattern recognition, reducing the probability of false triggering in environments with dust, reflections, and vehicle vibrations.

[0037] To use the reference plane identification results for elevation readings, a preset reference plane model is loaded into the industrial controller on the receiving side, and the spatial position and observation direction of the linear array laser receiver's photosensitive center are precisely determined. The determination of the spatial position relies on installation parameters and the real-time output of the airborne GNSS-RTK module and attitude and displacement measurement unit. Installation parameters include the fixed three-dimensional offset of the linear array laser receiver relative to the aircraft's coordinate system reference point and its fixed installation orientation. These quantities are given during entry calibration, with typical offsets in the range of tens of centimeters. The airborne GNSS-RTK module outputs east, north, and elevation values ​​at 10 Hz or 20 Hz, along with a solution status label. The attitude and displacement measurement unit outputs heading angle, roll angle, pitch angle, and hydraulic cylinder stroke at 100 Hz or 200 Hz. The industrial controller aligns the data with the time reference of the airborne GNSS-RTK module, extracting the latest set of synchronized data in each control cycle. The spatial position of the photosensitive center of the linear laser receiver is obtained in the following order: First, the installation offset is rotated in the aircraft coordinate system according to the heading angle to align with the direction of travel. Then, the vertical and horizontal tilt are corrected according to the roll and pitch angles. Finally, the corrected offset is superimposed on the reference point coordinates provided by the airborne GNSS-RTK module to obtain the east, north, and elevation values ​​of the photosensitive center of the linear laser receiver. The observation direction is determined using the same angular sequence. The fixed installation orientation of the linear laser receiver is rotated about the heading angle, roll angle, and pitch angle respectively to obtain the emission direction pointing towards the reference plane in the construction coordinate system. Using this sequence of heading first, then roll, and then pitch ensures that the effects of vehicle turning and body roll are independent, facilitating numerical monotonicity and convergence.

[0038] Combining the spatial position and observation direction of the reference plane model and the linear laser receiver, the industrial controller performs geometric solutions on both the left and right sides in each control cycle to generate elevation readings for the left and right cutting edges of the ironing plate. The geometric solution employs a combination of numerical stepping and bisection refinement, as follows: Starting from the photosensitive center of the linear laser receiver, the system advances outward along the observation direction with a fixed step size. The preferred step size is 5 mm. Upon reaching each sampling point, the east and north coordinates of that point are taken, and the reference plane elevation value at that location is obtained through the query interface of the reference plane model. The vertical coordinates of the sampling point are compared with the reference plane elevation value. When the vertical coordinates of the sampling point are first greater than or equal to the reference plane elevation value, it indicates that the beam has entered the region above the plane from below or within the plane. At this point, the interval between this sampling point and the previous sampling point is recorded. This enclosing interval is bisection refined, with the interval length halved successively and the comparison repeated until the interval length is no greater than 1 mm. The sampling point located on the reference plane at the time of the last refinement is taken as the intersection point. The vertical coordinates of the intersection point are the elevation readings of the corresponding side ironing plate cutting edges. The reason for adopting a rapid 5mm step size followed by 1mm precision refinement is that the distance scale at the construction site is relatively large. A coarse-to-fine approach balances real-time performance and accuracy within a single cycle. The large step size quickly locates the approximate intersection point, and the binary refinement ensures the final reading meets millimeter-level requirements, with a fixed number of iterations and controllable runtime. The same process is performed on both sides, obtaining the elevation readings of the left and right screed blade edges. If both sides have readings in the same control cycle, the industrial controller can simultaneously calculate the average elevation and cross slope angle quantization values. These quantization values ​​are used for target generation and channel arrangement in subsequent control.

[0039] Simultaneously, the airborne GNSS-RTK module and attitude and displacement measurement unit provide aircraft pose and motion information, serving as a common basis for receiver-side geometric solutions and trajectory extrapolation. The east and north coordinates of the airborne GNSS-RTK module are used to determine the horizontal position of the aircraft reference point in the construction coordinate system, while the elevation values ​​are used to estimate the vertical relationship of the aircraft reference point relative to the reference plane. Solving status labels are used as the basis for mode switching; labels are commonly categorized as fixed, floating, and unavailable. The heading angle of the attitude and displacement measurement unit is used to determine the vehicle's direction of travel, while the roll and pitch angles are used to correct the observation direction of the linear array laser receiver. The hydraulic cylinder stroke characterizes the instantaneous lifting and lowering of the screed. Vehicle speed is used to quantify the displacement extrapolation magnitude to be performed within a control cycle. After aligning the above information with the same time base, geometric solutions and continuous state tracking form a closed loop in each control cycle.

[0040] See Figure 2The diagram illustrating the geometric solution principle of the intersection point of the beam and the reference plane details the calculation process for the elevation reading of the ironing plate edge in this invention. A two-dimensional coordinate system is established in the diagram, with the horizontal axis representing the eastward coordinate axis and the vertical axis representing the elevation axis. The reference plane is represented by a thick black dashed line, which has a 2% slope, tilting from the lower left to the upper right. The photosensitive center of the linear laser receiver is marked as point O, located slightly to the left of the center of the coordinate system. The three-dimensional position of point O is determined through the following steps: First, the airborne GNSS-RTK module outputs the eastward, northward, and elevation values ​​of the aircraft reference point (i.e., the GNSS antenna position); second, based on the installation offset of the linear laser receiver relative to the GNSS antenna, which is given during the approach calibration; then, combining the heading angle, roll angle, and pitch angle provided by the attitude and displacement measurement unit, the installation offset is rotated; finally, the transformed offset is superimposed on the GNSS antenna position to obtain the spatial coordinates of the photosensitive center O. Starting from point O, a solid red arrow is drawn along the observation direction. This observation direction is determined by the fixed installation orientation of the linear laser receiver after three rotational transformations: heading angle, roll angle, and pitch angle. Numerical step sampling is performed along the observation direction. Sampling points P0, P1, P2, P3, P4, P5, and P6 are marked in the figure, with a step size Δs of 5 mm between adjacent sampling points. Sampling points are marked with small blue circles and proceed sequentially from the photosensitive center outwards. The core process of the geometric solution algorithm is as follows: At each sampling point, the east and north coordinates of that point are extracted, and the reference plane elevation value at that location is obtained through the query interface of the reference plane model. Simultaneously, the vertical coordinates of the sampling point itself are calculated. The vertical coordinates of the sampling point are compared with the reference plane elevation value. When the vertical coordinates of the sampling point are first greater than or equal to the reference plane elevation value, it indicates that the beam has entered the region above the plane from below or within the plane. At this time, the enclosing interval between this sampling point and the previous sampling point is recorded. The bisection refinement interval is marked with a green dashed rectangle in the figure, located between sampling points P5 and P6. This enclosed interval is bisected by halving its length repeatedly until the interval length is no greater than 1 mm. The final determined intersection point is marked as point I, represented by a solid red circle; this point is the precise intersection of the beam and the reference plane. A red vertical dashed line is drawn downwards from intersection point I to the east-facing coordinate axis; this vertical coordinate value is the elevation reading of the screed blade edge.

[0041] In the continuous state tracking phase, the industrial controller constructs a particle ensemble containing multiple modes, each corresponding to a solution state of the airborne GNSS-RTK module. The number of particles in the ensemble is fixed at four, representing the line-of-sight mode, weak multipath mode, strong multipath mode, and occlusion mode, respectively. Each particle has a built-in Kalman filter for tracking the continuous state, including the 3D position and attitude of the screed. The continuous state includes east coordinates, north coordinates, elevation values, heading angle, roll angle, pitch angle, estimated elevation of the left screed edge, estimated elevation of the right screed edge, and average elevation estimate. The Kalman filter performs two steps in each control cycle: prediction and update. Prediction uses the vehicle speed, heading angle change, roll angle change, pitch angle change, and hydraulic cylinder stroke change from the attitude and displacement measurement unit to estimate the magnitude of the 3D position and attitude changes within the current cycle and advances the continuous state from the previous cycle to the current cycle. The update phase employs different observation routing methods depending on the mode. In line-of-sight mode, the system receives two types of observations: planar observations from the airborne GNSS-RTK module and elevation readings from the left and right cutting edges. First, the planar observations are used to correct the east and north coordinates. Then, the elevation readings from the left and right cutting edges are used to correct the estimated elevations of the left and right screed cutting edges, and the average elevation estimate is updated accordingly. In weak multipath mode, both types of observations are received simultaneously, but prediction relies more heavily on short-term motion changes provided by the attitude and displacement measurement units to offset the impact of weak reflections on the planar observations. In strong multipath mode, only the elevation readings from the left and right cutting edges are received; planar observations are not received to avoid horizontal position jumps caused by strong reflections. Simultaneously, the system uses changes in hydraulic cylinder stroke and vehicle speed to amplify the extrapolation range during prediction, ensuring the continuity of three-dimensional position and attitude. In occlusion mode, only the elevation readings from the left and right cutting edges are received. During prediction, changes in hydraulic cylinder stroke and heading angle are used for attitude extrapolation to maintain the availability of the continuous state under short-term occlusion. The reason for adopting the above routing arrangement is that the quality of available information varies significantly in each mode. Explicit routing allows the Kalman filter to process only reliable observations in each mode, thereby avoiding unnecessary corrections in cases of strong reflections or occlusion.

[0042] To ensure that the output of 3D position and attitude can directly drive subsequent control, the industrial controller reads the continuous state from the particle corresponding to the current mode in each control cycle, forming a 3D positioning stream. The 3D positioning stream includes east coordinates, north coordinates, average elevation estimate, left screed edge elevation estimate, right screed edge elevation estimate, yaw angle, roll angle, and pitch angle, along with a source label for the current mode. The presence of the source label allows subsequent control layers to adopt different channel orchestration strategies. For example, when the source label is a strong multipath mode or an occlusion mode, it relies more on the left and right edge elevation estimates and hydraulic cylinder stroke changes to maintain closed-loop stability. When the source label is a direct-view mode, it makes full use of the high horizontal position accuracy provided by planar observation to improve thickness control and smoothness along the line.

[0043] In another implementation, the phase step count identification of the linear laser receiver can employ a maximum slope method based on a time window. This method calculates the slope sequence of the phase curve within each fixed-length time window, selects the time point corresponding to the slope peak as the step edge, and counts to obtain the phase step count value. The advantage of using the maximum slope method is that it can maintain stable edge positioning even with slow fluctuations in light intensity, making it particularly suitable for scenarios with long frames and low step speeds. In this approach, edge identification is less dependent on absolute amplitude and less affected by slow occlusion caused by dust.

[0044] In another implementation, the bisection refinement of the geometric solution can be replaced by a three-segment interpolation refinement. Specifically, after finding the enclosing interval with a coarse step, three sets of reference plane elevation values ​​and vertical coordinates of the sampling points are collected at the two ends and the midpoint of the interval. The difference between the three points is used for single-peak positioning to directly obtain the approximate position of the intersection point. Then, a short interval bisection is used again, with a step size preferably of 0.5 mm, to complete the final value determination. The advantage of the three-segment interpolation refinement is that fewer refinement steps are required, making it suitable for devices with short control cycles, such as high-speed controllers with a 5-millisecond control cycle.

[0045] In another implementation, the airborne GNSS-RTK module employs a dual-antenna layout to directly output the heading angle. With a dual-antenna layout, the heading angle of the attitude and displacement measurement unit can be used as an alternative. When there are short-term abrupt changes in the heading angle of the dual antennas, the industrial controller prioritizes the heading angle of the attitude and displacement measurement unit to maintain a smooth change in the observation direction. The reason for using a complementary dual-antenna and single-antenna configuration is that turns, ramps, and the passage of metal vehicles can cause short-term multipath propagation. A dual-antenna configuration can provide a more stable heading angle under most operating conditions, while the attitude and displacement measurement unit maintains better continuity under strong vibrations and short-term obstructions.

[0046] In another implementation, the particle set can be expanded to six, adding a low-speed operation mode and a shutdown mode. The low-speed operation mode is activated when the vehicle speed is below 0.2 meters per second, receiving only the left and right cutting edge elevation readings and reducing the extrapolation of predicted displacement to avoid positional jitter caused by micro-motions. The shutdown mode is activated when the vehicle speed is 0, not performing predictions, but only updating the left and right screed cutting edge elevation estimates using the left and right cutting edge elevation readings, so as to stably record the surface topography when the paver pauses feeding.

[0047] In another implementation, the observation direction of the linear laser receiver can be determined by a combination of fixed mounting orientation and online fine-tuning. Online fine-tuning employs a step-by-step execution method, correcting the observation direction by a very small angle in each control cycle, preferably between 0.05 and 0.2 degrees. The correction is retained when the reference plane signal energy increases in that direction; otherwise, it reverts to the previous direction. This fine-tuning can compensate for minor deformations and temperature drift of the support structure after long-term operation, ensuring a stable geometric relationship between the beam and the reference plane during intersection search.

[0048] Through the above implementation, the linear array laser receiver stably generates a reference plane number and phase step count value in each control cycle. After loading the reference plane model, it is converted into elevation readings of the left and right ironing plate edges through deterministic processes such as numerical stepping and bisection refinement. The airborne GNSS-RTK module and attitude and displacement measurement unit provide real-time constraints on the spatial position and observation direction of the linear array laser receiver, while providing motion quantization for continuous state prediction. The particle ensemble uses the solution state of the airborne GNSS-RTK module as an index to provide a clear observation route and Kalman filter update sequence for each working condition, ensuring that the three-dimensional positioning flow remains continuous, predictable, and directly usable for subsequent target generation and closed-loop control. The entire processing chain operates stably on a control cycle of 10 to 20 milliseconds. Field engineers can complete the deployment through configuration tables, installation parameters, and simple open-loop checks. Under actual working conditions with high dust, high reflection, and frequent start-stop cycles, it continuously obtains millimeter-level elevation readings of the left and right ironing plate edges and centimeter-level three-dimensional position and attitude estimates.

[0049] In one implementation, the industrial controller receives the calculated status tag and multipath indication tag from the airborne GNSS-RTK in each control cycle, and switches between line-of-sight mode, weak multipath mode, strong multipath mode, and occlusion mode according to a fixed mode switching table. The mode switching table uses a one-to-one mapping relationship without introducing scoring or proportional coefficients: when the calculated status tag is fixed, it enters line-of-sight mode; when the calculated status tag is floating and the multipath indication tag is weak, it enters weak multipath mode; when the calculated status tag is floating and the multipath indication tag is strong, it enters strong multipath mode; when the calculated status tag is unavailable, it enters occlusion mode. The reason for using this mapping relationship is that the calculated status tag and multipath indication tag are directly generated by the airborne GNSS-RTK within the observation domain, which can reflect the stability of horizontal positioning information with the shortest path; establishing mode switching based on the discrete values ​​of the tags can make subsequent observation routes completely determined, avoiding control lag caused by cross-dependencies under complex operating conditions.

[0050] like Figure 3As shown, the experimental curves of horizontal positioning accuracy under different GNSS-RTK modes demonstrate the positioning performance of the multi-mode particle ensemble of this invention under different working conditions. A time-positioning error coordinate system is established in the figure. The horizontal axis represents time in seconds, ranging from 0 to 80 seconds; the vertical axis represents horizontal positioning error in centimeters, ranging from 0 to 14 centimeters. The experimental curves are divided into five stages according to time sequence, each stage corresponding to a different GNSS-RTK solution state and a corresponding particle mode. The first stage, from 0 to 20 seconds, is in line-of-sight mode. The positioning error curve in this stage is drawn with a thick black solid line, with the error value fluctuating slightly around 1.5 centimeters, with a fluctuation range of approximately ±0.3 centimeters. In line-of-sight mode, the GNSS-RTK module's solution state label is "fixed," indicating the best planar observation quality and the highest horizontal positioning accuracy. The curve shows that the system can stably maintain a positioning error within 2 centimeters. The second stage, from 20 to 40 seconds, involves the system switching to weak multipath mode. At the 20-second mark, a gray vertical dashed line marks the mode switching point, with the label "Switched to Weak Multipath" above it. The positioning error curve for this stage is drawn with a black dashed line, alternating between longer segments and shorter intervals. The error value gradually increases from 1.5 cm to around 5 cm, with an average error of approximately 3.2 cm and a fluctuation range increasing to ±0.8 cm. The weak multipath mode corresponds to a GNSS-RTK solution status of "Floating" and a multipath indicator label of "Weak." At this time, the system still receives both planar and elevation observation data, but during the prediction phase, it relies more heavily on short-term motion changes of attitude and displacement measurement units to offset the impact of weak reflections on planar observations. In the third stage, from 40 to 60 seconds, the system further switches to strong multipath mode. The 40-second mark is labeled "Switched to Strong Multipath." The positioning error curve for this stage is drawn with a black dashed line, alternating between shorter segments and shorter intervals, visually presenting a denser dashed line effect. The error value increased rapidly from 5 cm, reaching approximately 12 cm at 60 seconds. The average error was 7.5 cm, with a standard deviation of ±2.1 cm. In the strong multipath mode, the GNSS-RTK solution status is "floating" and the multipath indicator label is "strong." In this mode, the system only receives elevation observations and not planar observations, using displacement extrapolation based on hydraulic cylinder stroke changes and vehicle speed to maintain the continuity of the horizontal coordinates. In the fourth stage, from 60 to 70 seconds, the system switches to occlusion mode. The 60-second mark indicates "Switched to Occlusion Mode." The positioning error curve for this stage is drawn as a black dotted dashed line, with the dashed line pattern consisting of alternating very short segments and intervals. The error value continued to rise, increasing from 12 cm to approximately 15 cm. The average error was 12.3 cm, with a standard deviation of ±3.5 cm. The occlusion mode corresponds to the GNSS-RTK solution status being "unavailable". The system relies solely on elevation observations, changes in heading angle, and changes in hydraulic cylinder stroke for pose extrapolation. Horizontal positioning accuracy decreases significantly, but continuous output can still be maintained.The fifth stage, from 70 to 80 seconds, saw the system return to line-of-sight mode. The 70-second mark is labeled "Return to Line-of-Sight Mode." The positioning error curve for this stage is again plotted as a thick black line. The error value rapidly decreased from 15 cm, and after a transition period of approximately 10 seconds, converged to within 2 cm at 80 seconds, eventually stabilizing around 1.5 cm. This recovery process demonstrates the system's rapid convergence capability after GNSS-RTK signal recovery. Figure 3 A legend box is drawn at the bottom, marked with a white rectangular background and a black border. The legend lists the curve styles and corresponding accuracy ranges for each mode in four rows: Direct view mode uses a thick solid line with an error of less than 2 cm; weak multipath mode uses a long dashed line with an error of 2 to 5 cm; strong multipath mode uses a short dashed line with an error of 5 to 12 cm; and occlusion mode uses a dotted dashed line with an error of more than 10 cm. Figure 3 The accuracy statistics box on the right lists the statistical data for each mode. The positioning accuracy is 1.5 ± 0.3 cm in line-of-sight mode; 3.2 ± 0.8 cm in weak multipath mode; 7.5 ± 2.1 cm in strong multipath mode; and 12.3 ± 3.5 cm in occlusion mode. The above experimental curves fully verify the effectiveness of the multi-mode particle ensemble of this invention. The system can automatically switch modes according to the GNSS-RTK solution status, maintain the continuity of positioning output under different operating conditions, and quickly converge to a high-accuracy state after signal recovery.

[0051] The particle swarm is created during system initialization and maintains a constant size. The swarm contains four particles, corresponding to the line-of-sight mode, weak multipath mode, strong multipath mode, and occlusion mode, respectively. Each particle has a built-in Kalman filter that tracks a continuous state including the screed's 3D position and attitude. This continuous state consists of east coordinates, north coordinates, elevation values, heading angle, roll angle, pitch angle, estimated elevation of the left screed edge, estimated elevation of the right screed edge, and average elevation estimate. Each particle's Kalman filter uses the same set of numerical dimensions and the same computational order to ensure consistent output structure when switching between different modes, avoiding interface incompatibility. The fixed swarm size directly benefits computational load prediction: within a 20-millisecond control cycle, the prediction and update of the four Kalman filters can be completed within 2 to 4 milliseconds, leaving time for geometric solving and electro-hydraulic servo control.

[0052] Before entering the particle swarm, observations are organized into two types of data packets. Planar observations include two fields: east and north coordinates, along with an accompanying solution status label. Elevation observations include two fields: the elevation readings of the left and right screed blade edges, along with the quantized values ​​for the average elevation and cross slope angle. These data packets are generated once per control cycle. Time alignment uses 1PPS pulses and 64-bit cycle counting from airborne GNSS-RTK, and a consistent time base within a single cycle is achieved through timestamp alignment within the industrial controller. The reason for classifying observations into planar and elevation observations is the significant differences in source and signal-to-noise characteristics at the construction site. Horizontal information is mainly affected by metal reflections and building obstructions, while elevation information is mainly affected by laser echo quality and local obstructions. The classified routing maintains clear input boundaries in each mode, ensuring that the Kalman filter only updates the state components matching the current operating condition, thus achieving a balance between millimeter-level elevation control and centimeter-level horizontal positioning.

[0053] In the current mode, the mode register records the data in each control cycle, and the observation route is executed exactly according to the mode register's instructions. In the direct-view mode and weak multipath mode, both planar and elevation observations are simultaneously fed into the Kalman filter of the corresponding particle. First, updates related to the horizontal position are performed, followed by updates related to the elevation. The advantage of updating the horizontal position first is that geometric solutions and control target generation require interpolation on a unified horizontal coordinate system; convergence of the horizontal position first reduces interpolation errors in the elevation target. In the strong multipath mode, only elevation observations are fed into the Kalman filter of the corresponding particle. Simultaneously, hydraulic cylinder stroke variation and vehicle speed are added to the prediction input, allowing the continuous state to be maintained through short-term displacement extrapolation when horizontal information is unavailable. The reason for adding hydraulic cylinder stroke variation is that the short-term change in the screed elevation is mainly determined by the actuator's action; this amount directly determines the extrapolation magnitude of the left and right screed edge elevation estimates. Adding vehicle speed allows the east and north coordinates to advance in controllable steps, ensuring that trajectory and design surface queries remain reliable. In occlusion mode, only elevation observations are fed into the Kalman filter of the corresponding particle, while changes in hydraulic cylinder stroke and heading angle are added to the prediction input to maintain pose continuity. The heading angle change can often still be provided by the attitude and displacement measurement unit during occlusion, and the controller updates the vehicle's orientation accordingly, thus maintaining the correct path direction during horizontal coordinate extrapolation. This is particularly crucial for generating the thickness of targets along the line.

[0054] Each Kalman filter performs condition prediction and condition update in a fixed sequence within a control cycle. Condition prediction takes vehicle speed, changes in heading angle, roll angle, pitch angle, and hydraulic cylinder stroke as inputs, and advances prior estimates of eastward coordinates, northward coordinates, elevation values, and attitude angles. The advancement sequence follows a structure of position first, then attitude, and finally edge elevation. Prioritizing position reduces coupling errors in the initial extrapolation phase of attitude. Placing edge elevation at the end of the extrapolation ensures the full impact of the actuator's actions within the current cycle is incorporated into the extrapolation. Condition update strictly follows the observation sequence after routing: in line-of-sight mode and weak multipath mode, eastward and northward coordinates are updated first based on planar observations, then the left and right screed edge elevation estimates are updated based on elevation observations, and the average elevation estimate is recalculated accordingly. In strong multipath mode and occlusion mode, the left and right screed edge elevation estimates are updated directly based on elevation observations, while retaining the position extrapolation result as the horizontal output for the current cycle. The essence of this update order is to prioritize observations with higher reliability for relevant state components, reducing mutual interference. For example, in a strong multipath mode, if horizontal observations are still involved in the update, the jumps caused by strong reflections will be introduced into the eastward and northward coordinates, resulting in unnecessary thickness perturbations during the target generation stage. By limiting the update order with observation routes, a stable effect can be achieved with a simple and clear process.

[0055] The industrial controller directly reads the fields of the continuous state from the particles corresponding to the current mode, combining them into a data frame containing east coordinates, north coordinates, average elevation estimate, left screed edge elevation estimate, right screed edge elevation estimate, heading angle, roll angle, and pitch angle, along with a source tag. The source tag records one of four modes: direct-view mode, weak multipath mode, strong multipath mode, or occlusion mode. The single source tag design allows the subsequent control layer to select the appropriate channel orchestration strategy immediately. For example, when the source tag is a strong multipath mode or occlusion mode, the left and right screed edge elevation estimates are used first to generate the control target; when the source tag is a direct-view mode, the high accuracy of the east and north coordinates is fully utilized for line interpolation to improve the spatial consistency of thickness control. The 3D positioning stream uses a fixed field order and fixed byte length to simplify the data interface with the electro-hydraulic servo side, with a typical frame length between 128 and 192 bytes. The control cycle is preferably 10 milliseconds or 20 milliseconds. In the 10-millisecond configuration, 3 milliseconds are allocated to geometry solving, 2 milliseconds to the Kalman filter, and 1 millisecond to data preparation on the electro-hydraulic servo side. The remaining time is used for data alignment and frame encapsulation. In the 20-millisecond configuration, the time budget for each stage is proportionally increased to allow for higher-resolution line search refinement or higher-bandwidth current output. In actual road construction, the 10-millisecond configuration can maintain stability at vehicle speeds from 0.5 m / s to 1.2 m / s; the 20-millisecond configuration is more suitable for constant speed ranges above 1.2 m / s, as the longer period combined with a longer phase sequence helps improve the smoothness of elevation readings.

[0056] In one optional implementation, the particle set is expanded to six, and a low-speed operation mode and a shutdown mode are added. The low-speed operation mode is activated when the vehicle speed is below 0.2 meters per second. The route retains only elevation observations, and the displacement extrapolation amplitude is reduced during the prediction phase. This allows the elevation estimates of the left and right screed blade edges to change slightly with the actuator's movements, preventing the amplification of micro-vibrations on the horizontal coordinate. The shutdown mode is activated when the vehicle speed is 0. During the prediction phase, the position and attitude are not adjusted; only the elevation estimates of the left and right screed blade edges are updated based on elevation observations. This ensures that the system can still output a stable average elevation estimate during periods of paused feeding or short waiting times, facilitating subsequent connections.

[0057] In another alternative implementation, a dual-antenna airborne GNSS-RTK is used to directly acquire the heading angle. The dual antennas provide a relatively stable heading angle output even when the vehicle is turning, with the heading angle provided by the attitude and displacement measurement unit serving as a backup. When the dual antennas briefly unlock and then re-engage, the heading angle may experience a small abrupt change within one cycle. During the prediction phase, the industrial controller prioritizes using the heading angle change from the attitude and displacement measurement unit to maintain the continuity of the observation direction. Then, during the update phase, it switches back to the heading angle provided by the dual antennas, allowing extrapolation and update to each utilize a more stable source.

[0058] In one implementation, the industrial controller receives query results from the 3D positioning flow and the digital design surface model in each control cycle. Using the east coordinates, north coordinates, heading angle, estimated elevation of the left and right ironing plate edges, average elevation estimate, roll angle, and pitch angle within the 3D positioning flow as inputs, it generates a control target and generates current commands for the left and right valves via adaptive feedforward and a PID controller, driving the electro-hydraulic servo mechanism to complete closed-loop control. The control cycle is preferably 10 milliseconds or 20 milliseconds; the following explanation uses 10 milliseconds as an example. First, in the target generation stage, the industrial controller determines the query positions for the left and right sides based on the east coordinates, north coordinates, and heading angle of the 3D positioning flow. The nominal width of the ironing plate is given in the equipment parameter table, typically 3.0 meters or 3.5 meters. The industrial controller takes half of the nominal width as the left and right offset along the lateral direction perpendicular to the heading angle, shifting from the horizontal position in the 3D positioning flow to the query positions on the left and right sides. The direct benefit of this arrangement is that the query point of the digital design surface model is consistent with the physical projection of the cutting edges of the left and right ironing plates, and the subsequent error calculation directly corresponds to the execution end without the need for additional geometric conversion.

[0059] The industrial controller initiates queries to the digital design surface model at the left and right query positions respectively, obtaining the left and right design elevations. The digital design surface model can use either a regular grid or an irregular triangular mesh, and the query method employs bilinear interpolation or mother triangle interpolation. Taking a regular grid as an example, the grid resolution is preferably 0.1 to 0.5 meters. When the grid resolution is 0.2 meters, each query position involves interpolation of 4 grid points in the horizontal direction, with a calculation time in the sub-millisecond range. The industrial controller then obtains the target set, which includes the left design elevation, the right design elevation, the average design elevation, and the target cross slope angle. The target cross slope angle is obtained in two ways: the first is to directly query the elevation difference between the left and right query positions in the digital design surface model and convert it to the nominal width to obtain the target cross slope angle; the second is to read the cross slope setting value of the construction section from the digital design surface model as the target cross slope angle. The choice between these two methods is explained as follows: When the road cross-section changes slowly with mileage, the first method can capture local gradual changes; when the construction section adopts a constant cross slope, the second method avoids repeated calculations and has a simpler interface.

[0060] The error calculation stage then begins. The industrial controller subtracts the estimated elevations of the left and right ironing plate edges within the 3D positioning flow from the left and right design elevations, respectively, to obtain the left and right channel elevation errors. Millimeters are used to match the resolution of the electro-hydraulic servo mechanism. The two channel elevation errors are added together and divided by 2 to obtain the average elevation channel error. The difference between the right and left channel elevation errors is converted to the nominal width to obtain the cross slope channel error. The reason for using sums and differences is that the electro-hydraulic servo mechanism has left-right coupling characteristics in its mechanical structure. Through channel decomposition, the average lift and lateral tilt can be distributed to two command channels. This ensures a clear correspondence when subsequently programmed back into the left and right valve currents, facilitating the separate application of amplitude limiting and slope constraints.

[0061] During the adaptive feedforward generation phase, the industrial controller reads the vehicle speed and the slope information of the digital design surface model in the direction of travel. The slope in the direction of travel is obtained by taking one sampling point in front of and one in front of the current position and performing a differential sampling along the path. The sampling interval is preferably 0.5 meters or 1.0 meter, and the industrial controller selects one of these two sets from the parameter table. For example, when the vehicle speed is 0.8 meters per second and the sampling interval is 1.0 meter, the industrial controller queries the average design elevation at the current position and 1.0 meter ahead, and the difference between the two is divided by the sampling interval to obtain the slope along the path. The slope along the path reflects the changing trend of the design elevation in front, and combined with the vehicle speed, it can predict the expected rise and fall of the screed within one or more control cycles. If the slope along the path is positive and the vehicle speed is high, it means that an uphill section is about to be entered. Increasing the current command in advance helps to compensate for the dynamic lag of the hydraulic system and the screed, thereby reducing the subsequent elevation error. If the slope along the path is negative, decreasing the current command in advance allows the screed to descend smoothly, avoiding the ripples caused by the peak and subsequent fall. To convert the friction gradient and vehicle speed into feedforward current, the industrial controller calls a feedforward function. At least three sets of curves are pre-set at the factory according to the equipment model and material type. For example, curve one is for room-temperature fine-grained materials, curve two for medium-temperature medium-grained materials, and curve three for high-temperature coarse-grained materials. The input to each curve is the friction gradient and vehicle speed, and the output is the feedforward current for the average elevation channel and the cross slope channel. The industrial controller selects the curve set according to the vehicle speed range. For example, curve one is used when the vehicle speed is below 0.5 m / s, curve two is used when the vehicle speed is between 0.5 m / s and 1.0 m / s, and curve three is used when the vehicle speed is above 1.0 m / s. Using a curve lookup table, the complex viscoelastic and hydraulic responses of the materials can be mapped to a stable output. Field personnel only need to select the curve according to the operating conditions without adjusting the calculation process.

[0062] During the PID control phase, the industrial controller is configured with a PID controller for both the average elevation channel and the cross slope channel. Each PID controller reads the error of the corresponding channel in each control cycle and forms the feedback current in three parts: proportional, integral, and derivative. The proportional part directly multiplies the current error by the proportional coefficient to generate the immediate correction, typically ranging from 0.6 mA / mm to 1.2 mA / mm, with a factory default of 0.8 mA / mm. The integral part accumulates the error into the integral register over the control cycle. The upper and lower bounds of the integral register are given in the parameter table, typically ±60 mA, to prevent long-term deviations from causing output bias. The derivative part calculates the change in error between the current cycle and the previous cycle, divides it by the control cycle to obtain the rate of change, and multiplies this rate of change by the derivative coefficient to generate the trend correction, typically ranging from 0.01 mA / mm / ms to 0.03 mA / mm / ms. The sum of these three parts constitutes the feedback current for that channel. The reason for using dual-channel PID is that the effects of average lifting and lateral tilting are different on the hydraulic mechanism and the ironing plate structure. Independent tuning can allow the two types of errors to converge at an appropriate speed, avoiding oscillations caused by mutual interference.

[0063] The industrial controller superimposes the adaptive feedforward current and the PID feedback current to obtain the total current of the average elevation channel and the total current of the cross slope channel. Limiting and slope constraints are then applied. Limiting restricts the total current to the range allowed by the actuator, typically from 0 mA to 20 mA. Slope constraints limit the current change between two adjacent control cycles to a maximum step size, typically no more than 2 mA per cycle, to suppress abrupt acceleration at the actuator. Limiting and slope constraints are applied separately at the channel level, allowing the dynamic responses of average elevation and lateral tilt to be set according to individual requirements. For example, in road sections with significant thickness variations, increasing the upper limit of the slope of the average elevation channel allows for faster design adaptation; in ramp sections with significant cross slope variations, increasing the upper limit of the slope of the cross slope channel makes lateral adjustments more sensitive.

[0064] See Figure 4The step response characteristic curve of the elevation PID controller demonstrates the dynamic performance of the closed-loop control system of this invention. A time-elevation deviation coordinate system is established in the figure. The horizontal axis represents time in seconds, ranging from 0 to 10 seconds; the vertical axis represents elevation deviation in millimeters, ranging from 0 to 30 millimeters. The step input signal is plotted as a gray dashed line. This signal remains at the 0 mm baseline from 0 to 2 seconds, undergoes a step change at 2 seconds, instantaneously rising to 30 mm and remaining constant. This step input simulates abrupt changes in the design elevation during actual construction, such as the transition zone from a flat slope to an uphill slope. The actual response curve of the system is plotted as a thick black solid line, representing the control scheme using PID with adaptive feedforward. The curve remains at the 0 mm baseline from 0 to 2 seconds. After responding to the step input, the elevation deviation rises from 0 mm. During the rise phase from 2 to 4 seconds, the curve exhibits a smooth upward trend with a gradually accelerating rate of increase. At approximately 3.4 seconds, the curve passes the 15mm mark, reaching 50% of the target value. This time point is defined as the 50% rise time and is labeled as... At approximately 4.6 seconds, the curve first approaches the target value of 30 mm, but due to system inertia, it continues to rise to approximately 31.5 mm, resulting in an overshoot of 1.5 mm. This overshoot is indicated by a blue vertical double arrow, signifying an overshoot of 1.5 mm. After the overshoot peak, the curve begins to decline and gradually approaches the target value. Between 5 and 7 seconds, the curve oscillates slightly around the target value, with the amplitude of the oscillation gradually decreasing. At approximately 7 seconds, the curve enters the steady-state error range, defined as an error within ±5% of the target value, i.e., between 29 mm and 31 mm. This time point is defined as the settling time, marked as ts = 5.0 seconds. Between 7 and 10 seconds, the curve remains stably near the 30 mm target value, indicating that the system has reached steady state. To compare control performance, the response curve of pure PID control is also plotted in the figure, represented by a black, medium-thickness dashed line, with the dashed line pattern consisting of alternating medium-length segments and intervals. The rise time of the pure PID curve is relatively slow, with a 50% rise time of approximately 3.2 seconds and a settling time of approximately 6.5 seconds, significantly slower than the PID plus feedforward scheme. The comparison of the two curves demonstrates that the introduction of adaptive feedforward can significantly improve the dynamic response speed of the system and shorten the settling time. Figure 4 Two key performance indicators are labeled in the lower middle section. The first label is located at 3.4 seconds, drawn with a red vertical dashed line from the 50% mark on the curve to the horizontal axis, labeled "50% rise time" and "...". The second mark is at 7 seconds, also marked with a red vertical dashed line, with "stable time" and "ts=5.0s" below it. Figure 4A legend box is drawn at the bottom, listing the meaning of the three curves. The first row is a thick black solid line, labeled "PID + Feedforward Control Response"; the second row is a black dashed line, labeled "Pure PID Control Response"; and the third row is a gray dashed line, labeled "Target Value". Figure 4 The control performance parameter box on the right shows the detailed parameter settings and performance indicators of the PID controller. The parameters include: proportional gain Kp of 0.8 mA / mm, integral gain Ki of 0.05 mA / mm / s, and derivative gain Kd of 0.02 mA / s / mm. The performance indicators include: rise time tr of 2.4 seconds and settling time ts of 5.0 seconds. These parameters and indicators fully demonstrate the excellent performance of the control system of this invention under actual working conditions, achieving rapid response while maintaining stability, thus meeting the high-precision and fast-response requirements of road construction for elevation control.

[0065] After constraints are met, the total channel current is converted into current commands for the left and right valves via a sum-difference matrix. The conversion rule is: the left valve current command equals the average elevation channel current minus the cross-slope channel current; the right valve current command equals the average elevation channel current plus the cross-slope channel current. The intuitive meaning of this matrix is ​​that average lifting is achieved by unidirectional movement on both sides, while lateral tilting is achieved by opposite movements on both sides. For example, when the average elevation channel current is 8 mA and the cross-slope channel current is 2 mA, the left valve current command is 6 mA, and the right valve current command is 10 mA. The ironing plate lifts overall while tilting slightly to the right. After generating the command, the industrial controller adds a small jitter current in each control cycle to reduce valve core static friction. Typical jitter frequencies are 40 Hz to 80 Hz, with amplitudes ranging from 2% to 5% of the rated range, for example, 0.4 mA to 1.0 mA. Jitter keeps the valve core in a small motion state, improving sensitivity in the low-speed range and thus enhancing the resolvability of millimeter-level elevation control.

[0066] The actuator drives the electro-hydraulic servo mechanism according to the current commands from the left and right valves. The hydraulic cylinder generates displacement, causing the ironing plate to rise, fall, and tilt. The readings from the displacement and pressure sensors are sent back to the industrial controller in the same control cycle, becoming the predictive input for the next cycle. Through this closed loop, the 3D positioning flow continuously interacts with the digital design surface model at a 10-millisecond rhythm. The error decreases cycle by cycle under the action of the PID controller, adaptive feedforward provides advance capability when the vehicle speed and slope along the route change, and current limiting and slope constraints maintain the smoothness and controllability of the action.

[0067] In one alternative implementation, the sampling interval for the forward query is adaptively selected. The industrial controller divides the vehicle speed into three intervals: a low-speed interval of 0.2 m / s to 0.5 m / s, a medium-speed interval of 0.5 m / s to 1.0 m / s, and a high-speed interval of 1.0 m / s to 1.5 m / s. Correspondingly, 0.5 m, 1.0 m, and 1.5 m are selected as sampling intervals. The benefit of this arrangement is that it provides a longer look-ahead distance at high speeds, giving more lead time to the adaptive feedforward and reducing overshoot caused by execution lag; and a shorter look-ahead distance at low speeds, avoiding premature introduction of distant slope features and maintaining local smoothness.

[0068] In another optional implementation, the proportional, integral, and derivative coefficients of the PID controller are range-based, adjusted according to oil temperature and pressure levels. The oil temperature levels are divided into low, medium, and high temperatures, while the pressure levels are divided into light load, medium load, and heavy load. The industrial controller reads the current level in each control cycle and selects a preset combination of coefficients. Taking the average elevation channel as an example, a higher proportional coefficient and a lower derivative coefficient are used under low temperature and heavy load conditions to give the system stronger thrust to overcome viscosity-induced hysteresis; while a lower proportional coefficient and a higher derivative coefficient are used under high temperature and light load conditions to ensure smooth operation even in low-viscosity environments. The advantage of range-based adjustment is that parameter selection is interpretable; on-site selection only requires information on oil temperature and load, avoiding untrackable response changes under complex operating conditions.

[0069] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A method for three-dimensional positioning and elevation control of a screed based on GNSS-RTK and domain laser coupling, characterized in that, The method comprises the following steps executed sequentially: establishing an encoding reference; deploying at least two domain laser transmitters in the construction domain, each transmitter configured with a multi-plane reference; assigning a unique phase sequence number and a unique frequency subcarrier number to each reference plane; using a non-overlapping set of subcarrier numbers for multiple transmitters; installing linear array laser receivers on both sides of the ironing plate to receive and decode the reference plane number and phase step count value; combining this with a pre-set reference plane model to generate the elevation readings of the left and right cutting edges of the ironing plate through geometric solution; simultaneously acquiring the aircraft's pose and motion information through an airborne GNSS-RTK module and attitude and displacement measurement unit; and constructing a particle model containing multiple modes. The system consists of a set of GNSS-RTK solutions, each corresponding to a specific mode. Each particle has a built-in Kalman filter to track the continuous state containing the three-dimensional position and attitude of the screed. The system switches between different modes based on the GNSS-RTK solution state label and routes the GNSS-RTK planar observations and the elevation readings of the left and right cutting edges of the screed to the corresponding particle's Kalman filter according to the current mode. Conditional prediction and updating are performed to generate a three-dimensional positioning flow. Based on the three-dimensional positioning flow and the digital design surface model, a control objective is generated, and elevation and cross slope errors are calculated. A total current command is generated through adaptive feedforward and a PID controller to drive the electro-hydraulic servo mechanism, completing closed-loop control. The constructed modes include four types: line-of-sight mode, weak multipath mode, strong multipath mode, and occlusion mode. The mode switching rules based on the GNSS-RTK's solution status label and multipath indicator label are as follows: a) If the solution status label is fixed, enter line-of-sight mode; b) If the solution status label is floating and the multipath indicator label is weak, enter weak multipath mode; c) If the solution status label is floating and the multipath indicator label is strong, enter strong multipath mode; d) If the solution status label is unavailable, enter occlusion mode. The routing is based on the rules for GNSS-RTK plane observations and the elevation readings of the left and right edges of the screed. The following parameters are specified: a) In direct-view mode, both the planar observation and the elevation readings of the left and right cutting edges of the screed are simultaneously fed into the Kalman filter of the corresponding particle; b) In weak multipath mode, both the planar observation and the elevation readings of the left and right cutting edges of the screed are simultaneously fed into the Kalman filter of the corresponding particle; c) In strong multipath mode, only the elevation readings of the left and right cutting edges of the screed are fed into the Kalman filter of the corresponding particle, and the hydraulic cylinder stroke change and vehicle speed are added for displacement extrapolation; d) In occlusion mode, only the elevation readings of the left and right cutting edges of the screed are fed into the Kalman filter of the corresponding particle, and the hydraulic cylinder stroke change and heading angle change are added for attitude extrapolation.

2. The method according to claim 1, characterized in that, In the step of establishing the coding reference, the phase sequence allocated to each reference plane is composed of uniform steps and has a length of not less than 8; the interval between frequency subcarriers is fixed and the subcarrier table length is not less than 8; the domain laser transmitter rotates the reference plane with a fixed-length frame structure and outputs the data in the order of the phase sequence number.

3. The method according to claim 1, characterized in that, The steps for generating the elevation readings of the left and right cutting edges of the ironing plate through geometric solution include: a) establishing a beam direction table and beam starting point for each linear laser receiver, generated by a combination of installation parameters and real-time attitude; b) using a numerical stepping line to search for and locate the intersection point between the beam and the reference plane, stepping outward from the starting point along the beam direction with a step size of no more than 5 mm. When the vertical coordinate of the sampling point is greater than or equal to the elevation value of the reference plane for the first time, the enclosed interval formed by the current sampling point and the previous sampling point is recorded; c) performing bisection refinement on the enclosed interval, successively halving the step size until the step size is no more than 1 mm, taking the sampling point of the last refinement as the intersection point, and using the vertical coordinate of the intersection point as the elevation reading of the cutting edge of the ironing plate on the corresponding side.

4. The method according to claim 1, characterized in that, In the step of generating the total current command, the adaptive feedforward is implemented as follows: read the vehicle speed and the slope of the design surface in the current travel direction, call the feedforward function to obtain the feedforward current command; the feedforward function is pre-set with at least 3 sets of curves according to the equipment model and material properties at the factory, and selects one of them according to the vehicle speed range.

5. The method according to claim 1 or 4, characterized in that, The steps for calculating errors and generating total current commands also include: subtracting the estimated elevations of the left and right blade edges in the three-dimensional positioning flow from the control target to obtain the left channel error and the right channel error; calculating the sum and difference of the two channel errors to generate the average elevation channel error and the cross slope channel error, and establishing a sum-difference mapping matrix between the two channels; performing PID regulation on the average elevation channel error and the cross slope channel error respectively, and adding them with the feedforward current command, and converting them into current commands for the left and right valves through a sum-difference back-compilation matrix to achieve decoupled control of the average elevation and cross slope.

6. The method according to claim 1, characterized in that, Before execution, a spatiotemporal alignment step is also included: using the 1PPS signal of GNSS-RTK as the system time reference, a synchronization pulse is allocated to each device; each device records a 64-bit cycle count along with the data during sampling; after receiving all the data, the industrial controller aligns each data packet to the same control cycle according to the cycle count.

7. The method according to claim 1, characterized in that, The preset reference plane model is generated in the following way: at least three calibration points with known elevations are set on the construction site, and calibration prisms are used to point to the reference plane in sequence. The industrial controller reads the east and north directions and their elevation values ​​of the calibration points, and at the same time reads the phase step count value of the linear laser receiver on the reference plane. The built-in plane fitting process is called to generate a set of reference plane model coefficients, which are then bound and stored with the reference plane number.

Images