Automatic driving truck coordination formation control system based on feedforward mechanism

Abstract

The invention discloses an automatic driving truck coordinated formation control system based on a feedforward mechanism, which comprises a front truck system and a rear truck system, and the front truck system and the rear truck system realize data interaction through a truck-to-truck communication technology to cooperatively complete formation control. The front vehicle system comprises a vehicle-mounted sensor module, a median filtering module, a state prediction module, a rear vehicle control calculation module and a communication sending module. The rear vehicle system comprises a vehicle-mounted sensor module, a median filtering module, a self-state detection module, a communication receiving module, a fusion control module and an execution mechanism. MPC continuous rolling optimization (control parameters are updated every 0.1-0.3 second) can adapt to the state change of a front vehicle in real time, and the influence of communication delay and delay of an executing mechanism is doubly counteracted in cooperation with a feed-forward signal for transmitting the control requirement 100 ms in advance.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving control technology, specifically to an autonomous truck coordinated platooning control system based on a feedforward mechanism. Background Technology

[0002] With the increasing demands for efficiency and cost reduction in the logistics and transportation industry, autonomous truck platooning technology has become a hot research topic due to its ability to reduce wind resistance, decrease the number of drivers, and improve road utilization. Current mainstream autonomous truck platooning control schemes generally use a "two-vehicle following" logic as their core logic, employing a pure feedback control mode: the following vehicle collects real-time status information such as the speed, position, and acceleration of the preceding vehicle through sensors, calculates control deviations based on its own driving status, and then adjusts its own power, braking, or steering systems through feedback regulation (such as PID control) to maintain the platoon formation. However, this pure feedback control mode has many intractable defects in practical applications. Two key delays exist during platooning: vehicle-to-vehicle communication latency (traditional V2X communication latency is mostly between 150-300ms) and vehicle actuator hysteresis (braking and steering system response latency is approximately 100-200ms). Feedback control requires the following vehicle to detect "changes in the state of the preceding vehicle - its own deviation" before it can start adjusting. This double delay causes the following vehicle's control response to lag, which can easily lead to "too slow following" or "fluctuations in distance". In high-speed driving scenarios, it can even cause the risk of rear-end collisions. There is an urgent need for an autonomous driving truck coordination and platooning control system based on a feedforward mechanism to solve the above problems. Summary of the Invention

[0003] The purpose of this invention is to provide an autonomous truck coordinated platooning control system based on a feedforward mechanism to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an autonomous driving truck coordinated platooning control system based on a feedforward mechanism, comprising a leading vehicle system and a trailing vehicle system, wherein the leading vehicle system and the trailing vehicle system interact with each other through vehicle-to-vehicle communication technology to collaboratively complete platooning control; the leading vehicle system includes an onboard sensor module, a median filtering module, a state prediction module, a trailing vehicle control estimation module, and a communication transmission module; the trailing vehicle system includes an onboard sensor module, a median filtering module, a self-state detection module, a communication receiving module, a fusion control module, and an actuator; the leading vehicle system collects its own state data through sensors, performs denoising and prediction, and uses a model predictive control (MPC) algorithm to estimate the control parameters of the trailing vehicle and sends a feedforward control signal to the trailing vehicle; the trailing vehicle system collects its own state data and performs denoising, receives the feedforward control signal, fuses its own state data with the feedback information from the leading vehicle to generate execution commands, drives the actuators to move, and achieves dynamic and stable platooning control, specifically including the following steps: S1 Forward Vehicle Status Measurement and Preprocessing: The forward vehicle collects its own real-time status data through on-board sensors at a preset sampling frequency. The status data includes position, vehicle speed, acceleration, steering angle and braking signal. The signal spike noise is removed by the median filtering module. S2 Forward Vehicle Status Prediction: Based on the denoised status data, the forward vehicle's status change trend within a preset time period is predicted through a preset motion prediction model. S3 Rear Vehicle Control Parameter Optimization and Calculation: The leading vehicle continuously optimizes the target control parameters of the following vehicle by combining its own real-time status, predicted status and formation constraints, using the MPC algorithm. S4 control signal transmission: The leading vehicle transmits the optimized feedforward control signal to the trailing vehicle in real time through the communication transmission module; S5 Rear Vehicle Status Preprocessing and Signal Reception: The rear vehicle collects real-time status data through its own on-board sensors and performs median filtering to remove noise, while simultaneously receiving feedforward control signals sent by the preceding vehicle. S6 Fusion Control and Execution: The rear vehicle fusion control module performs weighted fusion of the feedforward control signal and the feedback signal, generates execution instructions and transmits them to the actuator to complete the formation maintenance.

[0005] Preferably, the on-board sensor modules of both the front and rear vehicle systems include GPS, IMU, and millimeter-wave radar, and the sampling frequency is set to 100Hz. The median filtering module adopts a 3-5 point median filtering algorithm. The sliding window of the 3-5 point median filtering algorithm includes the current sampling point and at least one historical sampling point. The removal rate of instantaneous peak noise such as speed jumps caused by bumps and position deviations caused by electromagnetic interference is not less than 95%, ensuring the smoothness of the status data.

[0006] Preferably, the state prediction module uses a Kalman filter algorithm combined with a vehicle dynamics model to construct a motion prediction model, wherein the vehicle dynamics model is a three-degree-of-freedom vehicle dynamics model; the preset time is 0.5-2 seconds, and the future state of the preceding vehicle calculated based on the motion prediction model includes changes in future vehicle speed, position and steering angle, with the prediction error controlled within 5%.

[0007] Preferably, the MPC algorithm of the rear vehicle control calculation module is set to a prediction time domain of 1-3 seconds and a control time domain of 0.5-1.5 seconds. The formation constraints include safe distance, speed synchronization accuracy, and formation maintenance deviation threshold. The optimization objective function of the MPC algorithm is J=ω1×(vrear-vfront)²+ω2×(xrear-xfront-Δxtarget)², where ω1 is the speed difference weight, ω2 is the position deviation weight, and Δxtarget is the preset safe distance. The MPC algorithm updates the denoised state data of the front vehicle every 0.1-0.3 seconds, re-solves the constrained optimization objective function, and dynamically adjusts the target control parameters of the rear vehicle. The target control parameters include target acceleration, target steering angle, and braking force.

[0008] Preferably, the safe following distance is calculated using a dynamic safe following distance model, which is: following distance = reaction time × vehicle speed + safety redundancy distance, where the reaction time is 0.5 seconds and the safety redundancy distance is 2 meters; in the optimization objective function, ω1=0.6 and ω2=0.4, and the optimization problem is solved using a quadratic programming solver, which is an OSQP solver. The output adjustment step size of the target control parameters of the following vehicle satisfies the following conditions: the acceleration adjustment step size does not exceed 0.1 m / s², and the steering angle adjustment step size does not exceed 0.5°.

[0009] Preferably, both the communication sending module and the communication receiving module adopt 5G-V2X communication modules, and the communication mode is 5G-V2XPC5 direct connection communication mode; the feedforward control signal includes the target control parameters of the rear vehicle and the sending timestamp, the signal transmission delay is controlled within 100ms, the signal transmission success rate is not less than 99.9%, and a packet loss retransmission mechanism is supported, with a retransmission delay ≤50ms.

[0010] Preferably, the weighted fusion strategy of the fusion control module is as follows: the weight ratio of the feedforward control signal is 40%-60%, and the weight ratio of the feedback signal is 60%-40%; the feedback signal is generated based on the noise-reduced self-state data of the following vehicle and the real-time feedback information of the preceding vehicle; the fusion control module dynamically adjusts the weight allocation according to the driving conditions, with a feedforward weight of 50% and a feedback weight of 50% when driving on a straight road, and a feedforward weight of 60% and a feedback weight of 40% when driving on a curve.

[0011] Preferably, the actuator includes a power system, a steering system, and a braking system. The execution command generated by the fusion control module is transmitted to the actuator via a CAN bus. The response delay of the actuator does not exceed 200ms. After the actuator is activated, the speed difference between the following vehicle and the preceding vehicle is ≤0.5m / s, and the position deviation is ≤0.3m, which meets the requirements for maintaining the formation.

[0012] Preferably, the state prediction module of the preceding vehicle system works in collaboration with the control estimation module of the following vehicle. The future state data of the preceding vehicle output by the state prediction module is used as one of the input parameters of the control estimation module of the following vehicle, so that the MPC algorithm fully considers the subsequent motion trend of the preceding vehicle when estimating the control parameters of the following vehicle, thereby improving the foresight and adaptability of the control parameters. The self-state detection module of the following vehicle performs validity verification on the denoised self-state data, removes abnormal data, and then transmits it to the fusion control module.

[0013] Preferably, the formation constraints can be dynamically adjusted according to the actual driving scenario, which includes driving straight on highways, driving on curved mountain roads, and following other vehicles in congested areas. When following other vehicles in congested areas, the safety redundancy distance is adjusted to 1 meter, the prediction time domain of the MPC algorithm is adjusted to 1 second, the control time domain is adjusted to 0.5 seconds, and the upper limit of the number of retransmissions of the signal is set to 3 times to ensure the flexibility and scenario adaptability of formation control.

[0014] Compared with existing technologies, this solution designs an autonomous truck coordinated platooning control system based on a feedforward mechanism, which has the following beneficial effects: Significantly improved response performance and dynamic adaptability: MPC continuous rolling optimization (updating control parameters every 0.1-0.3 seconds) can adapt to changes in the status of the vehicle in front in real time. Combined with the feedforward signal to "transmit control requirements 100ms in advance", it can double offset the effects of communication latency and actuator lag. In scenarios where the vehicle in front accelerates urgently (acceleration 1.2m / s²) and turns quickly (steering angle 8°), the response latency of the vehicle behind is reduced by more than 60% compared with the existing technology, avoiding large fluctuations in the distance between vehicles.

[0015] The accuracy of state detection and the stability of control are significantly enhanced: the 5-point median filtering algorithm removes peak noise at a rate of ≥95%, effectively reducing sensor data deviation; the constraint optimization capability of the MPC algorithm can control the position deviation between the following vehicle and the preceding vehicle within 0.3m, which is 40%-62.5% lower than the existing technology (deviation 0.5-0.8m), and improves the stability of platooning driving by more than 50%. Attached Figure Description

[0016] Figure 1 This is a diagram of the architecture of the autonomous truck platooning control system based on the feedforward mechanism of the present invention. Detailed Implementation

[0017] Please see Figure 1The technical solution provided by this invention is as follows: an autonomous driving truck coordinated platooning control system based on a feedforward mechanism, comprising a leading vehicle system and a trailing vehicle system. The leading and trailing vehicle systems interact with each other via vehicle-to-vehicle communication technology to collaboratively complete platooning control. The leading vehicle system includes an onboard sensor module, a median filter module, a state prediction module, a trailing vehicle control estimation module, and a communication transmission module. The trailing vehicle system includes an onboard sensor module, a median filter module, a self-state detection module, a communication receiving module, a fusion control module, and an actuator. The leading vehicle system collects its own state data through sensors, performs denoising and prediction, and uses a model predictive control (MPC) algorithm to estimate the control parameters of the trailing vehicle and sends a feedforward control signal to the trailing vehicle. The trailing vehicle system collects its own state data and performs denoising. After receiving the feedforward control signal, it fuses its own state data with the feedback information from the leading vehicle to generate execution commands, driving the actuators to move and achieving dynamic and stable platooning control. Specifically, the system includes the following steps: S1 Forward Vehicle Status Measurement and Preprocessing: The forward vehicle collects its own real-time status data through on-board sensors at a preset sampling frequency. The status data includes position, vehicle speed, acceleration, steering angle and braking signal. The signal spike noise is removed by the median filtering module. S2 Forward Vehicle Status Prediction: Based on the denoised status data, the forward vehicle's status change trend within a preset time period is predicted through a preset motion prediction model. S3 Rear Vehicle Control Parameter Optimization and Calculation: The leading vehicle continuously optimizes the target control parameters of the following vehicle by combining its own real-time status, predicted status and formation constraints, using the MPC algorithm. S4 control signal transmission: The leading vehicle transmits the optimized feedforward control signal to the trailing vehicle in real time through the communication transmission module; S5 Rear Vehicle Status Preprocessing and Signal Reception: The rear vehicle collects real-time status data through its own on-board sensors and performs median filtering to remove noise, while simultaneously receiving feedforward control signals sent by the preceding vehicle. S6 Fusion Control and Execution: The rear vehicle fusion control module performs weighted fusion of the feedforward control signal and the feedback signal, generates execution instructions and transmits them to the actuator to complete the formation maintenance.

[0018] The onboard sensor modules of both the front and rear vehicle systems include GPS, IMU, and millimeter-wave radar, with a sampling frequency set to 100Hz. The median filtering module employs a 3-5 point median filtering algorithm. The sliding window of the 3-5 point median filtering algorithm includes the current sampling point and at least one historical sampling point. The removal rate of instantaneous peak noise such as speed jumps caused by bumps and position deviations caused by electromagnetic interference is no less than 95%, ensuring the smoothness of the status data.

[0019] The state prediction module uses a Kalman filter algorithm combined with a vehicle dynamics model to construct a motion prediction model. The vehicle dynamics model is a three-degree-of-freedom vehicle dynamics model. The preset time is 0.5-2 seconds. The future state of the vehicle in front, calculated based on the motion prediction model, includes changes in future vehicle speed, position, and steering angle. The prediction error is controlled within 5%.

[0020] The MPC algorithm of the rear vehicle control calculation module is set to a prediction time domain of 1-3 seconds and a control time domain of 0.5-1.5 seconds. The formation constraints include safe distance, speed synchronization accuracy, and formation maintenance deviation threshold. The optimization objective function of the MPC algorithm is J=ω1×(vrear-vfront)²+ω2×(xrear-xfront-Δxtarget)², where ω1 is the speed difference weight, ω2 is the position deviation weight, and Δxtarget is the preset safe distance. The MPC algorithm updates the denoised state data of the front vehicle every 0.1-0.3 seconds, re-solves the constrained optimization objective function, and dynamically adjusts the target control parameters of the rear vehicle. The target control parameters include target acceleration, target steering angle, and braking force.

[0021] The safe following distance is calculated using a dynamic safe following distance model, which is: following distance = reaction time × vehicle speed + safety redundancy distance, where the reaction time is 0.5 seconds and the safety redundancy distance is 2 meters. In the optimization objective function, ω1=0.6 and ω2=0.4. The optimization problem is solved using a quadratic programming solver, which is an OSQP solver. The output adjustment step size of the target control parameters of the following vehicle satisfies the following conditions: the acceleration adjustment step size does not exceed 0.1 m / s², and the steering angle adjustment step size does not exceed 0.5°.

[0022] Both the communication sending module and the communication receiving module adopt 5G-V2X communication modules, and the communication mode is 5G-V2XPC5 direct connection communication mode; the feedforward control signal includes the target control parameters of the rear vehicle and the sending timestamp, the signal transmission delay is controlled within 100ms, the signal transmission success rate is not less than 99.9%, and it supports a packet loss retransmission mechanism with a retransmission delay ≤50ms.

[0023] The weighted fusion strategy of the fusion control module is as follows: the weight ratio of the feedforward control signal is 40%-60%, and the weight ratio of the feedback signal is 60%-40%. The feedback signal is generated based on the noise-reduced self-state data of the following vehicle and the real-time feedback information of the preceding vehicle. The fusion control module dynamically adjusts the weight allocation according to the driving conditions: when driving on a straight road, the feedforward weight is 50% and the feedback weight is 50%; when driving on a curve, the feedforward weight is 60% and the feedback weight is 40%.

[0024] The actuator includes a power system, a steering system, and a braking system. The execution commands generated by the fusion control module are transmitted to the actuator via the CAN bus. The response delay of the actuator does not exceed 200ms. After the actuator is activated, the speed difference between the following vehicle and the preceding vehicle is ≤0.5m / s, and the position deviation is ≤0.3m, which meets the requirements for maintaining the formation.

[0025] The state prediction module of the preceding vehicle system works in collaboration with the control estimation module of the following vehicle. The future state data of the preceding vehicle output by the state prediction module is used as one of the input parameters of the control estimation module of the following vehicle, so that the MPC algorithm fully considers the subsequent motion trend of the preceding vehicle when estimating the control parameters of the following vehicle, thereby improving the foresight and adaptability of the control parameters. The self-state detection module of the following vehicle performs validity verification on the denoised self-state data, removes abnormal data, and then transmits it to the fusion control module.

[0026] The formation constraints can be dynamically adjusted according to the actual driving scenarios, which include driving straight on highways, driving on curved mountain roads, and following other vehicles in congested areas. When following other vehicles in congested areas, the safety redundancy distance is adjusted to 1 meter, the prediction time domain of the MPC algorithm is adjusted to 1 second, the control time domain is adjusted to 0.5 seconds, and the maximum number of retransmissions of the signal is set to 3 times to ensure the flexibility and adaptability of formation control.

[0027] Example 1: Straight-line platooning control of autonomous trucks on highways 1. Scene setting Scenario: A six-lane highway in both directions (speed limit 100km / h), with a dry road surface (coefficient of adhesion 0.85) and no significant electromagnetic interference; Formation configuration: 2 autonomous trucks (the lead truck is a tractor unit plus a semi-trailer, with a length of 16m; the follower truck is the same model), target safe distance = 0.5s × vehicle speed + 2m (e.g., at a vehicle speed of 100km / h, the safe distance is approximately 13.9m + 2m = 15.9m). Control targets: speed difference between the following vehicle and the preceding vehicle ≤ 0.3 m / s, longitudinal position deviation ≤ 0.2 m, lateral position deviation ≤ 0.1 m.

[0028] 2. Implementation Steps (1) Deployment and preprocessing of the front vehicle system Hardware deployment: The front vehicle is equipped with UBlox NEO-M8T GPS, ADIADIS16488IMU, Continental ARS408 millimeter-wave radar, and the median filter module, state prediction module, and rear vehicle control calculation module are integrated into the vehicle controller (model: NXPS32G399). State measurement and noise reduction: The sensor collects data at a sampling frequency of 100Hz. After the 5-point median filter module processes the vehicle speed signal (the raw data has a ±0.8m / s spike), the smoothness of the vehicle speed data is improved to 98%. State prediction: Based on the denoised data (vehicle speed 27.8m / s, i.e. 100km / h), the Kalman filter + three-degree-of-freedom model predicts that the speed of the vehicle in front will remain at 27.8m / s and the position will move forward by 27.8m in the next second, with a prediction error ≤0.2m / s.

[0029] Rear vehicle control parameter calculation (MPC optimization) MPC parameter settings: prediction time domain 2 seconds, control time domain 1 second, front vehicle state updated every 0.2 seconds; optimization objective function: \(J=0.6×(v_{rear}-27.8)^2+0.4×(x_{rear}-x_{front}-15.9)^2\); solution and output: OSQP solver solves within 45ms, outputting rear vehicle target control parameters: target acceleration 0m / s² (maintaining speed), target steering angle 0° (straight road), braking force 0%.

[0030] (3) Communication transmission The 5G-V2X communication transmitting module (model: Huawei MH5000-31) in the front vehicle transmits feedforward signals in PC5 mode with a transmission delay of 82ms. The communication receiving module (same model) in the rear vehicle successfully receives the signals and sends a "receive confirmation" response.

[0031] (4) Rear vehicle fusion control and execution Rear vehicle status preprocessing: The sensor collects its own vehicle speed of 27.6 m / s. After denoising by 5-point median filtering, the vehicle speed data is 27.7 m / s; the self-state detection module verification shows no abnormalities. Fusion control: In straight-line scenarios, the feedforward weight is 50% and the feedback weight is 50%; the feedforward signal requires "acceleration of 0 m / s²", and the feedback signal calculates "acceleration of 0.1 m / s² is required to reduce the speed difference". The final acceleration command after fusion is 0.05 m / s². Execution and effect: The actuator (power system) responded to the command, and after 150ms, the speed of the following vehicle increased to 27.75m / s, the speed difference with the preceding vehicle was 0.05m / s, the longitudinal position deviation was 0.15m, and the lateral position deviation was 0.08m, which fully met the control target.

[0032] Example 2: Autonomous Truck Formation Control on Mountain Roads 1. Scene setting Scenario: A two-lane road in a mountainous area (speed limit 60km / h), with a right curve of radius R=150m, wet road surface (adhesion coefficient 0.7), and local electromagnetic interference (radar signal has a positional deviation of ±0.5m). Formation configuration: 2 autonomous driving trucks (same as the model in Example 1), target safe distance = 0.5s × vehicle speed + 2m (when the vehicle speed is 60km / h, the safe distance is ≈8.3m + 2m = 10.3m). Control objectives: speed difference between the following vehicle and the preceding vehicle ≤ 0.5 m / s, longitudinal position deviation ≤ 0.3 m, lateral position deviation ≤ 0.2 m, and steering angle following error ≤ 1°.

[0033] 2. Implementation Steps (1) Deployment and preprocessing of the front vehicle system Hardware deployment: Same as in Example 1, except that an anti-electromagnetic interference shield is added to the millimeter-wave radar; State measurement and noise reduction: The sensor collected the vehicle speed of the vehicle in front of it at 16.7 m / s (60 km / h) and the steering angle at 5° (entering the right curve). The original position data contained a peak of ±0.5 m. After 5-point median filtering, the position deviation was reduced to ±0.05 m. State prediction: Predicts that within the next second, the steering angle of the vehicle in front will increase to 8° (mid-cornering) and the vehicle speed will decrease to 16.5m / s (cornering deceleration), with a prediction error of ≤0.3° (steering angle) and 0.1m / s (vehicle speed).

[0034] Rear vehicle control parameter estimation (MPC optimization) MPC parameter settings: prediction time domain 1.5 seconds, control time domain 0.8 seconds, update the status of the vehicle in front every 0.15 seconds (the status of the vehicle changes quickly in the curve, so shorten the update cycle). The objective function is optimized as follows: \(J=0.6×(v_{rear}-16.7)^2+0.4×(x_{rear}-x_{front}-10.3)^2\), with an additional "steering angle following constraint" (the difference between the steering angle of the rear vehicle and the predicted steering angle of the front vehicle is ≤1°). Solution and Output: The OSQP solver solves the problem within 48ms and outputs the target control parameters for the following vehicle: target acceleration -0.2m / s² (decelerating with the preceding vehicle), target steering angle 6° (adapting to the curve in advance), and braking force 10%.

[0035] (3) Communication transmission The leading vehicle sends a feedforward signal with a transmission delay of 95ms. When the following vehicle receives the signal, it experiences one packet loss. The leading vehicle triggers a retransmission mechanism with a retransmission delay of 45ms, and the following vehicle successfully receives the signal.

[0036] (4) Rear vehicle fusion control and execution Rear vehicle status preprocessing: The sensor collects its own vehicle speed of 16.9 m / s and steering angle of 3°. After denoising by 5-point median filtering, the vehicle speed is 16.8 m / s and the steering angle is 3.1°. The self-state detection module verification shows no abnormalities. Fusion control: In the curve scenario, the feedforward weight is 60% and the feedback weight is 40%; the feedforward signal requires "acceleration -0.2m / s², steering angle 6°", and the feedback signal calculates "deceleration 0.1m / s², steering angle increase 1.9°". The final command after fusion is: acceleration -0.16m / s², steering angle 5.16°. Execution and Effect: The actuators (braking system + steering system) responded to the command. After 180ms, the speed of the following vehicle dropped to 16.57m / s (0.13m / s difference from the speed of the preceding vehicle), the steering angle was 5.2° (0.2° error from the current steering angle of the preceding vehicle of 5°), the longitudinal position deviation was 0.22m, and the lateral position deviation was 0.15m. The control objectives were fully met, and there was no risk of following deviation.

[0037] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0038] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An autonomous truck coordinated platoon control system based on feedforward mechanism, characterized in that, The application relates to a platoon control system, which comprises a front vehicle system and a rear vehicle system, and the front vehicle system and the rear vehicle system realize data interaction through vehicle-to-vehicle communication technology and cooperatively complete platoon control; the front vehicle system comprises a vehicle-mounted sensor module, a median filter module, a state prediction module, a rear vehicle control calculation module and a communication sending module; the rear vehicle system comprises a vehicle-mounted sensor module, a median filter module, a self-state detection module, a communication receiving module, a fusion control module and an execution mechanism; the front vehicle system collects self-state data through sensors, removes noise and predicts the data, adopts a model predictive control (MPC) algorithm to calculate rear vehicle control parameters and sends the control parameters to the rear vehicle as a feedforward control signal; the rear vehicle system collects self-state data and removes noise, generates an execution instruction by fusing self-state data and front vehicle feedback information after receiving the feedforward control signal, drives the execution mechanism to act and realizes dynamic stable control of the platoon, and specifically comprises the following steps: S1. Front vehicle state measurement and pretreatment: the front vehicle collects real-time state data of itself through vehicle-mounted sensors at a preset sampling frequency, the state data includes position, vehicle speed, acceleration, steering angle and brake signal, and the state data is removed from signal peak noise through a median filter module; S2. Front vehicle state prediction: based on the state data after noise removal, the state change trend of the front vehicle in a future preset time is calculated through a preset motion prediction model; S3. Rear vehicle control parameter optimization calculation: the front vehicle continuously and rollingly optimizes the target control parameters of the rear vehicle by adopting an MPC algorithm in combination with real-time state, predicted state and platoon constraint conditions of itself; S4. Control signal transmission: the front vehicle transmits the optimized feedforward control signal to the rear vehicle in real time through a communication sending module; S5. Rear vehicle state pretreatment and signal receiving: the rear vehicle collects real-time state data through self-mounted sensors and removes noise through a median filter, and simultaneously receives the feedforward control signal sent by the front vehicle; S6. Fusion control and execution: the rear vehicle generates an execution instruction by weighting and fusing the feedforward control signal and the feedback signal through a fusion control module and transmits the execution instruction to an execution mechanism, thereby completing platoon formation maintenance.

2. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, characterized in that: The vehicle-mounted sensor modules of the front vehicle system and the rear vehicle system both comprise GPS, IMU and millimeter wave radar, and the sampling frequency is set to 100 Hz; the median filter module adopts a 3-5 point median filter algorithm, the sliding window of the 3-5 point median filter algorithm contains a current sampling point and at least one historical sampling point, the removal rate of instantaneous peak noise caused by vehicle speed jump due to bumping and position deviation caused by electromagnetic interference is not less than 95%, and the state data smoothness is ensured.

3. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, characterized in that: The state prediction module adopts a Kalman filter algorithm to combine a vehicle dynamics model to construct a motion prediction model, the vehicle dynamics model is a three-degree-of-freedom vehicle dynamics model, the preset time is 0.5-2 seconds, the future state of the front vehicle calculated based on the motion prediction model includes future vehicle speed, position and steering angle change, and the prediction error is controlled within 5%.

4. The automatic driving truck coordinated platoon control system based on feedforward mechanism according to claim 1, characterized in that, The MPC algorithm of the rear vehicle control calculation module sets the prediction time domain to 1-3 seconds and the control time domain to 0.5-1.5 seconds, and the platoon constraint conditions include a safe vehicle distance, a speed synchronization accuracy and a platoon keeping deviation threshold; the optimization objective function of the MPC algorithm is J = ω1 × (vback-vfront) 2 + ω2 × (xback-xfront-Δxtarget) 2, wherein ω1 is a speed difference weight, ω2 is a position deviation weight, and Δxtarget is a preset safe vehicle distance; the MPC algorithm updates the state data of the front vehicle after denoising once every 0.1-0.3 seconds, re-solves the optimization objective function with constraints, and dynamically adjusts the target control parameters of the rear vehicle, the target control parameters including a target acceleration, a target steering angle and a braking force.

5. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, wherein, The safe vehicle distance is calculated by a dynamic safe vehicle distance model, the dynamic safe vehicle distance model being: distance = reaction time × speed + safety redundancy distance, wherein the reaction time is 0.5 seconds and the safety redundancy distance is 2 meters; in the optimization objective function, ω1 = 0.6 and ω2 = 0.4, the optimization problem is solved by a quadratic programming solver, the quadratic programming solver being an OSQP solver, and the output target control parameter adjustment step of the rear vehicle satisfies: the acceleration adjustment step is not more than 0.1 m / s 2 and the steering angle adjustment step is not more than 0.5°.

6. The automatic driving truck coordination platoon control system based on feedforward mechanism of claim 1, wherein: The communication sending module and the communication receiving module both use a 5G-V2X communication module, and the communication mode is a 5G-V2X PC5 direct communication mode; the feedforward control signal includes the target control parameters of the rear vehicle and a sending timestamp, the signal transmission delay is controlled within 100 ms, the signal transmission success rate is not less than 99.9%, the lost packet retransmission mechanism is supported, and the retransmission delay is ≤50 ms.

7. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, characterized in that: The weighting fusion strategy of the fusion control module is that the weight proportion of the feedforward control signal is 40%-60% and the weight proportion of the feedback signal is 60%-40%; the feedback signal is generated based on the denoised state data of the rear vehicle and the real-time feedback information of the front vehicle, and the fusion control module dynamically adjusts the weight distribution according to the driving conditions, the front weight being 50% and the feedback weight being 50% when driving on a straight road, and the front weight being 60% and the feedback weight being 40% when driving on a curve.

8. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, characterized in that: The execution mechanism includes a power system, a steering system and a braking system, the execution instructions generated by the fusion control module are transmitted to the execution mechanism through a CAN bus, and the response delay of the execution mechanism is not more than 200 ms; after the execution mechanism acts, the speed difference between the rear vehicle and the front vehicle is ≤0.5 m / s and the position deviation is ≤0.3 m, meeting the platoon keeping requirements.

9. The automatic driving truck coordination platoon control system based on feedforward mechanism according to claim 1, characterized in that: The state prediction module of the front vehicle system cooperates with the rear vehicle control calculation module, the future state data of the front vehicle output by the state prediction module is used as one of the input parameters of the rear vehicle control calculation module, the MPC algorithm fully considers the subsequent motion trend of the front vehicle when calculating the control parameters of the rear vehicle, and the foresight and adaptability of the control parameters are improved; the self-state detection module of the rear vehicle effectively checks the denoised self-state data, and transmits the data to the fusion control module after removing abnormal data.

10. The automatic driving truck coordinated platoon control system based on feedforward mechanism of claim 1, wherein: The platoon constraint condition can be dynamically adjusted according to an actual driving scene, and the actual driving scene includes straight driving on a highway, curved driving on a mountain road, and following on a congested road section. When in the following scene on a congested road section, the safety redundant distance is adjusted to 1 meter, the prediction time domain of the MPC algorithm is adjusted to 1 second, the control time domain is adjusted to 0.5 second, and the upper limit of the retransmission number of signal transmission is set to 3 times, so as to ensure the flexibility and scene adaptability of the platoon control.

Images