Artificially driven vehicle behavior prediction system and method for mixed fleet control

Abstract

The application discloses a kind of artificial driving vehicle behavior prediction systems and methods for hybrid vehicle fleet control, belong to intelligent networked vehicle field.System is deployed on networked automatic driving vehicle, and the motion state information of target artificial driving vehicle and its surrounding vehicle is obtained by environment perception module;Artificial driving vehicle following behavior prediction model is built-in in behavior prediction calculation module, and the following intention of target artificial driving vehicle is modeled, and the future predicted motion state of target artificial driving vehicle is calculated and output based on motion state information;Predicted output interface sends future predicted motion state to the decision planning module of the car.The application solves the control problem caused by the uncertainty of artificial driving vehicle behavior in hybrid vehicle fleet, and provides a key basis for networked automatic driving vehicle to develop safe and efficient cooperative control strategy by providing high-precision behavior prediction, which significantly improves the overall stability and traffic efficiency of hybrid vehicle fleet.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle technology, and in particular to a system and method for predicting the behavior of manually driven vehicles for hybrid fleet control. Background Technology

[0002] In the development of intelligent connected vehicles, mixed traffic flow will be a long-term, normalized scenario. In this scenario, connected autonomous vehicles (CAVs) and human-driven vehicles coexist. CAVs can achieve efficient collaboration through precise communication and control, while the behavior of human-driven vehicles (HDVs) is influenced by various factors such as the driver's subjective state, driving habits, and reaction delays, exhibiting a high degree of uncertainty and randomness. This uncertainty makes it impossible for CAVs to accurately predict the next action of HDVs, becoming the biggest challenge to achieving safe and efficient collaborative control of mixed vehicle fleets.

[0003] Most existing solutions focus on improving the control algorithms of CAVs themselves or using overly simplified car-following models (such as Intelligent Driver Models (IDMs)) to simulate HDV behavior. These simplified models fail to fully consider the complexity of real human driver decision-making (such as paying attention to vehicles behind, observing multiple vehicles in front, and having an aggressive or conservative driving style), resulting in significant prediction bias. Control decisions based on biased predictions are either too conservative, reducing traffic efficiency, or too aggressive, creating safety hazards.

[0004] Therefore, there is an urgent need for a high-precision human-driven vehicle behavior prediction technology in the field of hybrid vehicle fleet control. This technology should be deployable on CAVs as a dedicated "environmental cognition" subsystem, making full use of sensor information and employing an advanced model that fully considers the characteristics of human driving behavior to predict the movement intentions of HDVs ahead or in the surrounding area in real time and accurately. This would provide reliable data input for the CAV's decision-making and planning system, fundamentally improving the performance of hybrid vehicle fleet control. Summary of the Invention

[0005] This invention provides a system and method for predicting the behavior of manually driven vehicles (HDVs) in hybrid vehicle fleet control. The system is deployed on a CAV and aims to accurately predict the future motion state of HDVs through a prediction model that deeply integrates multiple driving behavior factors. This solves the difficulty of control decision-making caused by CAVs' "lack of understanding" of HDV behavior, and ultimately improves the coordination, safety and efficiency of hybrid vehicle fleets.

[0006] A first aspect of the present invention provides a human-driven vehicle behavior prediction system for hybrid fleet control, deployed on a connected autonomous vehicle, the system comprising: The environmental perception module is used to acquire information on the vehicle motion status of the target manually driven vehicle and its surrounding vehicles within the target monitoring area. The behavior prediction calculation module is communicatively connected to the environment perception module. It is used to receive vehicle motion state information acquired by the environment perception module and calculate the future predicted motion state of the target manually driven vehicle using the built-in car-following behavior prediction model. The car-following behavior prediction model is used to mathematically represent the car-following behavior of the target manually driven vehicle through rear-view weight factor, memory strength coefficient and memory time, multi-front vehicle sensitivity coefficient and driving style weight factor, so as to simulate its decision-making process. The prediction output interface is communicatively connected to the behavior prediction calculation module and is used to send the predicted future motion state to the vehicle's decision-making and planning system to generate control commands.

[0007] Optionally, in one embodiment of the present invention, the future predicted motion state includes the future expected acceleration, future speed, and future position of the target manually driven vehicle.

[0008] Optionally, in one embodiment of the present invention, the behavioral characteristics of the target manually driven vehicle are quantified using the following parameters in the car-following behavior prediction model: Backsight weight factor p This is used to adjust the intensity of the influence of the distance between the vehicle in front and behind on the behavior of the target manually driven vehicle. The memory strength coefficient β and memory time m are used to jointly modulate the intensity of the influence of historical driving information on the behavior of the target manually driven vehicle. The multi-vehicle sensitivity coefficient q is used to adjust the intensity of the influence of the motion state information of the preceding vehicle and the vehicle before that on the behavior of the target manually driven vehicle. The driving style weighting factor h is used to quantify the individual characteristics of the driver of the target manually driven vehicle and distinguish driving style preferences.

[0009] Optionally, in one embodiment of the present invention, the car-following behavior prediction model is: ; In the formula, v n ( t Let n be the real-time speed of the target manually driven vehicle n. F f and F b Let represent the optimal speed functions for the driver's observation of vehicles in front and behind, respectively. C 1 represents the stimulus factor coefficient of the desired vehicle speed on the driver. C 2 represents the stimulus factor coefficient of the speed difference between multiple vehicles in front of the driver; Δ xn ( t )= x n+1 ( t )- x n ( t )express t The distance Δ between the target manually driven vehicle n and the preceding vehicle n+1 at any given time. v n ( t )= v n+1 ( t )- v n ( t ) represents the speed difference between the target manually driven vehicle n and the preceding vehicle n+1 at time t; α1 and α2 represent the prediction ability coefficients of the aggressive driver and the conservative driver, respectively; τ is the driver's reaction delay time. r As the forward weighting factor in the memory effect, m For memory time; Calculate the optimal speed function for the vehicle in front and the optimal speed function for the vehicle behind: ; ; In the formula, v max This indicates the maximum speed of the vehicles in the platoon. D max and D min These represent the minimum and maximum spacing between adjacent vehicles in the vehicle queue, respectively.

[0010] Optionally, in one embodiment of the present invention, the behavior prediction calculation module further includes a model parameter calibration unit. The model parameter calibration unit takes a historical real vehicle trajectory dataset as input and uses an optimization algorithm to automatically calibrate the rear-view weight factor, memory strength coefficient, multi-vehicle sensitivity coefficient and driving style weight factor, so as to minimize the error between the output of the car-following behavior prediction model and the actual driving behavior.

[0011] Optionally, in one embodiment of the present invention, the environmental perception module includes a vehicle-to-everything (V2X) communication device, and also includes at least one of an onboard camera, millimeter-wave radar, and lidar; the vehicle-to-everything (V2X) communication device is used to receive motion status information broadcast by the target manually driven vehicle or vehicles around it.

[0012] Optionally, in one embodiment of the present invention, the decision planning system makes a hybrid vehicle cooperative control decision based on the received future predicted motion state. The objectives of the hybrid vehicle cooperative control decision include at least one of maintaining vehicle speed consistency, maintaining a safe following distance, and suppressing the propagation of traffic flow disturbances. The system also generates longitudinal control commands for controlling the vehicle's state based on the hybrid vehicle cooperative control decision.

[0013] A second aspect of the present invention provides a method for predicting the behavior of manually driven vehicles for hybrid fleet control, used in the manually driven vehicle behavior prediction system for hybrid fleet control described in the above embodiments, the method comprising: Acquire information on the vehicle motion status of the target manually driven vehicle and its surrounding vehicles within the target monitoring area; Based on the vehicle motion state information, the future predicted motion state of the target manually driven vehicle is calculated using a car-following behavior prediction model. The car-following behavior prediction model is used to mathematically represent the car-following behavior of the target manually driven vehicle through rear-view weight factor, memory strength coefficient, memory time, multi-vehicle sensitivity coefficient and driving style weight factor, so as to simulate its decision-making process. Control commands are generated based on the predicted future motion state.

[0014] A third aspect of the present invention provides a connected autonomous vehicle, including the human-driven vehicle behavior prediction system for hybrid fleet control described in the above embodiments.

[0015] The system and method for predicting the behavior of manually driven vehicles for hybrid fleet control according to embodiments of the present invention have the following beneficial effects: 1. High prediction accuracy: By integrating four key factors—rearview effect, memory effect, multiple preceding vehicle observations, and driving style—the model provides a more comprehensive and accurate characterization of HDV behavior, significantly improving the reliability of predictions.

[0016] 2. Improved control performance: By providing accurate HDV behavior prediction, CAVs can formulate more forward-looking and smooth control strategies, effectively improving the overall stability and traffic efficiency of mixed vehicle fleets.

[0017] 3. Excellent adaptability: The system supports online parameter calibration, enabling it to adapt to the individual characteristics of different drivers and various traffic conditions.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram illustrating the overall architecture of a human-driven vehicle behavior prediction system for hybrid fleet control according to an embodiment of the present invention, and its application scenario in a hybrid fleet. Figure 2 This is a schematic diagram of the internal workflow of the behavior prediction calculation module in an embodiment of the present invention; Figure 3 This is a flowchart of a method for predicting the behavior of manually driven vehicles for hybrid fleet control, according to an embodiment of the present invention. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0021] Figure 1 This is a schematic diagram of the structure of a human-driven vehicle behavior prediction system for hybrid fleet control according to an embodiment of the present invention.

[0022] like Figure 1 As shown, this human-driven vehicle behavior prediction system for hybrid fleet control is deployed on a connected autonomous vehicle. The system includes: The environmental perception module is used to acquire information on the vehicle motion status of the target manually driven vehicle and its surrounding vehicles within the target monitoring area. The behavior prediction calculation module communicates with the environment perception module to receive vehicle motion state information acquired by the environment perception module and uses the built-in car-following behavior prediction model to calculate the future predicted motion state of the target manually driven vehicle. The car-following behavior prediction model mathematically represents the car-following behavior of the target manually driven vehicle through rear-view weight factor, memory strength coefficient, memory time, multi-front vehicle sensitivity coefficient and driving style weight factor to simulate its decision-making process. The prediction output interface communicates with the behavior prediction calculation module and is used to send the predicted future motion state to the vehicle's decision-making and planning system to generate control commands.

[0023] In an embodiment of the present invention, Figure 1 The invention also demonstrates typical application scenarios and complete data flow of the system in a hybrid fleet. The human-driven vehicle behavior prediction system for hybrid fleet control is integrated into the computing platform of connected autonomous vehicles in either post-installed or pre-installed form, and works in collaboration with the vehicle's existing system.

[0024] In embodiments of the present invention, the environmental perception module must include a vehicle-to-everything (V2X) communication device, and further include at least one of an onboard camera, millimeter-wave radar, and lidar. The V2X communication device is used to receive motion status information broadcast by the target manually driven vehicle or vehicles in its vicinity.

[0025] Specifically, the environmental perception module, as the data input source of the system, is responsible for collecting detailed motion information of one or more manually driven vehicles (HDVs) within the target monitoring area through onboard sensors (such as cameras, radar, and lidar) and V2X communication equipment. This includes the distance between the vehicle's front end and the vehicles in front, behind, left, and right, its relative speed, absolute speed, acceleration, etc., to construct a real-time, dynamic local traffic environment dataset.

[0026] In one specific embodiment, multiple onboard sensors simultaneously collect environmental data: a forward-facing millimeter-wave radar continuously detects the relative distance and speed of the target HDV; a V2X communication device receives BSM messages (containing information such as position, speed, and acceleration) broadcast from vehicles ahead (n+1, n+2) and behind (n-1); and a GPS / IMU integrated navigation system provides precise positioning for the vehicle. All these multi-source heterogeneous data are input to a data fusion unit for time synchronization, coordinate unification, and filtering, ultimately generating a unified, timestamped local traffic environment dataset.

[0027] The behavior prediction calculation module serves as the core computing engine of the system. This module pre-installs an advanced car-following behavior prediction model. The core innovation of this model lies in its mathematical framework, which simultaneously quantifies four key factors influencing human drivers: the rear-view effect (the psychological fear of being rear-ended), the memory effect (adjusting current actions based on past road conditions and the driving habits of surrounding vehicles), the multi-vehicle observation effect (looking further ahead to anticipate risks), and personalized driving style (an inherent tendency towards aggressive or conservative driving). This module receives real-time data from the environmental perception module, runs the prediction model, and outputs a predicted future motion state for the target HDV, such as expected acceleration, speed, or trajectory.

[0028] In embodiments of the present invention, the car-following behavior prediction model uses a real-time local traffic environment dataset constructed by the environmental perception module to predict the motion state of a human-driven vehicle. In the car-following behavior prediction model, the behavioral characteristics of the target human-driven vehicle are quantified through the following parameters: Backsight weight factor p This is used to adjust the intensity of the influence of the distance between the vehicle in front and behind on the behavior of the target manually driven vehicle. Memory strength coefficient β and memory time m This is used to jointly regulate the intensity of the influence of historical driving information on the behavior of target manually driven vehicles; Multiple front vehicle sensitivity coefficient q This is used to adjust the intensity of the influence of the motion state information of the preceding vehicle and the vehicle before that on the behavior of the target manually driven vehicle. Driving style weighting factor h It is used to quantify the individual characteristics of drivers of target manually driven vehicles and distinguish driving style preferences.

[0029] In one specific embodiment, the environmental perception module transmits the processed traffic environment dataset to the behavior prediction calculation module in real time. This module is the core computing engine, and its internally pre-built car-following behavior prediction model calls the latest parameter configuration (including...) from the model parameter library. p The model performs calculations based on the input data, including β, m, q, h, etc. During the calculation, the model comprehensively considers the current instantaneous state and historical information (by querying the historical driving database), and finally outputs the predicted future motion state of the target HDV, including the expected acceleration, velocity, and trajectory.

[0030] In embodiments of the present invention, such as Figure 2 The diagram illustrates the internal workflow and data processing of the behavior prediction calculation module in this invention. This module is responsible for calculating the future motion state of the target manually driven vehicle based on environmental perception data and a high-precision car-following behavior prediction model.

[0031] In one embodiment of the present invention, the behavior prediction calculation module further includes a model parameter calibration unit. The model parameter calibration unit takes a historical real vehicle trajectory dataset as input and uses an optimization algorithm to automatically calibrate the rear-view weight factor, memory strength coefficient, multi-vehicle sensitivity coefficient and driving style weight factor, so as to minimize the error between the output of the car-following behavior prediction model and the real driving behavior.

[0032] The established car-following behavior prediction model is as follows: ; In the formula, v n ( t Let n be the real-time speed of the target manually driven vehicle n. F f and F b Let represent the optimal speed functions for the driver's observation of vehicles in front and behind, respectively. C 1 represents the stimulus factor coefficient of the desired vehicle speed on the driver. C 2 represents the stimulus factor coefficient of the speed difference between multiple vehicles in front of the driver; Δ x n ( t )= x n+1 (t )- x n ( t )express t The distance Δ between the target manually driven vehicle n and the preceding vehicle n+1 at any given time. v n ( t )= v n+1 ( t )- v n ( t ) represents the speed difference between the target manually driven vehicle n and the preceding vehicle n+1 at time t; α1 and α2 represent the prediction ability coefficients of the aggressive driver and the conservative driver, respectively; τ is the driver's reaction delay time. r As the forward weighting factor in the memory effect, m For memory time; Calculate the optimal speed function for the vehicle in front and the optimal speed function for the vehicle behind: ; ; In the formula, v max This indicates the maximum speed of the vehicles in the platoon. D max and D min These represent the minimum and maximum spacing between adjacent vehicles in the vehicle queue, respectively.

[0033] In embodiments of the present invention, the future predicted motion state includes the target manually driven vehicle's expected future acceleration, future speed, and future position.

[0034] Specifically, the prediction model utilizes the motion state information of the target vehicle n and its preceding vehicle n+1, preceding vehicle n+2, and following vehicle n-1, collected in real time by the environmental perception module, including: Real-time headway Δ x n ( t ) and Δ x n-1 ( t ); Real-time speed v n ( t ), v n+1 ( t ),and v n+2 ( t ); Real-time speed difference Δ vn ( t ) and Δ v n+1 ( t ).

[0035] Based on this real-time data, the model performs calculations using the following mathematical relationships: 1. The real-time collected Δ x n ( t ) and Δ x n-1 ( t Substitute into the optimal velocity function F f (Δ x n ( t ))and F b (Δ x n-1 ( t The forward and backward expected velocities at the current moment are calculated. 2. Utilizing the real-time speed difference Δ v n ( t ) and Δ v n+1 ( t Combined with the sensitivity coefficient q of multiple preceding vehicles, the speed difference stimulus term is calculated; 3. Based on memory time m Extract from historical datasets t - m Speed ​​of time v n ( t - m ) and spacing Δ x n ( t - m ), Δ x n-1 ( t - m ), used for calculating the memory effect term; 4. Through driving style weighting factors h And the predictive ability coefficients α1 and α2 are used to adjust the expected speed calculation under different driving styles; 5. Finally, by combining all factors, the expected acceleration of the target vehicle at the next moment is calculated. dv n ( t ) / dt .

[0036] In this way, the car-following behavior prediction model makes full use of the multi-vehicle motion state information in the real-time local traffic environment dataset to achieve accurate prediction of the car-following behavior of manually driven vehicles.

[0037] In one specific embodiment, the behavior prediction calculation module first receives a real-time traffic environment dataset and parameter configurations from the model parameter library from the environmental perception module. The raw data enters the data preprocessing and caching unit for data validity verification and format standardization. Simultaneously, this unit maintains a historical data cache, continuously storing historical motion state information for a recent period (e.g., 10 seconds), providing data support for memory effect calculations.

[0038] The first step in the calculation process is to calculate the optimal velocity function. The system calculates two optimal velocity values ​​in parallel: the forward optimal velocity function. F f (Δ x n ( t Based on the real-time distance Δ between the target vehicle and the preceding vehicle n+1 x n ( t Calculate the optimal backward velocity function. F b (Δ x n-1 ( t Based on the real-time distance Δ between the target vehicle and the following vehicle n-1 x n-1 ( t ) Calculation. These two functions use a piecewise calculation method: when the distance is less than the minimum safe distance... D min When the function takes the boundary value 0 or v max When the spacing is within the normal range, the function transitions smoothly according to a cosine curve; when the spacing is greater than the maximum following distance... D max At that point, the function takes another boundary value.

[0039] The system is based on driving style weighting factors. h Multi-timescale prediction calculations are performed. For aggressive driving tendencies (h→1), the model performs forward predictions and calculations. F f (Δ x n ( t+ α1τ)) reflects the forward-looking nature of aggressive drivers; for conservative driving tendencies (h→0), the model performs lag prediction and calculates F f (Δ x n ( t-α2τ)) reflects the caution of a conservative driver. Ultimately, through h The prediction results of the two styles are weighted and fused.

[0040] The memory effect calculation unit extracts historical data from the historical data cache up to time tm (e.g., 2 seconds ago), including historical velocity. v n ( tm ), historical forward spacing Δ x n ( tm ) and historical backward spacing Δ x n-1 ( tm The system calculates the expected speed at a historical moment (by weighting and fusing the optimal forward and backward speeds from the past) and compares it with the actual historical speed to obtain the difference value. This difference value is adjusted by the memory strength coefficient β and added to the overall calculation as a memory effect term, reflecting the driver's behavioral characteristics of adjusting the current operation based on historical experience.

[0041] The multi-vehicle impact calculation unit handles the speed differences between multiple vehicles ahead. The system simultaneously considers the speed difference Δ between the target vehicle and the n+1 vehicles ahead. v n ( t The speed difference Δ between the preceding vehicle n+1 and the vehicle n+2 before it. v n+1 ( t (Through multiple front vehicle sensitivity coefficients) q Weighted fusion of these two speed differences [ q Δ v n ( t ) + (1 -q )Δ v n+1 ( t This reflects the driver's behavioral characteristic of observing vehicles ahead to anticipate risks. The weighted speed difference is calculated using the stimulus factor coefficient. Adjust the intensity of the influence.

[0042] All the aforementioned calculation results are aggregated into a comprehensive acceleration calculation unit. This unit calculates the expected acceleration of the target vehicle based on the differential equation of the car-following behavior prediction model established in this invention. Specifically, the multi-timescale prediction results are weighted and fused with the forward and backward optimal velocities through a rear-view weighting factor p, the current velocity is subtracted, and then multiplied by the stimulus factor coefficient. Adding the memory effect term adjusted by the memory strength coefficient β, and the multi-vehicle influence term, the final output is the expected acceleration of the target vehicle in the next moment. .

[0043] After the calculation is completed, the module outputs the predicted future motion state (including expected acceleration, velocity, and position trajectory), which is sent to the decision-making and planning system through the prediction output interface. Simultaneously, the latest parameters and calculation results used in this prediction are updated in the historical data cache to provide data support for subsequent predictions.

[0044] In one embodiment of the present invention, the decision planning system makes a hybrid vehicle cooperative control decision based on the received future predicted motion state. The objectives of the hybrid vehicle cooperative control decision include at least one of maintaining vehicle speed consistency, maintaining a safe following distance, and suppressing the propagation of traffic flow disturbances. The system also generates longitudinal control commands for controlling the state of its own vehicle based on the hybrid vehicle cooperative control decision data.

[0045] The prediction output interface serves as a bridge connecting the system to other parts of the vehicle. This interface sends the predicted future motion state generated by the behavior prediction calculation module to the CAV's decision-making and planning system in real time in a standardized data format. This system does not directly generate vehicle control commands, but rather provides crucial intelligence input to the higher-level system responsible for decision-making.

[0046] Specifically, the prediction output interface receives the results from the calculation module, converts them into a standard data format conforming to vehicle communication protocols (such as CAN, AutoSAR, or ROS), and sends them to the CAV's decision-making and planning system in real time. Based on this, the decision-making and planning system generates control commands and sends them to the vehicle's drive-by-wire actuators (accelerator, brake, steering), thereby achieving prediction-based cooperative adaptive cruise control.

[0047] The system has online learning capabilities, and the deviation between the prediction results and subsequent actual observations is recorded and stored in the historical driving database. The parameter online learning module periodically adjusts the model parameters and updates the model parameter library based on new driving data using optimization algorithms (such as recursive least squares), thereby continuously optimizing the model's predictive capabilities.

[0048] The present invention discloses a human-driven vehicle behavior prediction system for hybrid vehicle fleet control, deployed on a connected autonomous vehicle. An environmental perception module acquires motion state information of the target human-driven vehicle and surrounding vehicles. A behavior prediction calculation module incorporates a human-driven vehicle following behavior prediction model, which models the following intention of the target human-driven vehicle by coupling rear-view effects, memory effects, multiple preceding vehicle observations, and driving style factors. Based on the vehicle motion state information, the behavior prediction calculation module calculates and outputs the predicted future motion state of the target human-driven vehicle. The prediction output interface sends the predicted future trajectory to the vehicle's decision-making and planning system. This invention solves the control problem caused by the uncertainty of human-driven vehicle behavior in hybrid vehicle fleets. By providing high-precision behavior prediction, it provides a key basis for developing safe and efficient collaborative control strategies for connected autonomous vehicles, significantly improving the overall stability and traffic efficiency of hybrid vehicle fleets.

[0049] Next, with reference to the accompanying drawings, a method for predicting the behavior of manually driven vehicles for hybrid fleet control according to an embodiment of the present invention is described.

[0050] Figure 3 This is a schematic diagram of the process for predicting the behavior of manually driven vehicles for hybrid fleet control according to an embodiment of the present invention.

[0051] like Figure 3 As shown, the method for predicting the behavior of manually driven vehicles for hybrid fleet control is used in the aforementioned system for predicting the behavior of manually driven vehicles for hybrid fleet control. The method includes the following steps: Step S101: Obtain vehicle motion status information of the target manually driven vehicle and its surrounding vehicles within the target monitoring area; Step S102: Based on the vehicle motion state information, calculate the future predicted motion state of the target manually driven vehicle using the car-following behavior prediction model. The car-following behavior prediction model is used to mathematically represent the car-following behavior of the target manually driven vehicle by using the rear-view weight factor, memory strength coefficient, memory time, multi-front vehicle sensitivity coefficient, and driving style weight factor to simulate its decision-making process. Step S103: Generate control commands based on the predicted future motion state.

[0052] It should be noted that the foregoing explanation of the embodiment of the human-driven vehicle behavior prediction system for mixed fleet control also applies to the human-driven vehicle behavior prediction method for mixed fleet control in this embodiment, and will not be repeated here.

[0053] The method for predicting the behavior of manually driven vehicles for hybrid vehicle fleet control proposed in this invention solves the control problem caused by the uncertainty of the behavior of manually driven vehicles in hybrid vehicle fleets. By providing high-precision behavior prediction, it provides a key basis for formulating safe and efficient collaborative control strategies for connected autonomous vehicles, and significantly improves the overall stability and traffic efficiency of hybrid vehicle fleets.

[0054] The present invention also proposes a connected autonomous vehicle that includes the aforementioned human-driven vehicle behavior prediction system for hybrid fleet control.

[0055] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0056] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0057] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

Claims

1. A system for predicting the behavior of manually driven vehicles for hybrid fleet control, characterized in that, Deployed on connected autonomous vehicles, the system includes: The environmental perception module is used to acquire information on the vehicle motion status of the target manually driven vehicle and its surrounding vehicles within the target monitoring area. The behavior prediction calculation module is communicatively connected to the environment perception module. It is used to receive vehicle motion state information acquired by the environment perception module and calculate the future predicted motion state of the target manually driven vehicle using a built-in car-following behavior prediction model. The car-following behavior prediction model is used to mathematically represent the car-following behavior of the target manually driven vehicle through rear-view weight factor, memory strength coefficient, memory time, multi-vehicle sensitivity coefficient and driving style weight factor to simulate its decision-making process. The prediction output interface is communicatively connected to the behavior prediction calculation module and is used to send the predicted future motion state to the vehicle's decision-making and planning system to generate control commands. In the car-following behavior prediction model, the behavioral characteristics of the target human-driven vehicle are quantified using the following parameters: The rearview weighting factor p is used to adjust the intensity of the influence of the distance between the vehicle in front and behind and the vehicle on the behavior of the target manually driven vehicle. Memory strength coefficient and memory time This is used to jointly regulate the intensity of the influence of historical driving information on the behavior of target manually driven vehicles; Multiple front vehicle sensitivity coefficient This is used to adjust the intensity of the influence of the motion state information of the preceding vehicle and the vehicle before that on the behavior of the target manually driven vehicle. Driving style weighting factor It is used to quantify the individual characteristics of drivers of target manually driven vehicles and distinguish driving style preferences; The car-following behavior prediction model is as follows: In the formula, For the target, manually driven vehicle Real-time speed, and Let represent the optimal speed functions for the driver's observation of vehicles in front and behind, respectively. The coefficient of the stimulus factor for the driver's desired vehicle speed. The coefficient of the stimulus factor for the driver is the speed difference between multiple vehicles in front of the driver. express Constant target manual driving vehicle and the car in front The spacing between +1, express Constant target manual driving vehicle and the car in front The speed difference between +1; and These represent the predictive ability coefficients for aggressive and conservative drivers, respectively. For the driver's reaction time; As the forward weighting factor in the memory effect, For memory time.

2. The system according to claim 1, characterized in that, The predicted future motion state includes the expected future acceleration, future speed, and future position of the target manually driven vehicle.

3. The system according to claim 1, characterized in that, Calculate the optimal speed function for the vehicle in front and the optimal speed function for the vehicle behind: In the formula, This indicates the maximum speed of the vehicles in the platoon. and These represent the minimum and maximum spacing between adjacent vehicles in the vehicle queue, respectively.

4. The system according to claim 3, characterized in that, The behavior prediction calculation module also includes a model parameter calibration unit. The model parameter calibration unit takes historical real vehicle trajectory dataset as input and uses optimization algorithms to automatically calibrate the rear view weight factor, memory strength coefficient, multi-vehicle sensitivity coefficient and driving style weight factor, so as to minimize the error between the output of the car-following behavior prediction model and the real driving behavior.

5. The system according to claim 1, characterized in that, The environmental perception module includes a vehicle-to-everything (V2X) communication device, and also includes at least one of an onboard camera, millimeter-wave radar, and lidar; the V2X communication device is used to receive motion status information broadcast by the target manually driven vehicle or vehicles around it.

6. The system according to claim 1, characterized in that, The decision planning system makes a hybrid vehicle cooperative control decision based on the received future predicted motion state. The objectives of the hybrid vehicle cooperative control decision include at least one of the following: maintaining vehicle speed consistency, maintaining a safe following distance, and suppressing the propagation of traffic flow disturbances. Based on the hybrid vehicle cooperative control decision, the system generates longitudinal control commands for controlling the vehicle's state.

7. A method for predicting the behavior of manually driven vehicles for hybrid fleet control, characterized in that, The method for the human-driven vehicle behavior prediction system for hybrid fleet control as described in any one of claims 1-6 includes: Acquire information on the vehicle motion status of the target manually driven vehicle and its surrounding vehicles within the target monitoring area; Based on the vehicle motion state information, the future predicted motion state of the target manually driven vehicle is calculated using a car-following behavior prediction model. The car-following behavior prediction model is used to mathematically represent the car-following behavior of the target manually driven vehicle through rear-view weight factor, memory strength coefficient, memory time, multi-vehicle sensitivity coefficient and driving style weight factor, so as to simulate its decision-making process. Control commands are generated based on the predicted future motion state.

8. A connected autonomous driving vehicle, characterized in that, It includes a human-driven vehicle behavior prediction system for hybrid fleet control as described in any one of claims 1-6.

Images