Vehicle chassis global control method and device, electronic equipment and storage medium

Abstract

The invention provides a vehicle chassis global control method and device, electronic equipment and a storage medium, and belongs to the technical field of vehicle control, in the method, reward functions of a first agent and a second agent adopt dynamic weights, the self-adaptive switching capability of control strategies under complex working conditions can be improved, and the vehicle chassis global control efficiency is improved. After a continuous-discrete mixed action signal is obtained, the continuous-discrete mixed action signal is optimized and adjusted based on a system risk level, so that dynamic response conflicts of chassis multi-subsystem (suspension, steering and braking) control are eliminated, collaborative optimization is realized, road condition information in front of a vehicle is fully considered in the process, and the system risk level is improved. Decision delay can be reduced, the real-time performance of system response can be improved, accurate control of a continuous-discrete mixed action space can be finally achieved, and the algorithm is not limited.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle control, and in particular to a vehicle chassis global control method, device, electronic equipment and storage medium. Background Art

[0002] The current mainstream vehicle chassis control systems have the following key problems: Significant control conflicts: Each subsystem (suspension control subsystem, steering control subsystem, and braking control subsystem) makes independent decisions, which can easily lead to conflicting actions. For example, during braking, suspension adjustments may exacerbate vehicle pitch. Insufficient scenario adaptability: The fixed reward function makes it impossible to dynamically adjust the optimization objective, making it difficult to adaptively switch between emergency obstacle avoidance (safety first) and comfortable cruising (smoothness first). Low utilization of preview information: Traditional solutions fail to fully integrate preview data from cameras, lidar, V2X, and other technologies, causing control responses to lag behind actual road conditions. Algorithm application limitations: The traditional DQN algorithm only supports discrete action spaces and has difficulty in accurately controlling continuous variables such as the braking force distribution ratio.

[0003] In summary, traditional vehicle chassis global control has technical problems such as significant control conflict, insufficient scene adaptability, low utilization of preview information and limited algorithm application. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a vehicle chassis global control method, device, electronic device and storage medium to alleviate the technical problems of traditional vehicle chassis global control process such as significant control conflicts, insufficient scene adaptability, low utilization of preview information and limited algorithm application.

[0005] In a first aspect, an embodiment of the present invention provides a vehicle chassis global control method, comprising: Obtaining road condition information ahead of the vehicle and processing vehicle sensor data through the first agent to obtain vehicle state parameters; The second agent generates a continuous-discrete hybrid action signal according to the front road condition information and the vehicle state parameter, wherein the reward functions of the first agent and the second agent adopt dynamic weights; The system risk level is evaluated according to the vehicle dynamic response parameters, and the continuous-discrete hybrid action signal is optimized and adjusted based on the system risk level, and the optimized continuous-discrete hybrid action signal is output to the vehicle chassis execution system.

[0006] Furthermore, the road condition information ahead of the vehicle is obtained, including: Obtain raw data of the road ahead through on-board sensing devices and vehicle network communication equipment; The raw data of the road ahead from different data sources are fused to obtain road condition information ahead of the vehicle including the curvature radius of a preset distance ahead and the position information of obstacles ahead.

[0007] Furthermore, the vehicle-mounted sensing device includes: a laser radar, and the fusion processing adopts a Kalman filter algorithm.

[0008] Furthermore, the first intelligent agent is a parameter observer based on a recursive neural network, and the vehicle state parameters include at least: tire cornering stiffness and road adhesion coefficient.

[0009] Furthermore, the second agent includes: a suspension control agent and a steering and braking coordination agent; The suspension control agent is used to output a continuous damping coefficient and a discrete suspension height; The steering and braking collaborative intelligent agent is used to output a continuous braking force distribution ratio and a discrete target steering angle compensation amount.

[0010] Furthermore, the suspension control agent and the steering and braking collaborative agent are both implemented using a deep reinforcement learning framework, and their network structure includes a continuous action output module and a discrete action output module.

[0011] Furthermore, the vehicle dynamic response parameters include: roll angle and brake pressure; and evaluating the system risk level based on the vehicle dynamic response parameters includes: The system risk level is evaluated based on the roll angle and the brake pressure.

[0012] In a second aspect, an embodiment of the present invention further provides a vehicle chassis global control device, comprising: an acquisition unit, configured to acquire road condition information ahead of the vehicle and process vehicle sensor data through the first agent to obtain vehicle state parameters; a generating unit configured for a second agent to generate a continuous-discrete hybrid action signal based on the forward road condition information and the vehicle state parameter, wherein the reward functions of the first agent and the second agent adopt dynamic weights; An evaluation and optimization unit is used to evaluate the system risk level according to the vehicle dynamic response parameters, optimize and adjust the continuous-discrete hybrid action signal based on the system risk level, and then output the optimized continuous-discrete hybrid action signal to the vehicle chassis execution system.

[0013] In a third aspect, an embodiment of the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to execute any method described in the first aspect above.

[0015] In an embodiment of the present invention, a method for global control of a vehicle chassis is provided, including: obtaining road condition information ahead of the vehicle, and processing vehicle sensor data through a first intelligent agent to obtain vehicle state parameters; a second intelligent agent generates a continuous-discrete hybrid action signal based on the road condition information ahead and the vehicle state parameters, wherein the reward functions of the first intelligent agent and the second intelligent agent adopt dynamic weights; evaluating the system risk level based on the vehicle dynamic response parameters, and optimizing and adjusting the continuous-discrete hybrid action signal based on the system risk level, and then outputting the optimized continuous-discrete hybrid action signal to the vehicle chassis execution system. From the above description, it can be seen that in the vehicle chassis full-domain control method of the present invention, the reward function of the first intelligent agent and the second intelligent agent adopts dynamic weights, which can improve the adaptive switching ability of the control strategy under complex working conditions. After obtaining the continuous-discrete mixed action signal, the continuous-discrete mixed action signal will be optimized and adjusted based on the system risk level, thereby eliminating the dynamic response conflict of the chassis multi-subsystem (suspension, steering, braking) control and realizing collaborative optimization. In addition, the above process fully considers the road condition information in front of the vehicle, which can reduce decision-making delay and improve the real-time response of the system response. Ultimately, precise control of the continuous-discrete mixed action space can be achieved, and the algorithm is not restricted, alleviating the technical problems of the traditional vehicle chassis full-domain control process such as significant control conflict, insufficient scene adaptability, low utilization of preview information and limited algorithm application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flow chart of a vehicle chassis global control method provided by an embodiment of the present invention; Figure 2 A schematic diagram of the DQN algorithm framework provided by an embodiment of the present invention; Figure 3 A schematic diagram of the process implementation of the vehicle chassis global control method provided by an embodiment of the present invention; Figure 4A schematic diagram of a vehicle chassis global control device provided by an embodiment of the present invention; Figure 5 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] The traditional vehicle chassis global control process has significant control conflicts, insufficient scene adaptability, low preview information utilization, and limited algorithm application.

[0020] Based on this, in the vehicle chassis global control method of the present invention, the reward functions of the first intelligent agent and the second intelligent agent adopt dynamic weights, which can improve the adaptive switching ability of the control strategy under complex working conditions. After obtaining the continuous-discrete mixed action signal, the continuous-discrete mixed action signal will be optimized and adjusted based on the system risk level, thereby eliminating the dynamic response conflict of the chassis multi-subsystem (suspension, steering, braking) control and realizing collaborative optimization. In addition, the above process fully considers the road condition information in front of the vehicle, which can reduce decision-making delay and improve the real-time response of the system response. Ultimately, precise control of the continuous-discrete mixed action space can be achieved, and the algorithm is not restricted.

[0021] To facilitate understanding of this embodiment, a vehicle chassis global control method disclosed in an embodiment of the present invention is first introduced in detail.

[0022] Example 1: According to an embodiment of the present invention, an embodiment of a vehicle chassis global control method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0023] Figure 1 is a flow chart of a vehicle chassis global control method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S102, obtaining road condition information ahead of the vehicle, and processing vehicle sensor data through the first agent to obtain vehicle state parameters; In an embodiment of the present invention, the method is a vehicle chassis global control method based on multi-agent collaborative deep reinforcement learning. It adopts a multi-agent collaborative control framework to solve the dynamic response conflict problem of collaborative control of subsystems such as suspension, steering, and braking, and realizes adaptive optimization control of the chassis system under complex working conditions.

[0024] The above-mentioned road condition information in front of the vehicle is based on preview data and is used to generate continuous-discrete hybrid action signals, which can reduce decision-making delays and improve the real-time response of the system.

[0025] Specifically, the first agent uses an RNN network and vehicle dynamics equations to online identify key vehicle state parameters such as tire cornering stiffness and road adhesion coefficient. During implementation, the system is initialized and parameter configuration is completed through the following steps: The first agent uses a two-layer LSTM network (with 128 and 64 hidden layer neurons, respectively) to construct an RNN model. It inputs real-time signals from wheel speed sensors, steering wheel angle sensors, and body acceleration sensors (sampling frequency 100Hz). It is pre-trained on 100,000 sets of simulation data generated by the vehicle's magic formula dynamics model to achieve online identification of tire cornering stiffness and road adhesion coefficient, with an identification error of ≤5%. The communication layer protocol defines the second agent's action priority signal as a three-level code (EMERGENCY=3, WARNING=2, NORMAL=1), and transmits conflict arbitration instructions via the CAN bus (baud rate 500kbps).

[0026] Step S104: The second agent generates a continuous-discrete hybrid action signal based on the road condition information ahead and the vehicle state parameters, wherein the reward functions of the first agent and the second agent use dynamic weights; Step S106 , evaluating the system risk level according to the vehicle dynamic response parameters, optimizing and adjusting the continuous-discrete hybrid action signal based on the system risk level, and then outputting the optimized continuous-discrete hybrid action signal to the vehicle chassis execution system.

[0027] Specifically, this process can be completed by the conflict arbitration module. The agents (i.e., the second agent) exchange action priorities in real time through the communication layer (i.e., realize the interaction of action priority signals between agents). The conflict arbitration module uses fuzzy logic (i.e., fuzzy rules) to evaluate the system risk level. For example: def risk_evaluation(roll_angle, brake_pressure): if roll_angle>5° and brake_pressure>80MPa: return HIGH_RISK# Trigger steering and braking coordination priority strategy In emergency conditions (e.g., the braking priority (i.e., action priority) is "EMERGENCY"), the suspension control agent (i.e., one of the second agents) is only allowed to fine-tune the damping, and the height adjustment is frozen.

[0028] During implementation, after the communication layer collects the actions of each agent (the suspension control agent and the steering and braking coordination agent), the conflict arbitration module evaluates the system risk level according to the following fuzzy rules: Input variables: roll angle φ (membership function is trapezoidal, with a critical value of 5° corresponding to high risk), brake pressure P (Gaussian function, μ=80MPa, σ=5MPa); Output Decision: When the system risk level is HIGH, a "FREEZE_SUSPENSION" signal is sent to freeze suspension height adjustments, allowing only fine-tuning of the damping coefficient. The final action is transmitted to the actuator (i.e., the vehicle chassis actuation system) via the CAN bus. Braking ESP response delay is ≤30ms, and suspension solenoid valve action delay is ≤20ms.

[0029] The above process of optimizing and adjusting the continuous-discrete mixed action signal based on the system risk level can be adjusted by pre-set rules.

[0030] Figure 2 The reinforcement learning-DQN algorithm framework adopted by the present invention is shown in FIG. Figure 3 The process implementation of the vehicle chassis global control method of the present invention is shown in FIG.

[0031] In an embodiment of the present invention, a method for global control of a vehicle chassis is provided, including: obtaining road condition information ahead of the vehicle, and processing vehicle sensor data through a first intelligent agent to obtain vehicle state parameters; a second intelligent agent generates a continuous-discrete hybrid action signal based on the road condition information ahead and the vehicle state parameters, wherein the reward functions of the first intelligent agent and the second intelligent agent adopt dynamic weights; evaluating the system risk level based on the vehicle dynamic response parameters, and optimizing and adjusting the continuous-discrete hybrid action signal based on the system risk level, and then outputting the optimized continuous-discrete hybrid action signal to the vehicle chassis execution system. From the above description, it can be seen that in the vehicle chassis full-domain control method of the present invention, the reward function of the first intelligent agent and the second intelligent agent adopts dynamic weights, which can improve the adaptive switching ability of the control strategy under complex working conditions. After obtaining the continuous-discrete mixed action signal, the continuous-discrete mixed action signal will be optimized and adjusted based on the system risk level, thereby eliminating the dynamic response conflict of the chassis multi-subsystem (suspension, steering, braking) control and realizing collaborative optimization. In addition, the above process fully considers the road condition information in front of the vehicle, which can reduce decision-making delay and improve the real-time response of the system response. Ultimately, precise control of the continuous-discrete mixed action space can be achieved, and the algorithm is not restricted, alleviating the technical problems of the traditional vehicle chassis full-domain control process such as significant control conflict, insufficient scene adaptability, low utilization of preview information and limited algorithm application.

[0032] The above content briefly introduces the vehicle chassis global control method of the present invention. The specific contents involved are described in detail below.

[0033] In an optional embodiment of the present invention, obtaining road condition information ahead of the vehicle specifically includes the following steps: (1) Obtaining raw data of the road ahead through on-board sensing equipment and vehicle network communication equipment; Specifically, the on-board sensing equipment includes: LiDAR, LiDAR (scanning frequency 10Hz, ranging range 200m), which collects raw data of the road ahead, and V2X equipment (i.e., vehicle-to-vehicle communication equipment with communication delay ≤5ms), which also collects raw data of the road ahead.

[0034] (2) The raw data of the road ahead from different data sources are fused and processed to obtain the road condition information ahead of the vehicle, including the curvature radius of the preset distance ahead and the location information of the obstacles ahead.

[0035] Specifically, the raw data of the road ahead collected by the lidar and the raw data of the road ahead collected by the vehicle network communication equipment are fused through the Kalman filter (state transfer matrix F=1, observation matrix H=1, that is, the Kalman filter algorithm) to generate the curvature radius R of the 50m path ahead (that is, the curvature radius of the preset distance ahead) and the position information (x, y) of the obstacle ahead, and input them into the steering and braking collaborative intelligent agent.

[0036] In an optional embodiment of the present invention, the first intelligent agent is a parameter observer based on a recursive neural network, and the vehicle state parameters include at least: tire cornering stiffness and road adhesion coefficient.

[0037] Specifically, the vehicle sensor data includes wheel speed sensor data, steering wheel angle sensor data, and vehicle body acceleration sensor data. The vehicle sensor data is input into the first intelligent agent, and the tire cornering stiffness and road adhesion coefficient are output.

[0038] In an optional embodiment of the present invention, the second agent includes: a suspension control agent and a steering and braking coordination agent; the suspension control agent and the steering and braking coordination agent are both implemented using a deep reinforcement learning framework, and their network structures include a continuous action output module and a discrete action output module; A suspension control agent that outputs continuous damping coefficients and discrete suspension heights; The steering and braking collaborative intelligent agent is used to output the continuous braking force distribution ratio and discrete target angle compensation.

[0039] Specifically, the suspension control agent outputs the continuous damping coefficient and discrete suspension height based on the vehicle body acceleration sensor data and the road adhesion coefficient; The suspension control agent receives the vertical acceleration signal of the vehicle body (range ±5g, accuracy 0.01g), and outputs the continuous damping coefficient δ (value range 0-1, corresponding to the solenoid valve current 0-2A) and discrete suspension height instructions ("raise", "standard", "lower") through the improved DQN algorithm (i.e., NAF-DQN algorithm). Among them, the DQN network structure (i.e., the network structure of the suspension control agent) is: input layer (state dimension 12) → two fully connected hidden layers (256 neurons each, ReLU activation) → output layer (continuous values ​​are generated by the NAF module, and discrete values ​​are output through the Q value layer). The above-mentioned continuous damping coefficient and discrete suspension height are the action signals. The steering and braking collaborative agent outputs the continuous braking force distribution ratio and discrete target steering angle compensation based on the curvature radius, obstacle coordinates, tire cornering stiffness and road adhesion coefficient.

[0040] Specifically, the discrete target steering angle compensation Δθ (discrete value, 5° step size, range -15° to +15°) is calculated based on the curvature radius R, and the continuous braking force distribution ratio α (continuous value, front wheel distribution coefficient 0.6-0.8) is generated through the NAF-DQN algorithm. The advantage function formula of the NAF network (i.e., NAF module) is: A(s,a;θ)=Q(s,a;θ)-V(s;θ) , where state s includes: preview road condition, vehicle speed, vehicle roll angle, action a is α, θ represents the weight of the network, and the value function V(s;θ) Generated by the target network.

[0041] In an optional embodiment of the present invention, the network structure of the second intelligent agent includes: an input layer, two fully connected hidden layers and an output layer, the output layer includes: a NAF module and a Q-value layer, the NAF module (i.e., a continuous action output module) is used to generate continuous values, and the Q-value layer (discrete action output module) is used to generate discrete values.

[0042] The above-mentioned second intelligent agent includes: a suspension control intelligent agent and a steering and braking coordination intelligent agent, and the structures of both are: an input layer, two fully connected hidden layers and an output layer.

[0043] Specifically, continuous actions (such as the braking force distribution ratio) use the NAF-DQN architecture (using the Normalized Advantage Function (NAF) to improve the DQN algorithm). The network output layer uses the tanh activation function to map α to the interval [0.1, 0.9], corresponding to the front wheel braking force ratio. The loss function is: L(θ)=E s,a,r,s′ [(r+γV(s′)-Q(s,a;θ)) 2 ] , where the discount factor γ = 0.99, the target network parameters are soft-updated every 100 steps (τ = 0.001), and the braking force distribution error after training is ≤ ±2%. Discrete actions (such as suspension mode switching) retain the Q-value output layer of the traditional DQN module, which contains three neurons (corresponding to three height modes). Actions are selected using an ε-greedy strategy (ε initialized to 0.1 and decayed by 0.001 every 1000 steps). The measured mode switching accuracy reaches 99.2%. This process achieves "continuous + discrete" hybrid control.

[0044] In an optional embodiment of the present invention, the reward function includes: ,in, Represents the reward value, 、 、 represents the dynamic weight, represents the roll angle penalty function, represents the vertical displacement smoothness function, represents the braking time optimization function.

[0045] Specifically, represents the roll angle penalty function, when >5°, output -10, otherwise, output ; represents the vertical displacement smoothness function, is the rate of change of the vertical displacement of the vehicle body, and the output is (Threshold 0.5m / s); represents the braking time optimization function, and the output is (The target braking time is 0.8s).

[0046] In an optional embodiment of the present invention, when the vertical acceleration standard deviation σ_z<0.3g, ; When V2X receives the accident signal ahead, ; Specifically, when the vertical acceleration standard deviation σ_z < 0.3g, it is the comfort mode. , ,, , the vehicle body smoothness optimization has the highest priority, which can reduce the vibration acceleration perceived by passengers by 30%; when V2X receives the accident signal ahead, it will enter the accident section warning mode, , , , triggering the brake pre-fill function, which shortens the emergency braking response time by 40ms.

[0047] The above dynamic weights can also be generated through the self-attention mechanism.

[0048] Specifically, a Transformer architecture (6 encoder layers and 8 attention heads) is introduced. LiDAR point cloud features (dimension 512) and driving intention features (accelerator pedal travel, steering wheel angle) are used as query vectors, and dynamic weights are generated through the self-attention mechanism. The attention calculation formula is:

[0049] in, Q is the working condition feature query vector, K 、 V is a key-value pair matrix, d k is the key vector dimension (64).

[0050] Advantages: No need to preset weight rules, can autonomously learn emergency obstacle avoidance ( Improved to 0.7), comfortable cruise ( Increased to 0.7) and other scenarios’ weight distribution; Disadvantages: Requires 100,000 sets of labeled working condition data for training, 15M model parameters, and requires support from an on-board edge computing platform (computing power ≥ 10TOPS).

[0051] In addition, the first agent, suspension control agent, and steering and braking coordination agent can be integrated into the Huawei MDC610 central controller, exchanging data via 10Gbps Ethernet, replacing the distributed ECU architecture. Hardware interface definition: LiDAR interface: 10Gbps Ethernet; Actuator interface: CAN FD.

[0052] Advantages: Hardware costs are reduced by 20% (reducing 3 ECUs), and algorithm upgrade cycles are shortened from 3 months to 1 week; Disadvantages: Requires ASIL-D functional safety certification, and central controller redundancy design increases costs by 15%.

[0053] The method of the present invention can eliminate dynamic response conflicts in the control of multiple chassis subsystems (suspension, steering, and braking) to achieve collaborative optimization; design a dynamic weight mechanism to improve the adaptive switching capability of control strategies under complex working conditions; integrate preview information (i.e., the curvature radius at a preset distance ahead and the coordinates of the obstacle ahead) to reduce decision-making delays and improve the real-time response of the system; and expand the deep reinforcement learning algorithm (by adding the NAF module) to achieve precise control of the continuous-discrete hybrid action space.

[0054] The method of the present invention has the following advantages: (1) Improved collaborative control efficiency: In emergency obstacle avoidance scenarios, the conflict arbitration mechanism reduces the roll angle from 8° to 4.8°, a 40% reduction in roll angle, and shortens the braking distance from 32.5m to 31.6m, a 2.8% reduction in braking distance. (2) Dynamic response speed optimization: When switching working conditions (such as from comfortable cruising to emergency braking), dynamic weight adjustment makes the response time less than 100ms, which is 80% faster than the fixed weight scheme; (3) Breakthrough in the utilization of preview information: After integrating LiDAR and V2X data, decision-making delay is reduced from 50ms to 10ms, achieving real-time response from "road condition change to decision output"; (4) Control accuracy is significantly improved: the continuous action control error is reduced to ±0.15m (braking distance), and the discrete action switching accuracy reaches 99.2%.

[0055] Example 2: An embodiment of the present invention also provides a vehicle chassis global control device, which is mainly used to execute the vehicle chassis global control method provided in Example 1 of the present invention. The vehicle chassis global control device provided in the embodiment of the present invention is specifically introduced below.

[0056] Figure 4 is a schematic diagram of a vehicle chassis global control device according to an embodiment of the present invention. Figure 4 As shown, the device mainly includes: an acquisition unit 10, a generation unit 20, and an evaluation and optimization unit 30, wherein: an acquisition unit, configured to acquire road condition information ahead of the vehicle and process vehicle sensor data through the first agent to obtain vehicle state parameters; a generating unit configured for the second agent to generate a continuous-discrete hybrid action signal based on the road condition information ahead and the vehicle state parameters, wherein the reward functions of the first agent and the second agent adopt dynamic weights; The evaluation and optimization unit is used to evaluate the system risk level according to the vehicle dynamic response parameters, and optimize and adjust the continuous-discrete mixed action signal based on the system risk level, and then output the optimized continuous-discrete mixed action signal to the vehicle chassis execution system.

[0057] In an embodiment of the present invention, a vehicle chassis global control device is provided, including: obtaining road condition information ahead of the vehicle, and processing vehicle sensor data through a first intelligent agent to obtain vehicle state parameters; a second intelligent agent generates a continuous-discrete mixed action signal based on the road condition information ahead and the vehicle state parameters, wherein the reward function of the first intelligent agent and the second intelligent agent adopts dynamic weights; evaluating the system risk level according to the vehicle dynamic response parameters, and optimizing and adjusting the continuous-discrete mixed action signal based on the system risk level, and then outputting the optimized continuous-discrete mixed action signal to the vehicle chassis execution system. From the above description, it can be seen that in the vehicle chassis global control device of the present invention, the reward function of the first intelligent agent and the second intelligent agent adopts dynamic weights, which can improve the adaptive switching ability of the control strategy under complex working conditions. After obtaining the continuous-discrete mixed action signal, the continuous-discrete mixed action signal will be optimized and adjusted based on the system risk level, thereby eliminating the dynamic response conflict of the chassis multi-subsystem (suspension, steering, braking) control and realizing collaborative optimization. In addition, the above process fully considers the road condition information in front of the vehicle, which can reduce decision-making delay and improve the real-time response of the system response. Ultimately, precise control of the continuous-discrete mixed action space can be achieved, and the algorithm is not restricted, alleviating the technical problems of the traditional vehicle chassis global control process such as significant control conflict, insufficient scene adaptability, low utilization of preview information and limited algorithm application.

[0058] Optionally, the acquisition unit is also used to: obtain the original data of the road ahead through the vehicle-mounted sensing device and the Internet of Vehicles communication device; fuse the original data of the road ahead from different data sources to obtain the road condition information ahead of the vehicle including the curvature radius of the preset distance ahead and the position information of the obstacle ahead.

[0059] Optionally, the vehicle-mounted sensing device includes: a laser radar, and the fusion processing adopts a Kalman filter algorithm.

[0060] Optionally, the first intelligent agent is a parameter observer based on a recursive neural network, and the vehicle state parameters include at least: tire cornering stiffness and road adhesion coefficient.

[0061] Optionally, the second intelligent agent includes: a suspension control intelligent agent and a steering and braking coordination intelligent agent; the suspension control intelligent agent is used to output a continuous damping coefficient and a discrete suspension height; the steering and braking coordination intelligent agent is used to output a continuous braking force distribution ratio and a discrete target angle compensation amount.

[0062] Optionally, the suspension control agent and the steering and braking collaborative agent are both implemented using a deep reinforcement learning framework, and their network structure includes a continuous action output module and a discrete action output module.

[0063] Optionally, the vehicle dynamic response parameters include: roll angle and brake pressure; the evaluation and optimization unit is further used to: evaluate the system risk level according to the roll angle and brake pressure.

[0064] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those in the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference can be made to the corresponding content in the aforementioned method embodiment.

[0065] like Figure 5 As shown, an electronic device 600 provided in an embodiment of the present application includes: a processor 601, a memory 602 and a bus, wherein the memory 602 stores machine-readable instructions executable by the processor 601. When the electronic device is running, the processor 601 communicates with the memory 602 through the bus, and the processor 601 executes the machine-readable instructions to perform the steps of the above-mentioned vehicle chassis global control method.

[0066] Specifically, the above-mentioned memory 602 and processor 601 can be general-purpose memory and processor, which are not specifically limited here. When the processor 601 runs the computer program stored in the memory 602, it can execute the above-mentioned vehicle chassis global control method.

[0067] The processor 601 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 601 or by instructions in the form of software. The above-mentioned processor 601 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in memory 602, and processor 601 reads the information in memory 602 and performs the steps of the above method in conjunction with its hardware.

[0068] Corresponding to the above-mentioned vehicle chassis global control method, an embodiment of the present application also provides a computer-readable storage medium, which stores machine-executable instructions. When the computer-executable instructions are called and executed by the processor, the computer-executable instructions prompt the processor to execute the steps of the above-mentioned vehicle chassis global control method.

[0069] The vehicle chassis global control device provided in the embodiment of the present application can be specific hardware on the device or software or firmware installed on the device. The device provided in the embodiment of the present application, its implementation principle and the technical effect produced are the same as those in the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference can be made to the corresponding content in the aforementioned method embodiment. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.

[0070] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0071] For another example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0072] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0073] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0074] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the vehicle chassis global control method described in each embodiment of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program codes.

[0075] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0076] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. However, these modifications, changes, or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application. They should all be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A vehicle chassis global control method, characterized in that: include: Obtaining road condition information ahead of the vehicle and processing vehicle sensor data through the first agent to obtain vehicle state parameters; The second agent generates a continuous-discrete hybrid action signal according to the front road condition information and the vehicle state parameter, wherein the reward functions of the first agent and the second agent adopt dynamic weights; The system risk level is evaluated according to the vehicle dynamic response parameters, and the continuous-discrete hybrid action signal is optimized and adjusted based on the system risk level, and the optimized continuous-discrete hybrid action signal is output to the vehicle chassis execution system.

2. The method according to claim 1, characterized in that Obtain information about the road conditions ahead of the vehicle, including: Obtain raw data of the road ahead through on-board sensing devices and vehicle network communication equipment; The raw data of the road ahead from different data sources are fused to obtain road condition information ahead of the vehicle including the curvature radius of a preset distance ahead and the position information of obstacles ahead.

3. The method according to claim 2, characterized in that The vehicle-mounted sensing device includes: a laser radar, and the fusion processing adopts a Kalman filter algorithm.

4. The method according to claim 1, wherein The first intelligent agent is a parameter observer based on a recursive neural network, and the vehicle state parameters include at least: tire cornering stiffness and road adhesion coefficient.

5. The method according to claim 1, wherein The second intelligent agent includes: a suspension control intelligent agent and a steering and braking coordination intelligent agent; The suspension control agent is used to output a continuous damping coefficient and a discrete suspension height; The steering and braking collaborative intelligent agent is used to output a continuous braking force distribution ratio and a discrete target steering angle compensation amount.

6. The method according to claim 5, characterized in that The suspension control agent and the steering and braking collaborative agent are both implemented using a deep reinforcement learning framework, and their network structure includes a continuous action output module and a discrete action output module.

7. The method according to claim 1, characterized in that The vehicle dynamic response parameters include: roll angle and brake pressure; evaluating the system risk level based on the vehicle dynamic response parameters includes: The system risk level is evaluated based on the roll angle and the brake pressure.

8. A vehicle chassis global control device, characterized in that: include: an acquisition unit, configured to acquire road condition information ahead of the vehicle and process vehicle sensor data through the first agent to obtain vehicle state parameters; a generating unit configured for a second agent to generate a continuous-discrete hybrid action signal based on the forward road condition information and the vehicle state parameter, wherein the reward functions of the first agent and the second agent adopt dynamic weights; An evaluation and optimization unit is used to evaluate the system risk level according to the vehicle dynamic response parameters, optimize and adjust the continuous-discrete hybrid action signal based on the system risk level, and then output the optimized continuous-discrete hybrid action signal to the vehicle chassis execution system.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to execute the method according to any one of claims 1 to 7.

Images