A chassis system scene self-adaptive virtual calibration method based on reinforcement fine-tuning
By constructing a cross-vehicle PID parameter mapping generation model and combining it with an enhanced fine-tuning optimization mechanism, the problems of system coordination and scenario adaptation in chassis system calibration were solved, achieving efficient and accurate virtual calibration of the chassis electronic control system and reducing vehicle calibration costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2026-03-30
- Publication Date
- 2026-07-14
Smart Images
Figure CN122386650A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of vehicle chassis electronic control system calibration technology, specifically a scenario-adaptive virtual calibration method for chassis systems based on enhanced fine-tuning. Background Technology
[0002] The automotive chassis control system consists of multiple electronic control modules, including electronic stability control systems and anti-lock braking systems. These systems typically employ a proportional-integral-derivative (PID) control structure. Currently, chassis system calibration often uses individual system-specific methods. This involves pre-setting calibration conditions and manually adjusting system parameters based on the system's performance under those conditions. This approach requires extensive real-vehicle testing, making it unsuitable for the fast-paced, simulation-first principles of current vehicle development. Furthermore, it lacks consideration for the collaborative working processes of different systems, leading to antagonistic effects in parameter design when different systems work together. Currently, virtual calibration methods, using highly realistic vehicle kinematic environments built within simulation environments, have become the core approach for automotive chassis system calibration. Examples include a Bayesian optimization-based virtual acceleration calibration method for chassis electronic control system parameters (CN202411147630.1) and a virtual calibration method, device, equipment, and medium based on a driving simulator (CN202511574170.5). However, existing virtual calibration methods only consider the parameter characteristics of the current system to be calibrated, lacking calibration experience for other vehicle models; at the same time, they lack consideration of the parameter adjustment process of proportional-integral-derivative control structures in different scenarios. Summary of the Invention
[0003] To address the aforementioned issues, this invention provides a scenario-adaptive virtual calibration method for chassis systems based on enhanced fine-tuning. By constructing a cross-vehicle PID parameter mapping generation model and combining it with an enhanced fine-tuning optimization mechanism, the method achieves automatic optimization of control parameters for multiple chassis systems in a vehicle dynamics simulation environment, thereby improving overall vehicle dynamic performance and reducing vehicle calibration costs.
[0004] The technical solution of this invention is described below in conjunction with the accompanying drawings:
[0005] This invention provides a scenario-adaptive virtual calibration method for chassis systems based on enhanced fine-tuning, comprising the following steps:
[0006] Step 1: Construct the PID control system mapping parameter space for the chassis control system;
[0007] Step 2: Construct structural feature models for different vehicle types;
[0008] Step 3: Establish a driving scenario parameter model;
[0009] Step 4: Establish a driving scenario zoning method;
[0010] Step 5: Establish a basic PID system parameter mapping generation model;
[0011] Step 6: Perform cross-vehicle pre-training of the PID system parameter generation model;
[0012] Step 7: Enhance and fine-tune the PID system parameter generation model of the chassis system to be calibrated;
[0013] Step 8: Perform vehicle dynamics simulation evaluation of PID system parameters;
[0014] Step 9: Perform scenario risk weight-driven optimization;
[0015] Step 10: Output the PID mapping table of the chassis system to be calibrated.
[0016] Furthermore, the specific method for step one is as follows:
[0017] 11) Identify the chassis control systems that require collaborative calibration and establish a control parameter space;
[0018] The vehicle chassis system includes an electric power steering system, an electronic stability control system, an anti-lock braking system (ABS), and an active suspension system; the electronic stability control system uses a PID control structure, and the control law is expressed as:
[0019]
[0020] In the formula, K p For proportional gain; K i K is the integral gain; d This is the differential gain; To account for the system's control error, a PID gain mapping table is used to represent the control parameters under different vehicle speed conditions:
[0021]
[0022] In the formula, v is the vehicle speed; , , These are the mapping functions of proportional gain, integral gain, and differential gain with respect to vehicle speed, respectively.
[0023] 12) By selecting multiple discrete vehicle speeds v1, v2, ..., v m Construct a PID gain mapping table for the control system:
[0024]
[0025] For multiple chassis control systems, a joint parameter space for the control systems across these systems is constructed:
[0026]
[0027] In the formula, , , , These are the PID gain mapping parameter spaces for the electric power steering system, electronic stability control system, anti-lock braking system, and active suspension system, respectively.
[0028] Furthermore, the specific method for step two is as follows:
[0029] 21) Construct vehicle feature vectors by extracting key vehicle structural parameters, and use these vectors as inputs for generating subsequent control system parameters; encode vehicle mass, wheelbase, yaw inertia, suspension stiffness, and tire stiffness into vehicle feature vectors:
[0030]
[0031] In the formula, m is the total vehicle mass; This refers to the vehicle's wheelbase. The yaw inertia of the vehicle; For suspension stiffness; k t For tire stiffness;
[0032] 22) Normalize vehicle features:
[0033] .
[0034] Furthermore, the specific method for step three is as follows:
[0035] 31) Describe the vehicle's operating environment by defining a driving scenario parameter vector:
[0036]
[0037] In the formula, v is the vehicle speed; μ is the road surface adhesion coefficient; r is the road curvature; and h is the driving operation input.
[0038] Furthermore, the specific method for step four is as follows:
[0039] 41) Divide the driving scenario space, further subdividing the scenario based on two driving scenario parameters: vehicle speed and road surface adhesion coefficient.
[0040]
[0041] In the formula, C k This represents the k-th scene region.
[0042] Furthermore, the specific method for step five is as follows:
[0043] 51) Based on vehicle feature modeling and scenario modeling, construct the parameter mapping for the basic PID control system:
[0044]
[0045] In the formula, V' is the vehicle feature vector; S is the scene feature parameter; and ϕ0 is the model parameter.
[0046] Furthermore, the specific method for step six is as follows:
[0047] 61) Construct a historical calibration dataset based on all historical calibration data:
[0048]
[0049] In the formula, 𝐷 represents the historical calibration dataset; 𝑉 𝑖 Let be the vehicle structure feature vector of the nth vehicle sample; 𝑗 Let Θ be the parameter vector for the nth driving scenario; 𝑖𝑗 For vehicle 𝑉 𝑖 In driving scenarios 𝑗 The corresponding actual PID mapping parameters;
[0050] 62) Through supervised learning, based on a historical calibration dataset, the PID mapping parameters are trained using a neural network architecture supervised learning method; the loss function is defined as:
[0051]
[0052] In the formula, 𝐿 is the model training loss function; Θ 𝑝𝑟𝑒d The PID mapping parameters predicted by the model; Θ 𝑡𝑟𝑢𝑒 These are the PID mapping parameters obtained from actual calibration; ||⋅|| is the vector L2 norm;
[0053] 63) Update model parameters using gradient descent:
[0054]
[0055] In the formula, 𝜙 𝑡 Here are the model parameters at the nth iteration; 𝜙 𝑡+1 represents the updated model parameters; is the learning rate, used to control the step size of parameter updates; ∇ is the gradient of the loss function with respect to the model parameters.
[0056] Furthermore, the specific method for step seven is as follows:
[0057] 71) Based on the basic PID mapping parameter Θ base (v i Define the fine-tuning parameter ΔΘ(v) i This allows for the updating of PID parameters.
[0058]
[0059] 72) Constrain the fine-tuning parameters:
[0060]
[0061] In the formula, δ is the pre-set threshold for the change of the fine-tuning parameter;
[0062] 73) Use reinforcement learning methods to update PID parameters.
[0063] Furthermore, the specific method for step eight is as follows:
[0064] 81) Input the updated PID mapping parameters into the vehicle dynamics simulation model and calculate the vehicle dynamic response index X:
[0065]
[0066] In the formula, 𝑒 𝑦𝑎𝑤 yaw rate error; φ is the sideslip angle; φ is the body roll angle; d 𝑏𝑟𝑎𝑘𝑒 Braking distance;
[0067] 82) Introduce stability constraints:
[0068]
[0069] In the formula, J total For the overall performance indicators of the vehicle chassis system; J EPS For the performance indicators of electric power steering systems; J ESP For the performance indicators of the electronic stability control system; J ABS For the performance indicators of the anti-lock braking system, J Susp For the performance indicators of the active suspension system;
[0070] 83) The stability constraint is:
[0071]
[0072] In the formula, ϵ is the system stability threshold.
[0073] Furthermore, the specific method for step nine is as follows:
[0074] 91) Set reward weights according to different driving scenarios:
[0075]
[0076] In the formula, , , , These are the weighted indicators for yaw stability, roll stability, comfort, and braking performance, respectively.
[0077] 92) The reward for reinforcement learning is defined as:
[0078]
[0079] 93) Update the fine-tuning parameters of the reinforcement learning algorithm according to the reward function:
[0080]
[0081] In the formula, This represents the fine-tuning amount of the PID parameters at the t-th iteration; This is the gradient of the reward function with respect to the PID parameters.
[0082] Furthermore, the specific method for step ten is as follows:
[0083] 101) The final generated PID mapping parameters are:
[0084] .
[0085] The beneficial effects of this invention are as follows:
[0086] This invention provides a virtual calibration method for chassis electronic control systems that is cross-vehicle, multi-system collaborative, and scenario-adaptive. The cross-vehicle capability leverages experience from calibration data of other models; the multi-system collaboration allows for simultaneous calibration of different chassis subsystems to improve system performance; and the scenario-adaptive capability ensures that the calibrated parameter mappings can automatically switch parameters according to driving conditions. These advantages maximize the efficiency and accuracy of the calibration process while reducing calibration costs. Attached Figure Description
[0087] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0088] Figure 1 This is a schematic diagram of the overall process of the present invention;
[0089] Figure 2This is a schematic diagram illustrating the PID parameter calibration process for a specific electronic control system according to the present invention;
[0090] Figure 3 A schematic diagram illustrating the braking pressure control before calibrating a chassis electronic control system.
[0091] Figure 4 A schematic diagram showing the brake pressure control after calibration of a chassis electronic control system. Detailed Implementation
[0092] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.
[0093] Example 1
[0094] See Figure 1 and Figure 2 This invention provides a scenario-adaptive virtual calibration method for chassis systems based on enhanced fine-tuning, comprising the following steps:
[0095] Step 1: Construct the mapping parameter space for the PID control system of the chassis control system. The specific method is as follows:
[0096] 11) Identify the chassis control systems that require collaborative calibration and establish a control parameter space; the vehicle chassis system includes the electric power steering system, electronic stability control system, anti-lock braking system, and active suspension system; the electronic stability control system adopts a PID control structure, and the control law is expressed as:
[0097]
[0098] In the formula, K p For proportional gain; K i K is the integral gain; d This is the differential gain; This refers to the system's control error.
[0099] In practical vehicle controllers, PID parameters typically vary with the vehicle's operating state. A PID gain mapping table is used to represent the control parameters under different vehicle speed conditions.
[0100]
[0101] In the formula, v is the vehicle speed; , , These are the mapping functions of proportional gain, integral gain, and differential gain with respect to vehicle speed, respectively.
[0102] 12) By selecting multiple discrete vehicle speeds v1, v2, ..., v m Construct a PID gain mapping table for the control system:
[0103]
[0104] For multiple chassis control systems, a joint parameter space for the control systems across these systems is constructed:
[0105]
[0106] In the formula, , , , These are the PID gain mapping parameter spaces for the electric power steering system, electronic stability control system, anti-lock braking system, and active suspension system, respectively.
[0107] Step 2: Construct vehicle structural feature models for different vehicle types. The specific method is as follows:
[0108] 21) In order to enable the parameter generation system to fully utilize calibration knowledge across vehicle models and achieve calibration knowledge transfer, this step extracts key structural parameters of the vehicle to construct a vehicle feature vector, which is then used as the input for subsequent control system parameter generation.
[0109] Vehicle dynamics are mainly affected by factors such as vehicle mass, wheelbase, yaw inertia, suspension stiffness, and tire stiffness. These factors are encoded into a vehicle feature vector:
[0110]
[0111] In the formula, m is the vehicle mass; L is the vehicle wheelbase; k is the yaw inertia of the vehicle. s For suspension stiffness; k t For tire stiffness;
[0112] 22) To eliminate the differences in the dimensions of parameters of different vehicles, the vehicle characteristics are normalized:
[0113] .
[0114] Step 3: Establish a driving scenario parameter model. The specific method is as follows:
[0115] 31) Vehicles encounter various driving scenarios during actual operation, such as low-speed urban roads, highways, slippery surfaces, and emergency braking. These different scenarios significantly impact the performance and corresponding parameters of the chassis control system. This invention describes the vehicle's operating environment by defining a driving scenario parameter vector:
[0116]
[0117] In the formula, v is the vehicle speed; μ is the road surface adhesion coefficient; r is the road curvature; and h is the driving operation input.
[0118] Step 4: Establish a driving scenario zoning method, the specific method is as follows:
[0119] 41) To improve the targeting of control parameter optimization, this invention divides the driving scenario space, further partitioning the scenario based on two driving scenario parameters: vehicle speed and road surface adhesion coefficient.
[0120]
[0121] In the formula, C k This represents the k-th scene region.
[0122] The distinction results are shown in Table 1.
[0123] Table 1 Scene partitions and corresponding conditions
[0124] Low speed v < 30 km / h Medium speed 30 km / h < v < 80 km / h High speed v > 80 km / h Low μ ≤ 0.4 High μ > 0.4 Figure 3 Figure 4
[0125] Step 5: Establish a basic PID system parameter mapping generation model. The specific method is as follows:
[0126] 51) Based on vehicle feature modeling and scenario modeling, construct the parameter mapping for the basic PID control system:
[0127]
[0128] In the formula, V' is the vehicle feature vector; S is the scene feature parameter; and ϕ0 is the model parameter.
[0129] Step 6: Perform cross-vehicle pre-training of the PID system parameter generation model. The specific method is as follows:
[0130] 61) Construct a historical calibration dataset based on all historical calibration data:
[0131]
[0132] In the formula, 𝐷 represents the historical calibration dataset; Let be the vehicle structure feature vector of the nth vehicle sample; Let be the parameter vector for the nth driving scenario; For vehicle 𝑉 𝑖 In driving scenarios 𝑗 The corresponding actual PID mapping parameters;
[0133] 62) Through supervised learning, based on a historical calibration dataset, the PID mapping parameters are trained using a neural network architecture supervised learning method; the loss function is defined as:
[0134]
[0135] In the formula, θ is the model training loss function, used to measure the error between the predicted parameters and the true parameters; 𝑝𝑟𝑒d The PID mapping parameters predicted by the model; Θ 𝑡𝑟𝑢𝑒 These are the PID mapping parameters obtained from actual calibration; ||⋅|| is the vector L2 norm;
[0136] 63) Update model parameters using gradient descent:
[0137]
[0138] In the formula, 𝜙 𝑡 Here are the model parameters at the nth iteration; 𝜙 𝑡+1 represents the updated model parameters; is the learning rate, used to control the step size of parameter updates; ∇ is the gradient of the loss function with respect to the model parameters.
[0139] Step 7: Perform enhancement, fine-tuning, and optimization of the PID system parameter generation model for the chassis system to be calibrated. The specific method is as follows:
[0140] 71) Based on the basic PID mapping parameters Θ obtained in step six base (v i Define the fine-tuning parameter ΔΘ(v) i This allows for the updating of PID parameters.
[0141]
[0142] 72) To avoid instability in the generation of system parameters, constraints are imposed on the fine-tuning parameters:
[0143]
[0144] In the formula, δ is the pre-set threshold for the change of the fine-tuning parameter;
[0145] 73) Use reinforcement learning methods to update PID parameters.
[0146] Step 8: Perform vehicle dynamics simulation evaluation of PID system parameters. The specific method is as follows:
[0147] 81) Input the updated PID mapping parameters into the vehicle dynamics simulation model, calculate the vehicle dynamic response index X, and use it to evaluate vehicle stability and safety.
[0148]
[0149] In the formula, 𝑒 𝑦𝑎𝑤 yaw rate error; φ is the sideslip angle; φ is the body roll angle; d 𝑏𝑟𝑎𝑘𝑒 Braking distance;
[0150] 82) Simultaneously, due to the coupling relationship between the chassis and plate systems, stability constraints need to be introduced:
[0151]
[0152] In the formula, J total For the overall performance indicators of the vehicle chassis system; J EPS For the performance indicators of electric power steering systems; J ESP For the performance indicators of the electronic stability control system; J ABS For the performance indicators of the anti-lock braking system, J Susp For the performance indicators of the active suspension system;
[0153] 83) The stability constraint is:
[0154]
[0155] In the formula, ϵ is the system stability threshold.
[0156] Step 9: Perform scenario risk weight-driven optimization, the specific method is as follows:
[0157] 91) Set reward weights according to different driving scenarios:
[0158]
[0159] In the formula, , , , These are the weighted indicators for yaw stability, roll stability, comfort, and braking performance, respectively.
[0160] 92) Define the reward for reinforcement learning in step seven as:
[0161]
[0162] 93) Update the fine-tuning parameters of the reinforcement learning algorithm according to the reward function:
[0163]
[0164] In the formula, This represents the fine-tuning amount of the PID parameters at the t-th iteration; This is the gradient of the reward function with respect to the PID parameters.
[0165] Step 10: Output the PID mapping table for the chassis system to be calibrated. The specific method is as follows:
[0166] The final generated PID mapping parameters are:
[0167]
[0168] This mapping parameter can automatically generate optimal control parameters based on vehicle structural parameters and driving scenarios, thereby enabling adaptive virtual calibration of the chassis system scenario.
[0169] Example 2
[0170] See Figure 3 and Figure 4 , and The front and rear brake pressure control of a certain chassis electronic control system is calibrated; in the figure, L1, L2, R1, and R2 are the slip ratio curves of the left front, left rear, right front, and right rear wheels, respectively; pMC is the main cylinder pressure signal.
[0171] As shown in the figure, before calibration, the slip ratio of the four wheels fluctuated wildly and was difficult to stabilize at around 1.5, making it impossible to fully utilize the ground braking force. After calibration, the slip ratio of the four wheels was stable and between 0.5 and 2, which allowed for full utilization of the ground braking force and achieved better control.
[0172] In summary, by constructing a cross-vehicle PID parameter mapping generation model and combining it with an enhanced fine-tuning optimization mechanism, automatic optimization of chassis multi-system control parameters can be achieved in a vehicle dynamics simulation environment, thereby improving the overall vehicle dynamic performance and reducing vehicle calibration costs.
[0173] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A scenario-adaptive virtual calibration method for chassis systems based on enhanced fine-tuning, characterized in that, Includes the following steps: Step 1: Construct the PID control system mapping parameter space for the chassis control system; Step 2: Construct structural feature models for different vehicle types; Step 3: Establish a driving scenario parameter model; Step 4: Establish a driving scenario zoning method; Step 5: Establish a basic PID system parameter mapping generation model; Step 6: Perform cross-vehicle pre-training of the PID system parameter generation model; Step 7: Perform enhancement, fine-tuning, and optimization of the PID system parameter generation model for the chassis system to be calibrated; Step 8: Perform vehicle dynamics simulation evaluation of PID system parameters; Step 9: Perform scenario risk weight-driven optimization; Step 10: Output the PID mapping table of the chassis system to be calibrated.
2. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step one is as follows: 11) Identify the chassis control systems that require collaborative calibration and establish a control parameter space; The vehicle chassis system includes an electric power steering system, an electronic stability control system, an anti-lock braking system, and an active suspension system; the chassis electronic control system mostly adopts a PID control structure, and the control law is expressed as: ； In the formula, K p For proportional gain; K i K is the integral gain; d This is the differential gain; To account for the system's control error, a PID gain mapping table is used to represent the control parameters under different vehicle speed conditions: ； In the formula, v is the vehicle speed; , , These are the mapping functions of proportional gain, integral gain, and differential gain with respect to vehicle speed, respectively. 12) By selecting multiple discrete vehicle speeds v1, v2, ..., v m Construct a PID gain mapping table for the control system: ； For multiple chassis control systems, a joint parameter space for the control systems across these systems is constructed: ； In the formula, , , , These are the PID gain mapping parameter spaces for the electric power steering system, electronic stability control system, anti-lock braking system, and active suspension system, respectively.
3. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step two is as follows: 21) Construct vehicle feature vectors by extracting key structural parameters of the vehicle, and use them as inputs for the subsequent generation of control system parameters; encode the vehicle mass, wheelbase, yaw inertia, suspension stiffness, and tire stiffness into vehicle feature vectors: ； In the formula, m is the vehicle mass; L is the vehicle wheelbase; k is the yaw inertia of the vehicle. s For suspension stiffness; k t For tire stiffness; 22) Normalize vehicle features: 。 4. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step three is as follows: 31) Describe the vehicle's operating environment by defining a driving scenario parameter vector: ； In the formula, v is the vehicle speed; μ is the road surface adhesion coefficient; r is the road curvature; and h is the driving operation input.
5. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step four is as follows: 41) Divide the driving scenario space, further subdividing the scenario based on two driving scenario parameters: vehicle speed and road surface adhesion coefficient. ； In the formula, C k This represents the k-th scene region.
6. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step five is as follows: 51) Based on vehicle feature modeling and scenario modeling, construct the parameter mapping for the basic PID control system: ； In the formula, V' is the vehicle feature vector; S is the scene feature parameter; ϕ0 represents the model parameters.
7. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step six is as follows: 61) Construct a historical calibration dataset based on all historical calibration data: ； In the formula, 𝐷 represents the historical calibration dataset; 𝑉 𝑖 Let be the vehicle structure feature vector of the nth vehicle sample; 𝑗 Let Θ be the parameter vector for the nth driving scenario; 𝑖𝑗 For vehicle 𝑉 𝑖 In driving scenarios 𝑗 The corresponding actual PID mapping parameters; 62) Through supervised learning, based on historical calibration datasets, the PID mapping parameters are trained using a neural network architecture supervised learning method; The loss function is defined as: ； In the formula, 𝐿 is the model training loss function; Θ 𝑝𝑟𝑒d The PID mapping parameters predicted by the model; Θ 𝑡𝑟𝑢𝑒 These are the PID mapping parameters obtained from actual calibration; ||⋅|| is the vector L2 norm; 63) Update model parameters using gradient descent: ； In the formula, 𝜙 𝑡 Here are the model parameters at the nth iteration; 𝜙 𝑡+1 represents the updated model parameters; is the learning rate, used to control the step size of parameter updates; ∇ is the gradient of the loss function with respect to the model parameters.
8. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 7, characterized in that, The specific method for step seven is as follows: 71) Based on the basic PID mapping parameter Θ base (v i Define the fine-tuning parameter ΔΘ(v) i This allows for the updating of PID parameters. ； 72) Constrain the fine-tuning parameters: ； In the formula, δ is the pre-set threshold for the change of the fine-tuning parameter; 73) Use reinforcement learning methods to update PID parameters.
9. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 1, characterized in that, The specific method for step eight is as follows: 81) Input the updated PID mapping parameters into the vehicle dynamics simulation model and calculate the vehicle dynamic response index X: ； In the formula, 𝑒 𝑦𝑎𝑤 yaw rate error; φ is the sideslip angle; φ is the body roll angle; d 𝑏𝑟𝑎𝑘𝑒 Braking distance; 82) Introduce stability constraints: ； In the formula, J total For the overall performance indicators of the vehicle chassis system; J EPS For the performance indicators of electric power steering systems; J ESP For the performance indicators of the electronic stability control system; J ABS For the performance indicators of the anti-lock braking system, J Susp For the performance indicators of the active suspension system; 83) The stability constraint is: ； In the formula, ϵ is the system stability threshold.
10. The chassis system scene adaptive virtual calibration method based on enhanced fine-tuning according to claim 9, characterized in that, The specific method for step nine is as follows: 91) Set reward weights according to different driving scenarios: ； In the formula, , , , These are the weighted indicators for yaw stability, roll stability, comfort, and braking performance, respectively. 92) The reward for reinforcement learning is defined as: ； 93) Update the fine-tuning parameters of the reinforcement learning algorithm according to the reward function: ； In the formula, This represents the fine-tuning amount of the PID parameters at the t-th iteration; This represents the gradient of the reward function with respect to the PID parameters. The specific method for step ten is as follows: 101) The final generated PID mapping parameters are: 。