A method and system for reverse optimization of control parameters for electric passenger vehicles

Abstract

This invention relates to the field of motion sickness assessment technology for electric passenger vehicles, specifically to a method and system for reverse optimization of control parameters for electric passenger vehicles. It includes: a forward motion sickness evaluation and attribution module for calculating the motion sickness comfort index and locating motion sickness-inducing factors; an optimization target definition module for setting optimization targets and performance constraints based on diagnostic results; a vehicle dynamics simulation model module for constructing a parameterized vehicle model and conducting virtual testing; a reverse parameter optimization engine for performing reverse optimization of vehicle control parameters based on quantified optimization targets, performance constraints, and the vehicle dynamics simulation model; and an optimization result output and verification module for writing candidate control parameter sets to the vehicle controller and forming a closed-loop iteration through real-vehicle data acquisition and comparative verification. This technical solution can solve problems such as the disconnect between existing motion sickness assessment technologies and vehicle control execution, reliance on manual optimization processes, low efficiency, and lack of closed-loop verification.

Description

Technical Field

[0001] This invention relates to the field of motion sickness assessment technology for electric passenger vehicles, specifically to a method and system for reverse optimization of control parameters for electric passenger vehicles. Background Technology

[0002] With the rapid growth in the number of electric passenger vehicles, consumers are increasingly demanding higher levels of ride comfort. Compared to gasoline-powered vehicles, electric vehicles offer advantages such as faster torque response and greater energy recovery, which improves performance. However, this also leads to a significant increase in motion sickness due to issues like excessive jerk rate of acceleration and drastic pitch rate. Motion sickness comfort has become one of the key indicators affecting the user experience and market competitiveness of electric passenger vehicles.

[0003] In the quantitative evaluation of motion sickness comfort, the industry has developed various technical approaches. Early studies mostly assessed the human body's perception level of motion stimuli through frequency domain analysis of vehicle vibration signals. In recent years, with the development of machine learning technology, some studies have attempted to construct motion sickness prediction models based on vehicle motion parameters (such as longitudinal acceleration, lateral angular velocity, and vertical acceleration). For example, Chinese patent CN121740455A discloses an evaluation model that can infer the risk of motion sickness from vehicle motion parameters. Such models have made some progress at the perception and diagnosis levels, and can quantify motion sickness risk and locate motion sickness-causing factors to a certain extent.

[0004] In terms of control parameter optimization, traditional vehicle development processes primarily rely on engineers' subjective evaluations and experience-based tuning to improve motion sickness comfort. This involves manually adjusting control parameters such as throttle mapping curves, shock absorber damping curves, and torque response time constants to enhance the ride experience. Some research has also attempted to introduce intelligent optimization algorithms such as genetic algorithms and particle swarm optimization to perform offline optimization of control parameters based on simulation models. However, these optimization methods typically rely on idealized single-system simulation models for calculations, and the optimization process is often open-loop, rarely considering the coupling relationships between the powertrain, chassis, and suspension systems. Consequently, there is generally a certain deviation between the optimization results and the actual vehicle performance.

[0005] Regarding the connection between motion sickness assessment and control execution, in existing technologies, the output of motion sickness prediction models is mostly used for assessment or alarm purposes. For example, when the motion sickness index exceeds a threshold, a warning is issued to the driver, or adjustments to comfort assistance functions such as air conditioning and seat massage are passively triggered. However, the technical path for further transforming motion sickness prediction results into specific control parameter adjustment schemes has not been fully explored in existing technologies.

[0006] Furthermore, existing motion sickness control strategies mostly employ fixed modes or preset rules, such as limiting acceleration output in a fixed manner once comfort mode is activated. This type of open-loop control makes it difficult to dynamically and accurately adjust parameters based on real-time road conditions, vehicle speed changes, and the actual motion sickness risk level of passengers. This may lead to excessive restrictions on vehicle performance at unnecessary times, while insufficient intervention may occur during high-risk situations. Additionally, since optimization results typically require multiple iterations of simulation calculations, real-vehicle testing, and parameter correction to converge to a satisfactory solution, the entire calibration cycle is lengthy, and development efficiency needs improvement. Summary of the Invention

[0007] The purpose of this invention is to propose a method and system for reverse optimization of control parameters for electric passenger vehicles, which can solve problems such as the disconnect between existing motion sickness assessment technology and vehicle control execution, the reliance on manual optimization process, low efficiency, and lack of closed-loop verification.

[0008] To achieve the above objectives, in a first aspect, the present invention proposes a method for inverse optimization of control parameters for electric passenger vehicles, comprising: The positive motion sickness assessment and attribution module is used to calculate the motion sickness comfort index and locate the motion sickness-causing factors; The optimization target definition module is used to set optimization targets and performance constraints based on the diagnostic results of the positive motion sickness evaluation and attribution module. The vehicle dynamics simulation model module is used to build parametric vehicle models and perform virtual tests. The inverse parameter optimization engine is used to perform inverse optimization of vehicle control parameters based on quantized optimization objectives, performance constraints, and vehicle dynamics simulation models. The optimization result output and verification module is used to write the optimized candidate control parameter set to the vehicle controller and form a closed-loop iteration through real vehicle data acquisition and comparison verification.

[0009] Beneficial effects of the basic solution: Through the positive motion sickness evaluation and attribution module, a quantitative motion sickness comfort index can be output in real time, avoiding the limitations of traditional subjective body feeling evaluation, accurately identifying the core motion sickness-causing factors under various driving conditions such as acceleration, braking, and steering, realizing the transformation of driving discomfort problems from vague perception to precise source tracing, and providing a clear diagnostic basis for subsequent optimization.

[0010] The optimization target definition module formulates adaptability optimization indicators based on motion sickness diagnosis results, and simultaneously defines multiple performance constraints such as vehicle power response, driving stability, energy consumption, and handling. This avoids sacrificing the basic driving performance of the vehicle by solely pursuing anti-motion sickness comfort, and ensures that the optimization solution is compliant, practical, and in line with the overall vehicle development needs.

[0011] The vehicle dynamics simulation model builds a parameterized digital vehicle model, which can complete virtual simulation testing under all working conditions, replacing a large number of traditional real vehicle preliminary tests. This effectively reduces the cost of vehicle tuning and testing, site costs, and labor costs, shortens the development and iteration cycle of vehicle driving comfort, and improves R&D efficiency.

[0012] The reverse parameter optimization engine automatically completes the intelligent optimization of vehicle control parameters across the entire domain based on predetermined optimization goals and constraints, combined with the dynamic model. It abandons the traditional trial-and-error tuning mode that relies on engineers' experience, greatly improves the accuracy and efficiency of parameter tuning, and quickly outputs the optimal combination of control parameters.

[0013] The optimization result output and verification module includes the entire process of simulation optimization, controller flashing, real vehicle testing, and effect comparison, continuously correcting optimization deviations and ensuring that the vehicle's ride comfort tuning effect continues to converge to the optimal state.

[0014] As a feasible preferred embodiment, the positive motion sickness evaluation and attribution module includes: The motion sickness prediction submodule is used to calculate the motion sickness index based on vehicle motion parameters and vehicle state parameters using a machine learning motion sickness prediction model; and The interpretability analysis submodule is used to reverse analyze the key kinematic features that cause motion sickness and their contribution when the motion sickness index exceeds a preset threshold.

[0015] As a feasible and preferred option, the machine learning motion sickness prediction model adopts the XGBoost gradient boosting decision tree architecture. Its input feature vector includes vehicle motion parameters and vehicle state parameters after preprocessing and time / frequency domain feature extraction, and the output is the motion sickness index prediction value. Alternatively, a long short-term memory network temporal neural network architecture can be used, taking the time-series data of vehicle motion parameters in a fixed time window as input, and outputting the motion sickness index by stacking LSTM layers and fully connected layers.

[0016] As a feasible preferred embodiment, the optimization target definition module includes: The target setting submodule is used to set specific quantitative optimization targets based on the diagnostic results of the positive evaluation module; The constraint management submodule is used to set performance constraints, transforming optimization requirements into mathematical constraints that the optimization algorithm can handle.

[0017] As a feasible preferred embodiment, the vehicle dynamics simulation model module includes: The model parameter calibration submodule is used to perform high-precision calibration of the simulation model. The parameters include vehicle mass, moment of inertia, suspension K&C characteristic parameters, tire model parameters, and motor / brake model parameters. The virtual test submodule is used to receive specific vehicle control parameters, simulate their dynamic response under standard test conditions, and output a virtual sequence of vehicle motion parameters.

[0018] As a feasible preferred embodiment, the inverse parameter optimization engine includes: The optimization algorithm library submodule contains built-in intelligent optimization algorithms for iteratively searching for Pareto optimal solutions within the parameter space. The objective function and constraint processor submodule is used to formalize the optimization problem into a constrained multi-objective optimization problem and to construct the objective function and constraints. The Iteration Controller submodule is used to control the entire iterative process of reverse optimization.

[0019] As a feasible preferred option, the intelligent optimization algorithm adopts a non-dominated sorting genetic algorithm with an elitist strategy. By simulating the selection, crossover, and mutation operations of biological evolution, it selects superior individuals to enter the next generation based on non-dominated sorting and crowding distance until the convergence condition is met.

[0020] As a feasible preferred option, the objective function includes a core optimization objective function that minimizes the predicted motion sickness index, and an acceleration time penalty term that maintains dynamism; the constraints include at least one of the following: acceleration time deterioration threshold, energy recovery deceleration lower limit, suspension actuator travel limit, and control parameter feasible region.

[0021] As a feasible and preferred solution, the optimization result output and verification module includes: The parameter flashing submodule is used to receive the optimal parameter set and flash the parameters to the corresponding vehicle controller through a calibration tool; the parameter flashing submodule writes the optimized parameters into the calibration area of ​​the vehicle controller through the vehicle bus, or formally flashes them into the Flash memory of the vehicle controller and motor controller through the unified diagnostic service protocol. The real vehicle data acquisition submodule is used to collect actual data during the real vehicle verification phase; The comparison and verification submodule is used to compare key indicators before and after optimization to verify the optimization effect, and to feed back the deviation data to correct the simulation model when there is a deviation between the actual vehicle results and the simulation prediction.

[0022] A method for reverse optimization of control parameters of electric passenger vehicles utilizes the aforementioned reverse optimization system for control parameters of electric passenger vehicles. Attached Figure Description

[0023] Figure 1 This is a schematic block diagram of the overall architecture of a reverse optimization system for control parameters of an electric passenger vehicle.

[0024] Figure 2This is a schematic diagram illustrating the workflow of a reverse optimization method for control parameters of an electric passenger vehicle. Detailed Implementation

[0025] To make the technical solution and advantages of this application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only some embodiments of the present invention, and are only used to explain this application, not to limit it. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated; they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the accompanying drawings of the following embodiments represent the same features or components, and can be applied to different embodiments.

[0026] Furthermore, unless otherwise defined, the technical or scientific terms used in this invention description shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings.

[0028] Reference Figure 1 This disclosure provides an electric passenger vehicle control parameter reverse optimization system, including a forward motion sickness evaluation and attribution module, an optimization target definition module, a vehicle dynamics simulation model module, a reverse parameter optimization engine, and an optimization result output and verification module.

[0029] The connections between the modules are as follows: the output of the forward motion sickness evaluation and attribution module is connected to the input of the optimization target definition module; the output of the optimization target definition module is connected to the input of the inverse parameter optimization engine; the vehicle dynamics simulation model module is connected to both the forward motion sickness evaluation and attribution module and the inverse parameter optimization engine; the output of the inverse parameter optimization engine is connected to the input of the optimization result output and verification module. The feedback of the optimization result output and verification module is then connected to both the vehicle dynamics simulation model module and the forward motion sickness evaluation and attribution module, forming a complete closed-loop iterative system.

[0030] The system's hardware architecture primarily consists of cloud / edge computing servers, in-vehicle execution terminals, and data acquisition modules. These components are connected via in-vehicle Ethernet or high-speed CAN-FD bus to ensure real-time data transmission and low latency. The data processing center, as the core computing unit, typically comprises a high-performance server or in-vehicle domain controller (DCU), and internally houses motion sickness prediction models and optimization algorithm modules. The in-vehicle bus network, acting as the system's "neural network," connects the data processing center to various electronic control units (ECUs), specifically including the vehicle controller (VCU), motor controller (MCU), battery management system (BMS), and chassis controller (CDC controller) via a bus.

[0031] The positive motion sickness assessment and attribution module is used to calculate the motion sickness comfort index in real time and locate the motion sickness-causing factors. It includes a motion sickness prediction submodule and an interpretability analysis submodule.

[0032] The motion sickness prediction submodule integrates a machine learning motion sickness prediction model trained on a large amount of human factors test data. In this embodiment, the model adopts an XGBoost gradient boosting decision tree architecture, and its core hyperparameters are determined through grid search optimization. The specific parameters are set as follows: the base learner is gbtree, the maximum tree depth is set to 6, the learning rate is 0.1, the regularization parameter (gamma) used to control model complexity is set to 0.1, and the subsample ratio is 0.8.

[0033] The model's input feature vector includes vehicle motion parameters after preprocessing and time / frequency domain feature extraction, specifically including: longitudinal acceleration. a x Lateral acceleration a y Vertical acceleration a z yaw rate γ Roll angular velocity ， and their respective first derivative accelerometers j x , j y , j z (i.e., jerk), and vehicle status parameters such as vehicle speed. v The model outputs a motion sickness index (MSI) prediction value ranging from 0 to 10.

[0034] In another preferred embodiment, the positive motion sickness prediction model may also employ a Long Short-Term Memory (LSTM) temporal neural network architecture to better capture the cumulative effect of motion sickness. This network comprises an input layer, two stacked LSTM layers, and a fully connected output layer. The first LSTM layer has 128 memory units and returns the complete sequence; the second LSTM layer has 64 memory units and returns only the final output sequence. A Dropout layer is appended to each LSTM layer with a dropout rate of 0.2 to prevent overfitting. The LSTM layers use the tanh activation function, and the recurrent gates use the sigmoid activation function. Finally, the temporal features extracted by the second LSTM layer are integrated through a fully connected layer with 32 neurons using the ReLU activation function, and the motion sickness index is output by a single-neuron output layer using a linear activation function. This model uses temporal data of vehicle motion parameters within a fixed time window (e.g., 10 seconds) as input for end-to-end training and prediction.

[0035] The interpretability analysis submodule integrates a SHAP (SHapley Additive exPlanations) value analysis unit. When the MSI calculated by the motion sickness prediction submodule exceeds a preset threshold (e.g., MSI > 0.6), the interpretability analysis submodule is activated to reverse-engineer the key kinematic features causing motion sickness and their contribution. For example, the system can identify that "longitudinal jerk is the main cause of current motion sickness, contributing 45%", providing a diagnostic basis for subsequent precise optimization.

[0036] The optimization target definition module is a human-computer interaction interface, usually carried by an in-vehicle central control display or an engineer's workstation software interface, and includes a target setting submodule and a constraint management submodule.

[0037] The goal setting submodule is used to set specific, quantifiable optimization goals based on the diagnostic results of the positive evaluation module. For example, the main goal can be set as "to reduce the motion sickness index from 6.5 to below 4.0".

[0038] The constraint management submodule is used to set performance constraints, transforming fuzzy requirements into mathematical problems that the optimization algorithm can handle. For example, constraints such as "0-100km / h acceleration time degradation not exceeding 0.3 seconds" and "energy recovery deceleration not less than 0.1g" can be set. The constraint management submodule ensures that the optimization process is carried out while meeting the overall vehicle performance requirements.

[0039] The vehicle dynamics simulation model module is a parameterized vehicle model built based on multibody dynamics theory. It can run on high-performance computing units or cloud servers and includes a model parameter calibration submodule and a virtual testing submodule.

[0040] The model parameter calibration submodule is used to perform high-precision calibration of the simulation model. Its model parameters include: vehicle mass. m Moment of inertia I Suspension K&C characteristic parameters, tire model parameters (such as stiffness coefficients in the Pacejka magic formula) B Shape factor C Peak coefficient D Curvature coefficient E These parameters include motor / brake model parameters, etc. These parameters can be initially calibrated using basic vehicle data and then fine-tuned using real vehicle data.

[0041] The virtual testing submodule receives specific vehicle control parameters (such as motor torque response MAP, energy recovery curve, and active suspension damping coefficient) and simulates its dynamic response under standard test conditions (such as a 42-minute urban cycle or FTP-75 cycle). Its core output is a virtual sequence of vehicle motion parameters, including time-series data for triaxial acceleration, angular velocity, and jerk. The model's purpose is to provide an efficient and repeatable "digital twin" testing environment for the optimization process, avoiding frequent real-vehicle testing.

[0042] The inverse parameter optimization engine connects the presentation layer (motion sickness index) and the control layer (vehicle parameters), and includes an optimization algorithm library submodule, an objective function and constraint processor submodule, and an iterative controller submodule.

[0043] The optimization algorithm library submodule incorporates various intelligent optimization algorithms. In this embodiment, the Non-Dominated Sorting Genetic Algorithm with Elite Strategy (NSGA-II) is employed. This algorithm iteratively searches for a Pareto optimal solution set within the parameter space by simulating the selection, crossover, and mutation operations of biological evolution. In one iteration, the algorithm evaluates each individual in the population (i.e., a set of control parameters). x The objective function value corresponding to the non-dominated sort and crowding distance is used to select excellent individuals to enter the next generation until the convergence condition is met (such as reaching the maximum number of iterations or the Pareto front change rate is less than the threshold).

[0044] The objective function and constraint processor submodule is responsible for formalizing the optimization problem into a constrained multi-objective optimization problem and constructing the specific objective function.

[0045] Let the vector of control parameters to be optimized be:

[0046] in, The torque filtering time constant is... For example, the energy recovery intensity coefficient.

[0047] The core optimization objective function is defined as minimizing the predicted motion sickness index:

[0048] in, Indicates control parameters The motion parameters of the virtual vehicle obtained after inputting the vehicle dynamics simulation model are then used to calculate the motion index through the positive motion prediction model.

[0049] At the same time, performance constraints need to be considered, such as maintaining dynamics. An acceleration time penalty term is defined as one of the objective functions:

[0050] in, It is a parameter The simulated 0-100km / h acceleration time is as follows. This is the baseline acceleration time before optimization. This is the maximum allowable deterioration threshold (e.g., 0.3 seconds). This function ensures that a penalty is only incurred when the acceleration time deteriorates beyond the threshold.

[0051] The optimization problem can be expressed as:

[0052] in, Let the objective function vector be... For other inequality constraints (such as the lower limit of energy recovery deceleration, suspension actuator travel limit, etc.). and These are the lower and upper limits of the feasible domain for the control parameters, respectively.

[0053] The inverse parameter optimization engine employs a non-dominated sorting genetic algorithm with an elitist strategy (NSGA-II) to solve the aforementioned multi-objective optimization problem. The algorithm iteratively searches for a Pareto optimal solution set within the parameter space by simulating selection, crossover, and mutation operations in biological evolution.

[0054] In one iteration, the algorithm evaluates each individual in the population (i.e., a set of control parameters). ) corresponding and The values ​​are then used to select the best individuals for the next generation based on the non-dominated sort and crowding distance, until the convergence condition is met (such as reaching the maximum number of iterations or the Pareto front rate of change being less than the threshold).

[0055] The final output is a set of candidate control parameters that achieve the best trade-off between motion sickness comfort and vehicle performance.

[0056] The iterative controller submodule controls the entire iterative process of reverse optimization. Its sub-processes form an internal closed loop, and the specific steps are as follows: Step a: Variable definition. Define the vehicle control parameters to be optimized (such as torque filter time constant τ, energy recovery intensity coefficient, etc.). K regen (e.g., damping curve coefficient of CDC shock absorber) are defined as optimization variables.

[0057] Step b: Generate candidate parameters. Based on the current search state, the optimization algorithm proposes a new set of control parameters x. (k) (k is the number of iterations).

[0058] Step c: Virtual simulation test. This set of parameters x... (k) Given a calibrated vehicle dynamics model, run it under standard operating conditions and predict the vehicle's virtual motion response sequence {ax(t), ay(t), az(t), γ(t)}. (t),jx(t),jy(t),jz(t)}.

[0059] Step d: Virtual comfort assessment. Input the virtual motion parameters obtained in the previous step into the positive motion sickness prediction model to calculate the predicted "virtual" motion sickness index. MSI (k) = F 1( x (k) ).

[0060] Step e: Multi-objective assessment and decision-making. Evaluate the motion sickness index under this set of parameters. MSI (k) Acceleration time t acc ( x (k) Whether the preset goals and constraints are met.

[0061] Step f: Iterative judgment. If the evaluation result satisfies all convergence conditions (such as the motion sickness index meeting the standard and performance constraints being met), the loop is exited, and the current parameters are output as the optimal solution; if not, the optimization algorithm intelligently adjusts the search strategy based on the evaluation result, returns to step b, and generates the next set of candidate parameters x. (k+1) Continue iterating.

[0062] The optimization results output and verification module includes a parameter writing submodule, a real vehicle data acquisition submodule, and a comparison and verification submodule.

[0063] The parameter flashing submodule includes a parameter flashing interface for receiving the optimal parameter set output by the reverse optimization engine. The parameters are then flashed to the corresponding vehicle controller (such as the VCU, MCU, or CDC controller) using a calibration tool. Specifically, the optimized parameters are written to the VCU's calibration data memory via the vehicle bus, completing a software-level "virtual flashing".

[0064] The real-vehicle data acquisition submodule is responsible for collecting actual data during the real-vehicle verification phase. The test vehicle travels according to preset urban cycle conditions (such as FTP-75 or a custom condition). A motion sickness-inducing bionic robot (dummy) is fixed in the rear right seat. The dummy's built-in six-axis inertial measurement unit (IMU) collects three-axis acceleration data in real time. a x , a y , a z ) and angular velocity (γ, The data is transmitted to the data processing center via a bus.

[0065] The comparison and verification submodule compares key indicators (motion sickness index, jerk RMS value, etc.) before and after optimization to verify the optimization effect and form a closed loop. If there is an acceptable deviation between the actual vehicle results and the simulation prediction, the optimization is successful; if the deviation is large, the deviation data can be fed back to correct the simulation model and start a new round of optimization.

[0066] When the iteration converges, the system will formally write the optimal control parameters (such as a specific torque response curve MAP) into the Flash memory of the VCU and MCU through the UDS (Unified Diagnostic Service) protocol, thus completing the comfort calibration of the whole vehicle.

[0067] Reference Figure 2 This disclosure also provides a method for reverse optimization of control parameters for electric passenger vehicles, utilizing the aforementioned reverse optimization system for electric passenger vehicle control parameters. The method includes the following steps.

[0068] Step S1: Problem Diagnosis and Target Quantification. A standard motion sickness comfort test is performed on the vehicle to be optimized. The test vehicle travels under preset urban driving conditions, with a dummy fixed in the right rear seat, and the IMU collects motion parameters in real time. The positive motion sickness evaluation and attribution module calculates the initial motion sickness index (MSI) through the motion sickness prediction submodule. If the MSI exceeds a preset threshold (e.g., MSI > 0.6), the interpretability analysis submodule is activated to locate key motion sickness-inducing factors. Based on engineering requirements, a clear quantitative optimization target (e.g., reducing the MSI from 6.5 to below 4.0) and constraints (e.g., acceleration time deterioration not exceeding 0.3 seconds) are set in the optimization target definition module.

[0069] Step S2: High-precision calibration of the simulation model. The initial control parameters of the vehicle are input into the vehicle dynamics simulation model module, and the motion data collected by the real vehicle under standard operating conditions are compared and the parameters are fine-tuned to ensure that the digital model can reproduce the dynamic characteristics of the real vehicle with high fidelity, thus ensuring the effectiveness of subsequent virtual optimization results.

[0070] Step S3, model-based inverse collaborative optimization. The inverse parameter optimization engine 40 initiates internal closed-loop iteration: The NSGA-II algorithm in the optimization algorithm library submodule generates a set of candidate control parameters x. (k) For example, adjusting the torque filter time constant τ from 50ms to 80ms, while simultaneously increasing the energy recovery intensity coefficient. K regen The value was adjusted from 0.8 to 0.6.

[0071] The virtual testing submodule inputs this set of parameters into the calibrated vehicle dynamics model and runs it for 42 minutes under urban driving conditions. In the simulation, the kinematic description of the powertrain is as follows: During acceleration, the VCU sends a target torque command to the MCU based on the optimized throttle mapping curve. The MCU controls the rotor of the drive motor to accelerate, driving the wheels to rotate via the drive shaft, generating longitudinal acceleration. a x During deceleration, when the driver releases the accelerator pedal, the VCU, based on the new energy recovery strategy, controls the motor to enter generator mode. At this time, the inertial kinetic energy of the wheels drives the motor rotor to rotate through the drive shaft, generating braking torque. The optimized strategy limits the acceleration during this process. j x This avoids abrupt nose-nodding movements in the vehicle.

[0072] The motion sickness prediction submodule inputs the virtual motion parameters output from the simulation into the XGBoost or LSTM model to calculate the predicted virtual motion sickness index. MSI (k) .

[0073] Objective function and constraint processor submodule evaluation F 1( x (k) )and F 2( x (k) ), determine whether all constraints are satisfied.

[0074] The iterative controller submodule, based on the evaluation results, if MSI (k) If the error rate is >4.0 or the acceleration time deteriorates by more than 0.3 seconds, the algorithm will intelligently adjust its search strategy and generate the next set of candidate parameters. x (k+1)Continue iterating until the globally optimal solution is found.

[0075] Step S4: Result Application and Closed-Loop Verification. The optimization result output and verification module writes the optimal control parameter set obtained in step S3 into the Flash memory of the VCU and MCU via the UDS protocol. Subsequently, the vehicle is driven to undergo rapid real-vehicle retesting under real standard operating conditions. The real-vehicle data acquisition submodule collects actual motion parameters, and the comparison and verification submodule compares the MSI and jerk RMS values ​​before and after optimization. If the real-vehicle MSI decreases from 6.5 to 3.8, and the acceleration time only deteriorates by 0.2 seconds (within the 0.3-second constraint), the optimization is successful; if the deviation is large, the deviation data is fed back to the vehicle dynamics simulation model module to correct the model parameters and start a new round of optimization.

[0076] The above content is merely an embodiment of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can improve and implement this solution based on the guidance provided in this application and their own capabilities. Typical well-known structures or operating methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for inverse optimization of control parameters for electric passenger vehicles, characterized in that, include: The positive motion sickness assessment and attribution module is used to calculate the motion sickness comfort index and locate the motion sickness-causing factors; The optimization target definition module is used to set optimization targets and performance constraints based on the diagnostic results of the positive motion sickness evaluation and attribution module. The vehicle dynamics simulation model module is used to build parametric vehicle models and perform virtual tests. The inverse parameter optimization engine is used to perform inverse optimization of vehicle control parameters based on quantized optimization objectives, performance constraints, and vehicle dynamics simulation models. The optimization result output and verification module is used to write the optimized candidate control parameter set to the vehicle controller and form a closed-loop iteration through real vehicle data acquisition and comparison verification.

2. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 1, characterized in that, The positive motion sickness evaluation and attribution module includes: The motion sickness prediction submodule is used to calculate the motion sickness index based on vehicle motion parameters and vehicle state parameters using a machine learning motion sickness prediction model; and The interpretability analysis submodule is used to reverse analyze the key kinematic features that cause motion sickness and their contribution when the motion sickness index exceeds a preset threshold.

3. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 2, characterized in that, The machine learning motion sickness prediction model adopts the XGBoost gradient boosting decision tree architecture. Its input feature vector includes vehicle motion parameters and vehicle state parameters after preprocessing and time / frequency domain feature extraction. The output is the motion sickness index prediction value. Alternatively, a long short-term memory network temporal neural network architecture can be used, taking the time-series data of vehicle motion parameters in a fixed time window as input, and outputting the motion sickness index by stacking LSTM layers and fully connected layers.

4. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 1, characterized in that, The optimization target definition module includes: The target setting submodule is used to set specific quantitative optimization targets based on the diagnostic results of the positive evaluation module; The constraint management submodule is used to set performance constraints, transforming optimization requirements into mathematical constraints that the optimization algorithm can handle.

5. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 1, characterized in that, The vehicle dynamics simulation model module includes: The model parameter calibration submodule is used to perform high-precision calibration of the simulation model. The parameters include vehicle mass, moment of inertia, suspension K&C characteristic parameters, tire model parameters, and motor / brake model parameters. The virtual test submodule is used to receive specific vehicle control parameters, simulate their dynamic response under standard test conditions, and output a virtual sequence of vehicle motion parameters.

6. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 1, characterized in that, The inverse parameter optimization engine includes: The optimization algorithm library submodule contains built-in intelligent optimization algorithms for iteratively searching for Pareto optimal solutions within the parameter space. The objective function and constraint processor submodule is used to formalize the optimization problem into a constrained multi-objective optimization problem and to construct the objective function and constraints. The Iteration Controller submodule is used to control the entire iterative process of reverse optimization.

7. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 6, characterized in that, The intelligent optimization algorithm adopts a non-dominated sorting genetic algorithm with an elitist strategy. By simulating the selection, crossover and mutation operations of biological evolution, it selects excellent individuals to enter the next generation based on non-dominated sorting and crowding distance until the convergence condition is met.

8. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 6, characterized in that, The objective function includes a core optimization objective function that minimizes the predicted motion sickness index, and an acceleration time penalty term that maintains dynamism; the constraints include at least one of the following: acceleration time deterioration threshold, energy recovery deceleration lower limit, suspension actuator travel limit, and control parameter feasible region.

9. The method for reverse optimization of control parameters of an electric passenger vehicle according to claim 1, characterized in that, The optimization result output and verification module includes: The parameter flashing submodule is used to receive the optimal parameter set and flash the parameters to the corresponding vehicle controller through a calibration tool; the parameter flashing submodule writes the optimized parameters into the calibration area of ​​the vehicle controller through the vehicle bus, or formally flashes them into the Flash memory of the vehicle controller and motor controller through the unified diagnostic service protocol. The real vehicle data acquisition submodule is used to collect actual data during the real vehicle verification phase; The comparison and verification submodule is used to compare key indicators before and after optimization to verify the optimization effect, and to feed back the deviation data to correct the simulation model when there is a deviation between the actual vehicle results and the simulation prediction.

10. A method for inverse optimization of control parameters for electric passenger vehicles, characterized in that: The system employs a reverse optimization system for control parameters of an electric passenger vehicle as described in any one of claims 1-9.

Images