MPC-based vehicle adaptive path control method, apparatus and device, and medium

Abstract

The invention discloses a vehicle adaptive path control method and device based on MPC, equipment and a medium, and relates to the technical field of vehicle automatic driving, and the method comprises the steps: obtaining vehicle state data of a target vehicle; determining an optimization algorithm corresponding to the target vehicle based on the vehicle state data; based on an optimization algorithm, calculating a prediction time domain of the model prediction control MPC to obtain a target time domain; and obtaining a target time domain parameter of the target vehicle based on the target time domain and the MPC, and performing path control based on the target time domain parameter. The method can realize accurate path control.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a vehicle adaptive path control method, device, equipment and medium based on MPC. Background Technology

[0002] With the rapid development of autonomous driving technology, new energy sanitation vehicles, as equipment for urban cleaning operations, have placed higher demands on the intelligence, precision, and energy efficiency of trajectory tracking control. Model Predictive Control (MPC), with its advantage of explicitly handling multivariable and constrained optimization problems, has become the mainstream technical solution in the field of trajectory tracking control for autonomous vehicles.

[0003] To adapt to different driving conditions and operational needs, existing MPC-based trajectory tracking control schemes have formed three typical technical paths: fixed-parameter MPC schemes, rule-switching-based parameter adjustment schemes, and online optimization-based parameter adaptive schemes. However, none of these schemes fully consider the operating modes of new energy sanitation vehicles, leading to inaccurate path tracking.

[0004] Therefore, there is an urgent need for a method to achieve accurate path tracking control. Summary of the Invention

[0005] This application provides a vehicle adaptive path control method, device, equipment, and medium based on MPC, which can achieve accurate path control.

[0006] To achieve the above objectives, this application adopts the following technical solution: Firstly, this application provides a vehicle adaptive path control method based on MPC, including: Obtain vehicle status data for the target vehicle; Based on vehicle status data, determine the optimization algorithm corresponding to the target vehicle; Based on the optimization algorithm, the prediction time domain of Model Predictive Control (MPC) is calculated to obtain the target time domain; Based on the target time domain and MPC, the target time domain parameters of the target vehicle are obtained, and path control is performed based on the target time domain parameters.

[0007] In some possible implementations, the optimization algorithm corresponding to the target vehicle is determined based on vehicle state data, including: Based on vehicle status data, determine the driving mode of the target vehicle; Based on the driving mode, the corresponding optimization algorithm for the target vehicle is determined.

[0008] In some possible implementations, the driving mode of the target vehicle is determined based on vehicle status data, including: If the operation mode is identified as the first identifier and the path curvature meets the preset first operation conditions, the driving mode is determined to be the first mode. If the operating mode is identified as the second mode and the vehicle speed is higher than the preset first speed threshold, the driving mode is determined to be the second mode.

[0009] Among some possible implementations, the optimization algorithm corresponding to the target vehicle is determined based on the driving mode, including: When the driving mode is in the first mode, the optimization algorithm corresponding to the target vehicle is determined to be the improved whale optimization algorithm; where the first mode refers to the situation where the target vehicle is performing a cleaning task. When the driving mode is the second mode, the optimization algorithm corresponding to the target vehicle is determined to be the adaptive differential evolution algorithm; where the second mode refers to the situation where the target vehicle is performing a driving task.

[0010] In some possible implementations, the shrinkage factor of the improved whale optimization algorithm is determined in the following way:

[0011] in, It is a contraction factor; This represents the current iteration number; This represents the maximum number of iterations. As an indicator of population diversity; This represents the diversity impact coefficient.

[0012] In some possible implementations, the mutation factor of the adaptive differential evolution algorithm is determined in the following way:

[0013] in, It serves as an individual identifier to distinguish different individuals; Let be the variation factor corresponding to the i-th individual; This is the lower bound of the variable factor; This is the upper bound of the variable factor; Let i be the fitness of the i-th individual; The minimum fitness value of all individuals in the population. This represents the maximum fitness value among all individuals in the population.

[0014] In some possible implementations, the prediction time domain of Model Predictive Control (MPC) is calculated based on an optimization algorithm to obtain the target time domain, including: At least two candidate time-domain prediction values ​​are generated by optimizing the algorithm. MPC simulation is performed on each candidate time-domain prediction value and the fitness function is calculated. Elite selection and population update are performed based on the fitness function results until the convergence condition is met, thus obtaining the target time domain.

[0015] Secondly, this application provides a vehicle adaptive path control device based on MPC, comprising: The acquisition module is used to acquire vehicle status data of the target vehicle; The algorithm determination module is used to determine the optimization algorithm corresponding to the target vehicle based on vehicle status data. The time-domain determination module is used to calculate the prediction time domain of Model Predictive Control (MPC) based on optimization algorithms to obtain the target time domain. The path control module is used to obtain the target time domain parameters of the target vehicle based on the target time domain and MPC, and to perform path control based on the target time domain parameters.

[0016] Thirdly, this application provides a computing device, including a memory and a processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of the first aspects.

[0017] Fourthly, this application provides a computer-readable storage medium for storing a computer program for performing the method as described in any one of the first aspects.

[0018] Fifthly, this application provides a computer program product comprising one or more computer instructions, wherein when the computer instructions are executed by a computer, the computer performs the method as described in any one of the first aspects.

[0019] As can be seen from the above technical solution, this application has at least the following beneficial effects: In this application, by acquiring the vehicle state data of the target vehicle, real-time synchronous perception of the environment and its own state is achieved, providing objective and quantitative input basis for all subsequent adaptive decisions and ensuring that control decisions closely align with actual operating conditions. Furthermore, based on the vehicle state data, the optimization algorithm corresponding to the target vehicle is determined, achieving dynamic matching between the control strategy and the driving scenario. Subsequently, based on the optimization algorithm, the prediction time domain of the Model Predictive Control (MPC) is calculated to obtain the target time domain, ensuring on-demand allocation and efficient utilization of computing resources while maintaining or even improving control performance. Finally, based on the target time domain and MPC, the target time domain parameters of the target vehicle are obtained, and path control is performed based on these parameters, achieving high-performance real-time control under parameter adaptation. This solution, by determining the optimization algorithm corresponding to the target vehicle based on vehicle state data, achieves dynamic selection or switching of the most suitable optimization algorithm for the current scenario, overcoming the limitations of the traditional one-size-fits-all approach. Moreover, by introducing the target time domain, online real-time optimization of the prediction time domain is ensured, ultimately achieving accurate path control.

[0020] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0021] Figure 1 An application environment diagram for a vehicle adaptive path control method based on MPC provided in this application embodiment; Figure 2 This is a flowchart illustrating a vehicle adaptive path control method based on MPC according to an embodiment of this application. Figure 3 This is a schematic diagram of the structure of a vehicle adaptive path control device based on MPC according to an embodiment of this application; Figure 4 This is a schematic diagram of a computing device provided in an embodiment of this application. Detailed Implementation

[0022] The terms "first," "second," and "third," etc., used in this application specification and accompanying drawings are used to distinguish different objects, not to limit a specific order.

[0023] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0024] To ensure clarity and conciseness in the description of the following embodiments, a brief introduction to the related technologies is given first: Model Predictive Control (MPC) is an advanced process control method. Its principle can be summarized as rolling optimization and feedback correction. In vehicle control, the MPC controller, in each control cycle, predicts the vehicle's trajectory over a future time domain based on the current vehicle state (such as position, speed, and heading angle) and a predictive model describing the vehicle's dynamics or kinematics. Then, by solving a constrained optimization problem (the objectives are typically tracking a desired path, ensuring ride comfort, and ensuring safety), it calculates the optimal control sequence from the current moment to several future moments, such as steering wheel angle and acceleration, but only executes the first control command in the sequence. This process is repeated in the next cycle. Its advantage lies in its ability to explicitly handle multi-objective optimization and system constraints (such as actuator limits and road boundaries).

[0025] The prediction time domain is a crucial parameter in MPC (Multi-Process Control), referring to the length of the future time range from which the controller makes predictions. The length of the prediction time domain directly determines the scale of the optimization problem and how far the controller can "see." If the time domain is too short, the controller may lack foresight and be unable to handle sharp turns or sudden situations, leading to decreased control performance or even instability. If the time domain is too long, the computational load of the optimization problem increases dramatically, potentially making it impossible to solve within the real-time control cycle, thus affecting the real-time performance of the control.

[0026] In traditional vehicle MPC path control, the prediction time domain (or together with the control time domain) is usually set to a fixed value. This one-size-fits-all approach can lead to a performance-efficiency trade-off in complex real-world vehicle driving scenarios, specifically manifested as follows: Poor scene adaptability and unstable performance. High-speed / low curvature roads: Vehicles need a longer prediction time domain to ensure stability and smoothness at high speeds; if the fixed time domain is set too short, the controller's look-ahead is insufficient, which may generate aggressive or oscillating control commands.

[0027] Low-speed / high-curvature roads (such as parking spaces, sharp bends). The path changes drastically, and the uncertainty of future state predictions is high and the reference value decreases. If the fixed time domain is set too long, the controller will solve an optimization problem containing a lot of unnecessary and unreliable prediction information, which not only increases the computational burden, but may also affect the current tracking accuracy of key path segments due to over-optimization of uncertain long-term trajectories.

[0028] Low utilization of computing resources and significant challenges in real-time performance. To ensure performance in the worst-case scenarios, traditional methods tend to conservatively set a sufficiently long fixed time domain. This results in the controller solving an excessively large and computationally redundant optimization problem under most normal driving conditions, needlessly consuming the valuable resources of the onboard computing unit, increasing computational latency, and limiting the possibility of deploying MPC on lower-cost hardware.

[0029] The root cause of the above problems is that the dynamic characteristics of the vehicle, the curvature of the road, and the required control granularity change in real time with the driving state and the scene. A fixed prediction time domain cannot match this change, and it cannot achieve intelligent adjustment such as "looking far when it should look far and looking near when it should look near".

[0030] In view of this, this application provides a vehicle adaptive path control method based on MPC. This method dynamically determines an optimization algorithm based on vehicle state data and adaptively calculates the prediction time domain (target time domain) accordingly. This breaks away from the fixed prediction time domain paradigm and introduces a mechanism for decision-making based on real-time vehicle state data (which may include vehicle speed, yaw rate, lateral deviation from the reference path, road curvature, etc.), ultimately achieving online adaptive adjustment of the prediction time domain. It automatically extends the time domain to improve stability during high-speed straight driving and automatically shortens the time domain to improve calculation speed and tracking agility during low-speed turning or parking. Thus, without sacrificing control performance, it improves the computational efficiency and real-time performance of the entire MPC controller, enabling it to better adapt to complex and changing real-world driving environments and facilitating deployment on resource-constrained vehicle platforms.

[0031] To make the technical solution of this application clearer and easier to understand, the application scenarios of the technical solution of this application are described below with reference to the accompanying drawings. Figure 1 As shown in the figure, this figure is a schematic diagram of an application scenario provided by an embodiment of this application.

[0032] In this application scenario, terminal 102 continuously collects and reports real-time vehicle status data (such as position, speed, heading angle, etc.) to server 104. Based on the received status data, server 104 analyzes the current driving scenario, dynamically selects the matching optimization algorithm, and calculates the optimal prediction time domain (target time domain) accordingly, thereby completing the solution of the model predictive control problem. Subsequently, server 104 sends the calculated target time domain parameters and reference control sequence to terminal 102. After receiving the instruction, terminal 102 immediately updates the configuration of the local controller according to the new time domain parameters, and combines the sent reference sequence with local real-time sensor information to perform high-frequency rolling optimization and precise path tracking control.

[0033] To make the technical solution of this application clearer and easier to understand, the following describes a vehicle adaptive path control method based on MPC provided by the embodiments of this application, in conjunction with the above application scenarios. Figure 2 As shown, this figure is a flowchart of a vehicle adaptive path control method based on MPC provided in an embodiment of this application.

[0034] S201. Obtain the vehicle status data of the target vehicle.

[0035] Among them, target vehicles refer to vehicles selected as data collection objects in specific application scenarios, such as new energy sanitation vehicles, which are selected as algorithm adaptation objects, including operation vehicles performing cleaning tasks, driving tasks, and other different scenarios.

[0036] Vehicle status data refers to a set of various parameters that can objectively reflect the operating status, hardware condition and functional status of a target vehicle, including but not limited to driving parameters (driving speed, acceleration, mileage, etc.), power system parameters (remaining battery charge, motor output power, fuel level, etc.), handling status parameters (braking status, steering angle, gear information, etc.), hardware health parameters (tire pressure data, coolant temperature, fault diagnostic codes, component wear data, etc.), and environmental interaction parameters (current geographical location, driving road condition related data, etc.).

[0037] For example, vehicle status data can be collected from a target vehicle through multi-dimensional acquisition methods. From the perspective of acquisition dimensions, these methods can include direct acquisition from sensors, reading data from in-vehicle systems, acquisition through communication transmission, and fusion of multi-source data.

[0038] Sensor direct acquisition involves using various onboard sensors (such as speed sensors, acceleration sensors, tire pressure sensors, temperature sensors, fault diagnosis sensors, etc.) on the target vehicle to perceive various physical quantities and status information during vehicle operation in real time, and convert them into transmittable electrical or digital signals. Vehicle system data reading involves retrieving recorded vehicle operating parameters, fault logs, historical status data, etc., from the storage units of the target vehicle's vehicle control system (such as the body control system, battery management system, in-vehicle infotainment system, etc.). Communication transmission acquisition involves using vehicle communication modules (such as vehicle networking communication modules, Bluetooth, Wi-Fi, CAN bus communication, etc.) to transmit real-time data collected by sensors or historical data stored in the vehicle system to vehicle terminals, cloud servers, or external data receiving devices, thereby achieving remote or local data acquisition. Multi-source data fusion is used to meet the high-precision data requirements in complex scenarios. It can combine the above methods and use mechanisms such as data synchronization, redundancy verification, and cross-validation to integrate vehicle status data from different sources, thereby improving the integrity and accuracy of the data.

[0039] S202. Based on vehicle status data, determine the optimization algorithm corresponding to the target vehicle.

[0040] One possible approach is to determine the driving mode of a target vehicle based on vehicle status data, and then determine the corresponding optimization algorithm for the target vehicle based on the driving mode.

[0041] The vehicle status data includes vehicle speed, path curvature, and operation mode identifier.

[0042] For example, based on vehicle status data, the driving mode of the target vehicle is determined; based on the driving mode, the corresponding optimization algorithm for the target vehicle is determined; when the operation mode is identified as the first identifier and the path curvature meets the preset first operation condition, the driving mode is determined as the first mode; wherein, the first mode is the situation where the target vehicle is performing a cleaning task; when the operation mode is identified as the second identifier and the vehicle speed is higher than the preset first speed threshold, the driving mode is determined as the second mode; wherein, the second mode is the situation where the target vehicle is performing a driving task.

[0043] Optionally, when the driving mode is the first mode, the optimization algorithm corresponding to the target vehicle is determined to be the improved whale optimization algorithm; when the driving mode is the second mode, the optimization algorithm corresponding to the target vehicle is determined to be the adaptive differential evolution algorithm.

[0044] The shrinkage factor of the improved whale optimization algorithm is determined in the following way:

[0045] in, It is a contraction factor; This represents the current iteration number; This represents the maximum number of iterations. As an indicator of population diversity; This represents the diversity impact coefficient.

[0046] The mutation factor of the adaptive differential evolution algorithm is determined in the following way:

[0047] in, It serves as an individual identifier to distinguish different individuals; Let be the variation factor corresponding to the i-th individual; This is the lower bound of the variable factor; This is the upper bound of the variable factor; Let i be the fitness of the i-th individual; The minimum fitness value of all individuals in the population. This represents the maximum fitness value among all individuals in the population.

[0048] Among them, vehicle status data refers to a set of technical parameters that reflect the operation and work of the target vehicle, including vehicle speed, path curvature, and work mode identification. It is the basis for judging the driving mode and selecting the optimization algorithm.

[0049] Optimization algorithms refer to intelligent algorithms used for adaptive adjustment in the time domain of MPC prediction, including the improved Whale Optimization Algorithm (WOA) and the Adaptive Differential Evolution (ADE) algorithm, which can achieve performance optimization according to the needs of vehicle scenarios.

[0050] Driving mode refers to the type of target vehicle operation scenario determined based on vehicle status data, mainly including the first mode for performing cleaning tasks and the second mode for performing driving tasks.

[0051] The operation mode identifier is an identifier used to distinguish the operation type of the target vehicle. It is divided into a first identifier (corresponding to the cleaning task) and a second identifier (corresponding to the driving task).

[0052] The preset first working condition is a path curvature threshold standard set for determining the first mode. When the path curvature meets this condition, it can be determined as a cleaning task scenario by combining it with the first identifier.

[0053] The preset first speed threshold is a critical value for vehicle speed set to determine the second mode. When the vehicle speed is higher than this threshold, it can be determined as a driving task scenario by combining it with the second identifier.

[0054] For example, the environmental perception module of the target vehicle can collect vehicle status data such as vehicle speed, path curvature, and operation mode identifier to ensure the real-time performance and integrity of the data. Furthermore, based on the collected vehicle status data, two-layer judgment actions can be performed: first, determining the operation mode identifier type (first identifier or second identifier) ​​to clarify the vehicle's basic operation direction; second, determining whether key parameters meet preset conditions (path curvature meets the first operation condition, vehicle speed is higher than the first speed threshold), ultimately determining the target vehicle's driving mode (first mode or second mode). Then, based on the judgment result, an algorithm matching action can be performed: if the driving mode is the first mode (cleaning task), the improved whale optimization algorithm is matched; if it is the second mode (driving task), the adaptive differential evolution algorithm is matched, achieving a precise correspondence between the scenario and the algorithm.

[0055] Furthermore, for the matched optimization algorithm, key parameter calculations can be performed. For example, for the improved whale optimization algorithm, the shrinkage factor can be calculated by combining the current iteration count, maximum iteration count, population diversity index, and diversity influence coefficient; for the adaptive differential evolution algorithm, the mutation factor corresponding to each individual can be calculated by combining individual fitness, population fitness extreme value, and mutation factor boundary, ensuring that the algorithm parameters adapt to the scenario requirements. After completing algorithm matching and parameter calculation, the optimized algorithm and key parameters adapted to the target vehicle's driving mode are output, providing support for adaptive adjustment in the MPC prediction time domain, and ultimately achieving performance optimization of the target vehicle in different scenarios.

[0056] It should be noted that the improved whale optimization algorithm and the adaptive differential evolution algorithm share the same goal: to quickly and accurately find an optimal prediction time domain for the MPC controller, i.e., the target time domain N value, by maximizing the fitness function fitness(N).

[0057] The improved whale optimization algorithm first randomly generates a group of individual whales, each representing a candidate predicted time-domain value N, forming an initial population. The fitness (N) of each individual in the initial population is then calculated. The main loop iterates, updating the parameters of each whale (i.e., each candidate N value) in each iteration, such as the shrinkage factor. The first coefficient A and the second coefficient C, the random number p, and the spiral shape parameter L.

[0058] Optionally, the contraction factor The update method is as shown in the above formula, and adaptive updates are performed.

[0059] The value of the random number p determines the current whale's behavior; specifically, when p < 0.5, the absolute value of the coefficient A, |A|, is checked. If |A| ≥ 1, a global search is performed, that is, a non-current whale is randomly selected as the target; the current position update formula is X(t+1) = X rand (t)-A D, D=|C X rand The formula is X(t) - X(t)|, which moves the whale away from the random target and allows it to explore a wide range of solutions to avoid local optima. If |A| < 1, then local exploitation (surrounding the prey) is performed, which means targeting the currently globally optimal individual whale. The current position update formula is X(t+1) = X best (t)-A·D',D'=|C·X best (t)-X(t)|, to make the whale move closer to the currently known optimal solution, and to perform a local fine search.

[0060] When p≥0.5, a bubble net attack (spiral update) is performed, which also targets the current globally optimal individual; the current position update formula is X(t+1)=D'' cos(2π L)+X best (t), D''=|X best (t)-X(t)|, to simulate the process of a whale spiraling towards its prey, searching around the optimal solution; where b is the spiral constant, controlling the tightness of the spiral, such as taking b=1; L is a random parameter, and the position on the spiral path is determined by a uniformly distributed random number in L∈[-1,1]. The term is an exponentially growing term that controls the spiral radius; cos(2πL) is a cosine function that generates periodic oscillations; the combined effect is to generate a logarithmic spiral path, that is, to spirally approach the optimal solution as the center, b controls the spiral shape, and L is a random number.

[0061] Where A is a coefficient, and the calculation formula is: A=2 r1- r1 is a random number uniformly distributed in the range [0,1]; since r1 is a random number, the range of values ​​for A is [- , When a is large, the absolute value of A is large; when a is small, the absolute value of A is small; |A| (the absolute value of A) determines the algorithm's exploration capability, that is, when |A|≥1, global exploration is forced; even if it should have been local exploration (p<0.5), if |A|≥1, it will switch to global exploration to ensure that the algorithm does not get trapped in local optima too early; when |A|<1, local exploration is allowed, and a fine search is performed around the current optimal solution.

[0062] The convergence criterion of the improved whale optimization algorithm is mainly based on the population state and the stability of the optimal solution, that is, the population state reaches the maximum number of iterations, t> It is the most direct and reliable stopping condition, preventing infinite loops and ensuring real-time performance. The maximum number of iterations can be adjusted according to the problem complexity and computational budget. The stability of the optimal solution can be determined based on the optimal solution stability criterion. For example, if the fitness improvement of the global optimal solution for K consecutive generations is less than a minimum threshold ε, i.e., |fitness_best(t) - fitness_best(tk)| < ε, k = 1, 2, ..., K, it indicates that the algorithm has not found a better solution and the quality of the solution has stabilized. For example, if ε = e -5 K=10.

[0063] The improved whale optimization algorithm calculates the fitness of a whale after it moves to a new location; if the fitness of the new location is better than the fitness of the original location, the new location replaces the original location; otherwise, the original location remains unchanged; after all individuals have been updated, the global optimal solution X for the entire population is checked and updated. best Check if the convergence condition is met (e.g., reaching the maximum number of iterations). (or the optimal solution shows no significant improvement over multiple generations).

[0064] If convergence is not achieved, the next iteration begins; if convergence is achieved, the algorithm terminates, and the current global optimal solution X is obtained. best The candidate optimal N value found by the improved whale optimization algorithm is the target time domain.

[0065] The Adaptive Differential Evolutionary Algorithm searches for the optimal solution by simulating the evolutionary process of mutation, crossover, and selection. Specifically, it first randomly generates a group of individuals, each representing a candidate predicted time-domain value N, forming an initial population, and evaluates the fitness of each individual in the initial population. It then performs iterative iterations, calculating and updating the population diversity for each individual in each iteration, such as calculating the current diversity index based on the distribution of individuals in the population. The calculation process can be represented as follows: =population_std / search_range, where population_std is the standard deviation of the population, reflecting the dispersion of individuals (predicted time-domain N value) in the current population, and search_range is the search range, i.e., the range of values ​​for the predicted time-domain N value, used to normalize the standard deviation and map the diversity index to the range of 0 to 1; based on the calculated diversity index, one of the three mutation strategies is selected: If the diversity index is >0.3, indicating high diversity, then an exploration-oriented strategy is selected to broaden the search scope; the exploration-oriented strategy specifically includes: V_i=X_r1+F (X_r2-X_r3) Among them, the new candidate individuals generated after the mutation of V_i are the new predicted time-domain N values); X_r1, X_r2, and X_r3 are different candidate individuals randomly selected from the population, and r1, r2, and r3 are random indices representing other candidate N values ​​in the population.

[0066] 0.1 Diversity Indicators A value of 0.3 indicates diversity; therefore, a balanced strategy is chosen to balance individual and global information. The specific details of the balanced strategy are as follows: V_i=X_i+F (X best -X_i)+F (X_r1-X_r2) Where X_i is the i-th individual to be updated in the current iteration, i.e., the original candidate N value; X best N represents the individual with the best fitness in the current population, i.e., the "best" predicted time-domain value at present; F is the scaling factor, which is a parameter that controls the magnitude of variation. It is usually a constant between 0 and 1 and is used to adjust the degree of influence of differences between individuals on new individuals.

[0067] If the diversity index is <0.1, indicating low diversity, a development-oriented strategy is selected, focusing on local fine-grained search. The development-oriented strategy specifically involves: V_i=X best +F (X_r1-X_r2) Furthermore, the variation factor F is adaptively calculated; based on the current relative fitness level of an individual in the population, the formula is used. Calculate an adaptive variation factor for it. High-fit individuals have a larger F-value (≈0.9), enhancing their exploration capabilities; low-fit individuals have a smaller F-value (≈0.1), allowing for fine-tuning; and medium-fit individuals have a moderate F-value, balancing exploration and development.

[0068] Based on the chosen strategy and The process involves performing a mutation operation on the current individual to generate a trial individual N_trial; then, performing a crossover operation between the trial individual and the original individual to fuse their features and finally determine the trial individual; calculating the fitness of the trial individual N_trial; comparing the fitness of the trial individual with that of the original individual, if the fitness of N_trial is better, then replacing the original individual with N_trial in the next generation of the population, otherwise, retaining the original individual in the next generation of the population; after all individuals have been updated, checking and updating the global optimal solution of the entire population.

[0069] Check if the convergence condition is met (e.g., reaching the maximum number of iterations, or the improvement rate of the optimal solution is too low, or the success mutation rate continues to decrease); if not converged, start the next iteration; if converged, the algorithm ends and outputs the current global optimal solution as the candidate optimal N value found by the adaptive differential evolution algorithm, i.e., the target time domain.

[0070] Wherein, the fitness function is fitness(N) = 1 / (1+J(N)); J(N) is the comprehensive performance index, the smaller the J(N) value, the better the performance, J(N) = Jlateral + Jheading; Jlateral is the lateral error index, Jlateral = α1×RMSE(e_y) + α2×max(|e_y|) + α3×ISE(e_y), RMSE(e_y) is the root mean square error of the lateral error, e_y is the lateral error, T is the total data duration / total steps, ISE(|e_y|) is the integral of the square of the lateral error (Δt = 0.05s); Jheading is the heading error index, Jheading = ×RMSE(e_φ)+ ×max(|e_φ|); RMSE(e_φ) is the root mean square error of the longitudinal error (such as angle / velocity error), e_φ is the longitudinal error, max(|e_φ|) is the maximum absolute value of the longitudinal error, and α1, α2 and α3 are the weight coefficients of the lateral error (the weight allocation is different in different scenarios, such as the weight of road edge cleaning / vehicle water filling scenarios). and The weighting coefficients for longitudinal errors (weighting allocation under different scenarios).

[0071] For example, an improved whale optimization algorithm is used when cleaning road edges. =0.3, =0.7; Adaptive differential evolution algorithm is used in high-speed environments such as when vehicles are adding water. =0.7, =0.3.

[0072] S203. Based on the optimization algorithm, the prediction time domain of Model Predictive Control (MPC) is calculated to obtain the target time domain.

[0073] One possible approach is to generate at least two candidate predicted time-domain values ​​through an optimization algorithm, perform MPC simulation on each candidate predicted time-domain value and calculate the fitness function, and then perform elite selection and population update based on the fitness function results until the convergence condition is met to obtain the target time-domain value.

[0074] Among them, optimization algorithms refer to intelligent algorithms used for time-domain optimization of MPC prediction, including improved whale optimization algorithm and adaptive differential evolution algorithm, which achieve efficient optimization through multi-strategy collaboration.

[0075] Model predictive control is an advanced control strategy that constructs a vehicle dynamics model and solves a finite-time open-loop optimization problem in each control cycle, achieving rolling optimization by applying only the first element of the control sequence.

[0076] The prediction time domain refers to the time step (denoted as N) used in MPC to predict the future state of the vehicle. Its value directly affects the control accuracy, computational efficiency, and system stability.

[0077] The target time domain refers to the optimal predicted time domain value that meets the convergence condition after iterative calculation through optimization algorithms. It is an important parameter for MPC trajectory tracking control.

[0078] Candidate prediction time-domain values ​​refer to alternative prediction time-domain values ​​generated by the optimization algorithm, and their performance needs to be verified through MPC simulation.

[0079] MPC simulation is a virtual test that simulates the vehicle trajectory tracking process based on the vehicle dynamics model and candidate predicted time-domain values, and is used to evaluate the actual control effect of the candidate values.

[0080] The fitness function is an evaluation function that quantifies the merits of candidate prediction time-domain values, comprehensively considering performance such as tracking accuracy, energy economy, and control smoothness.

[0081] Elite selection is a strategy that selects the candidate value with the best fitness from the population and retains it, ensuring that high-quality solutions are not eliminated.

[0082] Population update is based on the results of elite selection. It generates a new set of candidate prediction time-domain values ​​through operations such as mutation and crossover, thereby achieving iterative evolution of the population.

[0083] Convergence criteria are the standards for determining when an optimization algorithm stops iterating. These include the maximum number of iterations, the optimal solution stability criterion, and the improvement rate threshold, ensuring that the algorithm converges efficiently to the optimal solution.

[0084] For example, by improving the whale optimization algorithm with strategies such as global exploration and local exploitation, or by using the adaptive differential evolution algorithm through population initialization and mutation operations, at least two candidate predicted time-domain values ​​can be generated to construct an initial optimization population. Then, MPC simulation is run for each candidate predicted time-domain value, simulating the trajectory tracking process over the next N steps based on the vehicle dynamics model, reconstructing the actual control scenario corresponding to that candidate value. The fitness function is then substituted into the MPC simulation results, and the fitness of each candidate value is calculated by comprehensively considering indicators such as lateral tracking error, energy consumption, and control smoothness to quantify its performance. Further, fine-tuning is performed... The algorithm selects the candidate predicted time domain value with the best fitness from the current population and retains it for the next iteration to avoid losing high-quality solutions. Then, based on the selected elite individuals, new candidate values ​​are generated through mutation, crossover and other operations of the algorithm to update the population structure and achieve dynamic adjustment and optimization of the search range. After each iteration, it is judged whether the convergence conditions are met, such as reaching the maximum number of iterations, the optimal solution being stable for multiple generations, and the improvement rate being lower than the threshold. If not met, the simulation, calculation, selection and update process is repeated. If met, the iteration stops. Finally, the optimal candidate predicted time domain value that meets the convergence conditions is determined as the target time domain.

[0085] For example, the improved whale optimization algorithm dynamically switches between three stages: global exploration, local exploitation, and bubble net attack. By balancing the search capability through an adaptive shrinkage factor, it ultimately outputs the optimal prediction time domain N that adapts to the MPC of new energy sanitation vehicles. The specific steps are as follows: The initialization phase includes setting core parameters, generating the initial population, and initial fitness evaluation; setting core parameters includes the maximum number of iterations. =50 (adapted to vehicle computing power), population size (default 20 candidate N values), search range of N (dynamically adjusted according to the operation mode, [15,40] for edge cleaning, [30,50] for water navigation), diversity influence coefficient λ=0.3, convergence threshold ε=e -5 The initial population generation involves randomly generating 20 integers within the range N, forming the initial whale population X(t) = {X1(t), X2(t), ..., X20(t)}, where Xi(t) represents the i-th candidate N value in the t-th iteration. Initial fitness evaluation involves substituting each initial candidate N value into the MPC simulation to calculate the fitness function fitness(Ni) = 1 / (1+J(Ni)), and selecting the initial global optimum X. best (The candidate N value with the highest fitness).

[0086] The iterative optimization phase includes calculating key parameters, dynamically selecting optimization strategies, re-evaluating fitness, and updating the population; specifically, calculating key parameters includes population diversity indicators. That is, calculated by the ratio of the population standard deviation to the search range (range 0-1, 0 indicates population aggregation, 1 indicates population dispersion); adaptive contraction factor. (t is the current iteration number, which decreases linearly from 1 to 0 as iterations progress); Explore the balance coefficient (r1 is a uniformly distributed random number in the range [0,1], and the value of A is in the range [-a,a]); the strategy selection switch p generates uniformly distributed random numbers in the range [0,1], which is used to switch optimization strategies. For example, dynamically selecting optimization strategies includes: Strategy 1: Global exploration (p<0.5 and |A|≥1): Randomly select an individual Xrand(t) from the population (not the current optimal solution); Calculate distance (C=2) r2, where r2 is a uniformly random number in the range [0,1], controlling the search step size); Update individual position: X(t+1) = Xrand(t) A D enables a wide-ranging exploration in the opposite direction of random individuals, avoiding getting trapped in local optima.

[0087] Strategy 2: Localized development (p<0.5 and |A|<1): Calculate the distance D′=|C between the current individual and the global optimal solution. X best (t) X(t)∣; Update individual position: X(t+1) = X best (t) A D′ performs a small-scale fine-grained search around the global optimal solution to improve the accuracy of the solution.

[0088] Strategy 3: Bubble Web Attack (p≥0.5): Calculate the distance D′′ between the current individual and the global optimal solution. best (t) X(t)∣; Set the helix constant b=1 (to control the tightness of the helix) and the random parameters. (Determines the location of the spiral path); Update individual position: X(t+1)=D′′ cos(2π L)+X best (t) simulates the spiraling motion of a whale surrounding its prey, converging toward the optimal solution.

[0089] Fitness reassessment involves recalculating the fitness (Ni) for each individual X(t+1) in the updated population.

[0090] Population updates involve retaining individuals with higher fitness; if a new individual has a higher fitness than the original individual, the original individual is replaced, and the global optimum X is updated. best (t+1) (Takes the N value corresponding to the maximum fitness of the current population).

[0091] When the current iteration number t > Tmax (reaching the maximum iteration number), or the fitness improvement amount |fitnessbest(t) of the global optimal solution for 10 consecutive generations is satisfied. If fitnessbest(t−k)∣<ε (k=1,2,...,10, the solution quality is stable), then stop the iteration and set the final global optimal solution X. best The target time domain N of MPC is used to adapt to the current working conditions.

[0092] The adaptive differential evolution algorithm achieves efficient search of the time domain N for MPC prediction through dynamic selection of multiple mutation strategies and adaptive parameter adjustment, balancing exploration breadth and development accuracy. The specific steps are as follows: The initialization phase includes setting core parameters, generating the initial population, and initial fitness evaluation. The core parameters are: maximum number of iterations Tmax=30 (to adapt to the real-time requirements of vehicle-mounted systems), population size T=20 (number of candidate N values), search range of N [5,30] (basic range, which can be fine-tuned according to the operation mode), and upper and lower bounds of the mutation factor. =0.1、 =0.9, improvement rate threshold εimprove=0.001, success mutation rate threshold εsuccess=0.05, statistical window M=10 (used to judge the success mutation rate); the initial population is generated by randomly generating 20 integers in the range [5,30] to form the initial population X(0)={X1(0),X2(0),...,X20(0)}; The initial fitness evaluation involves calculating the fitness (Ni) = 1 / (1+J(Ni)) for each initial candidate N value using MPC simulation, recording the minimum fitness fmin and maximum fitness fmax, and determining the initial global optimum X. best (0).

[0093] The iterative optimization phase includes calculating population state indices, individual mutations, crossover and selection, and population and optimal solution updates; calculating population state indices, i.e., population diversity; and calculating the historical improvement rate, which is the relative improvement rate (fbest(t)) between the previous generation and the current generation's global optimal solution. fbest(t 1)) / fbest(t 1) Successful mutation rate is the proportion of new individuals that replace original individuals through mutation in the most recent M generations (number of successful replacements / total number of individuals). Individual mutation is performed on each individual Xi(t).

[0094] In the crossover and selection process, the crossover operation uses a binomial crossover to generate experimental individuals N_trial; the greedy selection process calculates the fitness of experimental individuals N_trial. If fitness(N_trial) > fitness(Xi(t)), then the new population accepts N_trial; otherwise, the original individual Xi(t) is retained.

[0095] Population update and optimal solution update involve updating the population X(t+1) based on the greedy selection result, recalculating fmin and fmax, and updating the global optimal solution X. best (t+1).

[0096] S204. Based on the target time domain and MPC, obtain the target time domain parameters of the target vehicle, and perform path control based on the target time domain parameters.

[0097] Among them, the target time domain parameters are a set of MPC control parameters that are dynamically adjusted in combination with the operation mode, including the target time domain, weight matrix, safety boundary, performance constraints, etc.

[0098] Path control is a control process that converts target time-domain parameters into actual actuator commands and drives the target vehicle to accurately follow a preset path, while balancing accuracy, energy consumption, and comfort.

[0099] For example, the target time domain N output by the multimodal optimization module is first input to the MPC control module as the basis for subsequent parameter adaptation and optimization. Then, the MPC controller dynamically adapts the target time domain parameters according to the current working mode of the target vehicle (such as edge cleaning, water navigation). In edge cleaning mode, the lateral error weight is increased and N is limited to the range of [15,40] to ensure edge-fitting accuracy. In water navigation mode, the lateral error weight is reduced and N is limited to the range of [30,50] to balance safety and comfort. Subsequently, the MPC's state predictor predicts the state sequence of the next N steps based on the target time domain N, the vehicle dynamics model, and the current vehicle state (position, speed, etc.). The rolling time domain actuator solves the finite time domain optimization. The problem is solved by generating an optimal control sequence [u(t), u(t+1), ..., u(t+N-1)]. The first control variable u(t) in the optimal control sequence is then converted into actual actuator commands (such as steering angle, throttle / brake), while ensuring that the rate of change of the control variable is within the allowable range of the mechanical system. Subsequently, the actuators of the target vehicle respond to the commands to complete steering, acceleration or braking actions, realize the tracking control of the preset path, and meet the requirements of cleaning, navigation and other tasks. Finally, the environmental perception module integrates data from multiple sensors such as GPS and IMU to update the vehicle status in real time and evaluate the control effect. The feedback information is transmitted to the optimization module to provide support for the dynamic adjustment of subsequent time-domain parameters, forming a closed-loop control.

[0100] Based on the above, the MPC-based vehicle adaptive path control method achieves real-time synchronous perception of the environment and its own state by acquiring the vehicle state data of the target vehicle. This provides objective and quantitative input for all subsequent adaptive decisions, ensuring that control decisions closely align with actual operating conditions. Furthermore, based on the vehicle state data, the corresponding optimization algorithm for the target vehicle is determined, achieving dynamic matching between the control strategy and the driving scenario. Then, based on the optimization algorithm, the prediction time domain of the Model Predictive Control (MPC) is calculated to obtain the target time domain, ensuring on-demand allocation and efficient utilization of computing resources while maintaining or even improving control performance. Finally, based on the target time domain and MPC, the target time domain parameters of the target vehicle are obtained, and path control is performed based on these parameters, achieving high-performance real-time control under parameter adaptation. This scheme determines the corresponding optimization algorithm for the target vehicle based on vehicle state data, enabling dynamic selection or switching of the most suitable optimization algorithm for the current scenario, overcoming the limitations of the traditional one-size-fits-all approach. Furthermore, by introducing the target time domain, online real-time optimization of the prediction time domain is ensured, ultimately achieving accurate path control.

[0101] The above text combined Figures 1 to 2 The vehicle adaptive path control method based on MPC provided in the embodiments of this application has been described in detail. The apparatus and equipment provided in the embodiments of this application will be described below with reference to the accompanying drawings.

[0102] This application also provides a vehicle adaptive path control device based on MPC, such as... Figure 3 As shown in the figure, this is a structural schematic diagram of a vehicle adaptive path control device 300 based on MPC provided in an embodiment of this application. The device includes: The acquisition module 301 is used to acquire vehicle status data of the target vehicle; The algorithm determination module 302 is used to determine the optimization algorithm corresponding to the target vehicle based on vehicle state data; The time domain determination module 303 is used to calculate the prediction time domain of Model Predictive Control (MPC) based on an optimization algorithm to obtain the target time domain; The path control module 304 is used to obtain the target time domain parameters of the target vehicle based on the target time domain and MPC, and to perform path control based on the target time domain parameters.

[0103] In some possible implementations, the algorithm determination module 302 is specifically used for: Based on vehicle status data, determine the driving mode of the target vehicle; Based on the driving mode, the corresponding optimization algorithm for the target vehicle is determined.

[0104] In some possible implementations, the algorithm determination module 302 is specifically used for: When the operation mode is identified as the first identifier and the path curvature meets the preset first operation conditions, the driving mode is determined to be the first mode; wherein, the first mode is the situation where the target vehicle performs the cleaning task. When the operation mode is identified as the second identifier and the vehicle speed is higher than the preset first speed threshold, the driving mode is determined to be the second mode; wherein, the second mode is the situation where the target vehicle performs the driving task.

[0105] In some possible implementations, the algorithm determination module 302 is specifically used for: When the driving mode is the first mode, the optimization algorithm corresponding to the target vehicle is determined to be the improved whale optimization algorithm; When the driving mode is the second mode, the optimization algorithm corresponding to the target vehicle is determined to be the adaptive differential evolution algorithm.

[0106] In some possible implementations, the algorithm determination module 302 is specifically used for: The shrinkage factor of the improved whale optimization algorithm is determined in the following way:

[0107] in, It is a contraction factor; This represents the current iteration number; This represents the maximum number of iterations. As an indicator of population diversity; This represents the diversity impact coefficient.

[0108] In some possible implementations, the algorithm determination module 302 is specifically used for: The mutation factor of the adaptive differential evolution algorithm is determined in the following way:

[0109] in, It serves as an individual identifier to distinguish different individuals; Let be the variation factor corresponding to the i-th individual; This is the lower bound of the variable factor; This is the upper bound of the variable factor; Let i be the fitness of the i-th individual; The minimum fitness value of all individuals in the population. This represents the maximum fitness value among all individuals in the population.

[0110] In some possible implementations, the time-domain determination module 303 is specifically used for: At least two candidate time-domain prediction values ​​are generated by optimizing the algorithm. MPC simulation is performed on each candidate time-domain prediction value and the fitness function is calculated. Elite selection and population update are performed based on the fitness function results until the convergence condition is met, thus obtaining the target time domain.

[0111] The MPC-based vehicle adaptive path control device according to the embodiments of this application can correspond to the execution of the method described in the embodiments of this application, and the other operations and / or functions of each module / unit of the MPC-based vehicle adaptive path control device are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.

[0112] This application also provides a computing device. For example... Figure 4 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 400 includes a bus 401, a processor 402, a communication interface 403, and a memory 404. The processor 402, the memory 404, and the communication interface 403 communicate with each other via the bus 401.

[0113] Bus 401 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0114] Processor 402 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0115] The communication interface 403 is used for communication with external devices. For example, if the computing device is a first switch, the communication interface 403 can be used for communication between the first switch and a first user terminal, or for communication between the first switch and a second switch.

[0116] Memory 404 may include volatile memory, such as random access memory (RAM). Memory 404 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0117] The memory 404 stores executable code, and the processor 402 executes the executable code to perform the aforementioned MPC-based vehicle adaptive path control method.

[0118] Specifically, in achieving Figure 3 In the case of the illustrated embodiment, and Figure 3 When the modules or units of the MPC-based vehicle adaptive path control device described in the embodiments are implemented in software, the execution... Figure 3The software or program code required for the functions of each module / unit can be partially or wholly stored in memory 404. Processor 402 executes the program code corresponding to each unit stored in memory 404 to execute the aforementioned MPC-based vehicle adaptive path control method.

[0119] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium capable of being stored by a computing device, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the aforementioned MPC-based vehicle adaptive path control method.

[0120] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.

[0121] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0122] When the computer program product is executed by a computer, the computer executes any of the aforementioned MPC-based vehicle adaptive path control methods. The computer program product can be a software installation package; when any of the aforementioned MPC-based vehicle adaptive path control methods needs to be used, the computer program product can be downloaded and executed on the computer.

[0123] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0124] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.

Claims

1. A vehicle adaptive path control method based on MPC, characterized in that, The method includes: Obtain vehicle status data for the target vehicle; Based on the vehicle status data, determine the optimization algorithm corresponding to the target vehicle; Based on the optimization algorithm, the prediction time domain of Model Predictive Control (MPC) is calculated to obtain the target time domain; Based on the target time domain and MPC, the target time domain parameters of the target vehicle are obtained, and path control is performed based on the target time domain parameters.

2. The method according to claim 1, characterized in that, The optimization algorithm for determining the target vehicle based on the vehicle status data includes: Based on the vehicle status data, determine the driving mode of the target vehicle; Based on the driving mode, the optimization algorithm corresponding to the target vehicle is determined.

3. The method according to claim 2, characterized in that, The vehicle status data includes vehicle speed, path curvature, and operation mode identifier; Based on the vehicle status data, the driving mode of the target vehicle is determined, including: When the operation mode is identified as the first identifier and the path curvature meets the preset first operation condition, the driving mode is determined to be the first mode; wherein, the first mode is the situation where the target vehicle performs the cleaning task; When the operation mode is identified as the second identifier and the vehicle speed is higher than the preset first speed threshold, the driving mode is determined to be the second mode; wherein, the second mode is the situation in which the target vehicle performs the driving task.

4. The method according to claim 3, characterized in that, The optimization algorithm for determining the target vehicle based on the driving mode includes: When the driving mode is the first mode, the optimization algorithm corresponding to the target vehicle is determined to be the improved whale optimization algorithm; When the driving mode is the second mode, the optimization algorithm corresponding to the target vehicle is determined to be the adaptive differential evolution algorithm.

5. The method according to claim 4, characterized in that, The shrinkage factor of the improved whale optimization algorithm is determined in the following way: in, It is a contraction factor; This represents the current iteration number; This represents the maximum number of iterations. As an indicator of population diversity; This represents the diversity impact coefficient.

6. The method according to claim 4, characterized in that, The mutation factor of the adaptive differential evolution algorithm is determined in the following way: in, It serves as an individual identifier to distinguish different individuals; Let be the variation factor corresponding to the i-th individual; This is the lower bound of the variable factor; This is the upper bound of the variable factor; Let i be the fitness of the i-th individual; The minimum fitness value of all individuals in the population. This represents the maximum fitness value among all individuals in the population.

7. The method according to claim 1, characterized in that, The step of calculating the prediction time domain of Model Predictive Control (MPC) based on the optimization algorithm to obtain the target time domain includes: At least two candidate time-domain prediction values ​​are generated using the optimization algorithm. MPC simulation is performed on each candidate time-domain prediction value, and the fitness function is calculated. Elite selection and population update are performed based on the fitness function results until the convergence condition is met, thus obtaining the target time domain.

8. A vehicle adaptive path control device based on MPC, characterized in that, The device includes: The acquisition module is used to acquire vehicle status data of the target vehicle; The algorithm determination module is used to determine the optimization algorithm corresponding to the target vehicle based on the vehicle state data. The time-domain determination module is used to calculate the prediction time domain of Model Predictive Control (MPC) based on the optimization algorithm to obtain the target time domain. The path control module is used to obtain the target time domain parameters of the target vehicle based on the target time domain and MPC, and to perform path control based on the target time domain parameters.

9. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 7.

Images