Human-machine co-driving networked vehicle platoon cooperative control method and system based on fuzzy rules

Abstract

The application discloses a man-machine co-driving networked vehicle platoon cooperative control method and system based on fuzzy rules, which comprises the following steps: taking the normalized values of lateral position error, heading angle error and longitudinal speed error as three-dimensional fuzzy inputs, dynamically calculating man-machine control weights through a Mamdani inference mechanism, and weighting and fusing automatic driving control inputs and operator control inputs; aiming at the time-varying switching problem of a communication topology, designing a distributed observer containing a networked vehicle observation correction term and a neighbor consistency correction term, establishing an explicit gain design condition to ensure exponential convergence of an estimation error; introducing an extensible preset performance function containing a basic contraction term, an obstacle-induced expansion term and a curvature-induced expansion term, combining a CLF-CBF-QP safety filter to online correct a nominal control input, and realizing cooperative guarantee of platoon keeping and safety obstacle avoidance.

Description

Technical Field

[0001] This invention relates to the field of connected vehicle cluster collaborative control and human-machine sharing control technology, and in particular to a method and system for human-machine co-driving connected vehicle platooning collaborative control based on fuzzy rules. Background Technology

[0002] In multi-vehicle platooning control with a connected vehicle-follower architecture, the complete state of the lead connected vehicle is usually not directly available to the following connected vehicles, which can only rely on local neighbor information for state estimation. Meanwhile, in complex operating environments, wireless communication links are susceptible to terrain obstruction and interference, and the communication topology of connected vehicles exhibits time-varying characteristics. Existing distributed observers based on fixed topology assumptions lack explicit gain design conditions and cannot guarantee the convergence of estimation errors under topology switching conditions.

[0003] Furthermore, when human operators are included in the control loop of the navigation connected vehicle, the operator and the autonomous driving system have independent control intentions. Traditional fixed-weight fusion methods cannot dynamically adjust control rights according to real-time operating conditions, while existing game theory-based dynamic permission allocation methods have high computational complexity and poor real-time performance, making it difficult to meet the control requirements of highly maneuverable connected vehicle swarms. In terms of safety constraints, when the formation performs obstacle avoidance maneuvers, it needs to temporarily deviate from the reference trajectory. Traditional fixed performance boundary constraints cannot accommodate the transient errors required for obstacle avoidance, creating an inherent contradiction with the safety obstacle avoidance requirements and easily leading to an infeasible control optimization problem. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a human-machine co-driving connected vehicle platooning collaborative control method based on fuzzy rules. This invention achieves integrated control of human-machine collaboration, safe obstacle avoidance, and platooning maintenance in complex dynamic environments by dynamically allocating human-machine control weights using three-dimensional fuzzy rules, estimating connected vehicle states using a topology distributed observer, and correcting control inputs online using an expandable preset performance safety filter.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control method, comprising the following steps: A human-machine co-driving control model for a leading connected vehicle is established. The problem of control allocation between the operator and the autonomous driving system is modeled as a weighted fusion of control inputs. The optimal control inputs for autonomous driving and the driver are solved independently, and the fusion weights are determined based on three-dimensional fuzzy rules to obtain the fused control input of the connected vehicle. The lead vehicle executes fused control inputs, and each following vehicle runs a distributed observer. The distributed observer includes a vehicle observation correction term and a neighbor consistency correction term to update the vehicle state estimate and obtain the vehicle state estimate result. The observer gain design conditions are established based on the common Lyapunov function method to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology. Each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced expansion term, and a curvature-induced expansion term. Each following connected vehicle executes the modified control inputs to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

[0006] A further technical solution is to determine the fusion weights based on three-dimensional fuzzy rules, specifically as follows: First, calculate the lateral position error of the connected vehicle. Heading angle error and longitudinal velocity error And after normalization, it is expressed as: ; in, , and These are the normalized values ​​for lateral position error, heading angle error, and longitudinal velocity error, respectively. This is the upper bound of the lateral position error. This is the upper bound of the heading angle error. This is the upper bound for the longitudinal speed error; if any normalized value reaches or exceeds 1, then autonomous driving will take full control. Secondly, the membership degrees of the small, medium, and large triangles are calculated for the normalized error values, and expressed as follows:

[0007]

[0008] ; in, , , These are the membership functions for small, medium, and large triangles, respectively. This represents the normalized value of the lateral position error, the normalized value of the heading angle error, or the normalized value of the longitudinal velocity error. A three-dimensional Mamdani rule base is established, the activation intensity of each rule is calculated through product reasoning, the centroid method is used to defuzzify and obtain the autonomous driving weight coefficients, and the autonomous driving weight coefficients are subjected to amplitude limiting processing.

[0009] A further technical solution is that the fusion control input is represented as: ; in, To integrate control inputs, For autonomous driving weighting coefficients, For driver weighting coefficients, This is the optimal control input for autonomous driving. This provides the optimal control input for the driver.

[0010] A further technical solution is that the dynamic equation of the distributed observer is expressed as: ; in, ; In the formula, To follow connected vehicles The constantly updated estimate of the connected vehicle's status. To follow connected vehicles Local estimates of the status of the connected vehicle at any given time. To lead the way in connected vehicles The actual state at any given moment Follow the connected car for your neighbor exist Local estimates of the status of the connected vehicle at any given time. The sampling period is For drift term, For comprehensive correction items, To enhance the observation gain of connected vehicles in the navigation network, For neighbor coupling gain, To follow connected vehicles exist The status of the communication link between the vehicle and the navigation network at all times. To follow connected vehicles exist Connected vehicles at all times and following The status of the communication link between them In order to keep up with the total number of connected vehicles, For connected vehicle observation and correction items, This is a neighbor consistency correction item.

[0011] A further technical solution is to establish the observer gain design conditions based on the common Lyapunov function method, expressed as:

[0012] in, For topology The corresponding Laplace matrix, It is the Lipschitz constant; Under the condition of satisfying the gain design, the estimation error satisfies: ; in, , , It is a constant that depends on the observer gain and network parameters.

[0013] A further technical solution is that the dynamic boundary of the scalable preset performance function is: ; in, Basic contraction term, For obstacle-inducing extensions, This is a curvature-induced extension term.

[0014] A further technical solution is to find the cost function for solving the local MPC optimization problem as follows: ; in, For formation tracking error, The expected formation position is calculated based on the estimated state of connected vehicles. For the neighbor consensus item, For obstacle avoidance, For communication maintenance items, This is a saturation penalty term.

[0015] Secondly, this invention provides a fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control system, comprising the following modules: The human-machine co-driving fusion module is configured to: establish a human-machine co-driving control model for the connected vehicle, model the control rights allocation problem between the operator and the autonomous driving system as a weighted fusion of control inputs, solve the optimal control inputs for autonomous driving and driver independently, and determine the fusion weights based on three-dimensional fuzzy rules to obtain the fusion control inputs for the connected vehicle; The distributed state observation module is configured as follows: the lead connected vehicle executes the fusion control input, and each following connected vehicle runs a distributed observer. The distributed observer includes a connected vehicle observation correction term and a neighbor consistency correction term to update the connected vehicle state estimate and obtain the connected vehicle state estimate result. The observer gain design conditions are established based on the common Lyapunov function method to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology. The observer gain design module is configured such that each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced extension term, and a curvature-induced extension term. The execution module is configured to: each following connected vehicle executes the modified control inputs to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention introduces a fuzzy inference mechanism into the human-machine co-driving control of a leading connected vehicle. Utilizing a 27-rule base, it achieves real-time mapping from error states to control permissions, eliminating the need for complex iterative solutions required by game theory methods. This results in low computational complexity and fast real-time response, effectively solving the technical problems of fixed-weight fusion's inability to adapt to real-time conditions and the insufficient real-time performance of game theory methods. It enables smooth dynamic switching of human-machine control between stable driving and high-risk maneuvers. Furthermore, by designing a distributed observer and using the common Lyapunov function method, it derives an explicit gain design inequality under the switching topology. Mathematically, it rigorously proves the exponential convergence of the connected vehicle state estimation error under time-varying communication link interruption and reconstruction conditions, overcoming the limitations of existing methods that rely on fixed topology assumptions.

[0017] This invention introduces an extensible preset performance function, which automatically and dynamically widens the tracking performance envelope when the connected vehicle approaches an obstacle or performs a sharp turn, and automatically restores the exponentially contracted envelope after obstacle avoidance. Combined with an online quadratic programming safety filter composed of higher-order control obstacle function constraints and control Lyapunov function constraints, this invention fundamentally solves the inherent contradiction between traditional fixed performance boundaries and safe obstacle avoidance requirements, and maintains the preset performance of formation tracking while ensuring the forward invariance of the formal safety set. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 This is a schematic diagram of the process framework of the human-machine co-driving connected vehicle platooning collaborative control method based on fuzzy rules of the present invention; Figure 2 This invention provides a fuzzy rule-based formation motion trajectory diagram of a connected vehicle cluster in an obstacle environment. Figure 3 This is a graph showing the change in adaptive permission allocation for connected vehicles according to the present invention. Figure 4 This is a graph showing the estimation error of the distributed observer in this invention. Figure 5 This invention provides a preset performance envelope and composite error diagram, wherein (a) represents the following connected vehicle. Preset performance envelope and composite error diagram, (b) for following connected vehicles Preset performance envelope and composite error plot; Figure 6 This is a diagram illustrating the formation tracking error of the present invention. Detailed Implementation The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0022] Terminology Explanation: Connected vehicle-follower architecture: refers to a multi-connected vehicle formation control structure in which a lead connected vehicle determines the overall movement trajectory, and several follow connected vehicles adjust their own movement according to their relative geometric relationship with the connected vehicle to maintain the formation configuration; Human-machine co-driving: refers to a control mode in which human operators and autonomous driving systems are jointly incorporated into the same connected vehicle control loop, and control commands are jointly generated through weighted fusion. Model predictive control (MPC) refers to a control method that uses a system dynamics model to perform rolling optimization in the prediction time domain within each control cycle to find the optimal control sequence. Distributed observer: refers to a state estimator deployed locally in each following connected vehicle and reconstructing the state of the connected vehicle by interacting with information from neighboring connected vehicles; Preset Performance Function (PPF): refers to a function that applies a time-varying upper bound constraint to the tracking error and ensures that the error converges within a specified performance envelope; Control Lyapunov function (CLF): refers to a scalar function used to verify system stability and generate convergence constraints; Control Barrier Function (CBF): A scalar function used to describe the forward invariance of safety sets and generate safety constraints; Quadratic Programming (QP): refers to a convex optimization problem with a quadratic objective function and linear inequalities as constraints, which can be solved in real time within each control cycle.

[0023] Example 1 like Figure 1 As shown, this embodiment provides a fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control method, including the following steps: S101: Establish a human-machine co-driving control model for the connected vehicle, model the control rights allocation problem between the operator and the autonomous driving system as a weighted fusion of control inputs, solve the optimal control inputs of autonomous driving and driver independently, and determine the fusion weights based on three-dimensional fuzzy rules to obtain the fusion control input of the connected vehicle.

[0024] In step S101, a connected vehicle dynamics model is first established to obtain system state information.

[0025] Each connected vehicle is described using a dynamic bicycle model. The state vector is defined as follows: ,in and For connected vehicles Vertical and horizontal positions in the global coordinate system For longitudinal velocity, For lateral velocity, For heading angle, This represents the yaw rate. The control input vector is defined as follows: ,in This refers to the front wheel steering angle. This is the longitudinal acceleration.

[0026] The continuous-time dynamic equation is Among them, the drift term and input matrix Based on vehicle mass yaw moment of inertia Distance from front and rear axles to the center of gravity and and front and rear tire lateral stiffness and The calculations determined that the key dynamic parameters are:

[0027]

[0028] By sampling period The Euler discretization method with step size transforms the continuous-time model into a discrete-time dynamic model:

[0029] in The sampling period is defined as . The Euler discretization method facilitates real-time solution in digital controllers, providing a unified predictive model foundation for subsequent distributed observers and model predictive controllers following connected vehicles.

[0030] Optimal control input for autonomous driving This refers to the first instruction of the optimal control sequence obtained by the model predictive controller of an autonomous driving system in solving the optimization problem within the current control cycle. , To predict the front wheel steering angle output by the controller for the model, The longitudinal acceleration is predicted by the model to be the output of the controller.

[0031] Driver optimal control input This refers to the control commands input by a human operator through the user interface within the current control cycle. , The front wheel steering angle input by the operator. The longitudinal acceleration is input by the operator.

[0032] Within each control cycle, the model predictive controller and the operator of the autonomous driving system independently generate their respective control inputs. The cost function of the model predictive controller in the autonomous driving system... Defined as: ; in, The tracking error is the result of the preset performance function transformation. The state error weighting matrix is... This is the weighting coefficient for the front wheel steering angle. These are the longitudinal acceleration weighting coefficients. To control the rate of change weighting matrix, The human-machine consistency weighting coefficient. To reduce the cost of obstacle avoidance, For terminal costs, For prediction in the time domain. The preset performance function transformation error is defined by logarithmic transformation as follows: ,in The first of the original tracking errors One portion, A performance function is preset for standard exponential decay. As the initial boundary, For steady-state boundary, This represents the decay rate.

[0033] Driver's cost function Defined as:

[0034] in This is the original tracking error. The operator's effective error weighting matrix, As a penalty factor for driving comfort, The weighting factor for the operator's steering angle. The operator's longitudinal acceleration weighting factor. For operator obstacle weighting, Cost of operator terminals.

[0035] The results were obtained through independent optimization and solution. and Then, the fuzzy rule permission allocation module is entered to calculate the fusion weight.

[0036] The fusion weights are determined based on three-dimensional fuzzy rules, specifically as follows: First, calculate the lateral position error of the connected vehicle. ( ), heading angle error ( and longitudinal velocity error ( ),in, This represents the actual state of connected vehicles. This is the reference state for the desired trajectory.

[0037] Then, normalization is performed, and it is expressed as: ; in, , and These are the normalized values ​​for lateral position error, heading angle error, and longitudinal velocity error, respectively. The upper limit of the lateral position error ( ), The upper bound of the heading angle error ( ), The upper bound of the longitudinal velocity error ( If any normalized value reaches or exceeds 1 ( or or If the condition is high-risk, then it is directly ordered to be... The autonomous driving system takes full control, skipping fuzzy reasoning steps. This mechanism provides a safety net for the system, ensuring that the autonomous driving system takes over first in extreme deviation conditions.

[0038] If the normalization error is less than 1, calculate the triangular membership degrees for the normalized error values ​​at three levels: small (S), medium (M), and large (B), as follows:

[0039]

[0040] ; in, , , These are the membership functions for small, medium, and large triangular shapes, respectively. This represents the normalized value of the lateral position error, the normalized value of the heading angle error, or the normalized value of the longitudinal velocity error. Establish a 3D Mamdani fuzzy rule base containing 27 rules, with the rule antecedents being... The S / M / B combination, with the rule consequent being the output single-point value, is divided into five levels: Very Small (VS), Small (S), Medium (M), Large (B), and Very Large (VB), corresponding to the following single-point output values: .

[0041] The rule design follows these principles: smaller errors result in higher driver privileges; larger errors result in higher autonomous driving privileges; longitudinal speed errors have a significant impact on safety and therefore have a faster weighting. The rule format is as follows: ; in The activation strength of each rule is calculated through product inference. The centroid method is used to solve fuzzy calculations for autonomous driving weight coefficients.

[0042] Finally, the results are subjected to amplitude limiting: To ensure that the weighting coefficients are always in Within the legal scope.

[0043] Based on autonomous driving weight coefficients The optimal control input for autonomous driving is calculated using the following formula. With driver optimal control input Perform linear weighted superposition: ; in, To integrate control inputs, For autonomous driving weighting coefficients, For driver weighting coefficients, This is the optimal control input for autonomous driving. This provides the optimal control input for the driver.

[0044] Fusion control input The following constraint is applied to the chassis actuators of the leading connected vehicle, and a saturation constraint is imposed on the fusion result: front wheel steering angle. longitudinal acceleration .when When the system is in control, the autonomous driving system takes the lead; when When the operator is in control, the operator takes the lead; when At that time, human-machine shared control.

[0045] S102: The lead connected vehicle executes the fusion control input, and each following connected vehicle runs a distributed observer. The distributed observer includes a connected vehicle observation correction term and a neighbor consistency correction term to update the connected vehicle state estimate and obtain the connected vehicle state estimate result. The observer gain design conditions are established based on the common Lyapunov function method to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology.

[0046] In step S102, each following connected vehicle Maintain the status of connected vehicles Local estimates The dynamic equations of the distributed observer consist of a superposition of three terms: ; in, ; In the formula, To follow connected vehicles The constantly updated estimate of the connected vehicle's status. To follow connected vehicles Local estimates of the status of the connected vehicle at any given time. To lead the way in connected vehicles The actual state at any given moment Follow the connected car for your neighbor exist Local estimates of the status of the connected vehicle at any given time. The sampling period is This is the drift term, used for forward state prediction based on the current estimate. For comprehensive correction items, To enhance the observation gain of connected vehicles in the navigation network, For neighbor coupling gain, To follow connected vehicles exist The communication link status between the vehicle and the lead vehicle at any given time (i.e., the status of the following vehicle) At any moment Can connected vehicles be directly observed? When can they be observed? Otherwise ), To follow connected vehicles exist Connected vehicles at all times and following The communication link status between them (i.e., indicating the status of the following connected vehicles) With connected vehicles At any moment The communication connectivity status, when connected Otherwise ), In order to keep up with the total number of connected vehicles, For connected vehicle observation and correction items, This is a neighbor consistency correction item.

[0047] The observer gain design conditions are established based on the common Lyapunov function method and are expressed as follows:

[0048] in, For topology The corresponding Laplace matrix (whose elements are derived from the observation gain of the connected vehicle) Coupling gain with neighbor Decide), This indicates taking the smallest eigenvalue of the matrix. For the system dynamics function The Lipschitz constant, The sampling period.

[0049] Under the premise that the gain condition is satisfied and the communication topology satisfies the joint connectivity assumption, the estimation error vector is defined. By constructing a common Lyapunov function, it can be proven that the estimation error satisfies the following exponential convergence bound: ; in, As a constant that depends on the observer gain and network parameters, The rate of convergence is exponential. This is the actual time. For the initial estimation error norm, This is the upper bound of the system disturbance. This represents the steady-state residual limit.

[0050] like Figure 6 As shown, when the communication topology is switched, the formation holding error decays rapidly after a brief disturbance and remains within a small range under different topology modes, verifying the exponential convergence and robustness of the topology-switching distributed observer.

[0051] S103: Each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced extension term, and a curvature-induced extension term.

[0052] In step S103, each following connected vehicle uses the connected vehicle state estimation results obtained from the distributed observer. Solve the local model predictive control optimization problem to generate nominal control inputs. The cost function for solving the local MPC optimization problem is: ; in, For formation tracking error, The expected formation position is calculated based on the estimated state of connected vehicles. For the neighbor consensus item, For obstacle avoidance, For communication maintenance items, This is a saturation penalty term. The formation error weighting matrix, To control the input weighting matrix, To control the rate of change weighting matrix, For formation tracking error weights, For neighbor consistency weight, As the terminal cost weight, To control the time domain, For prediction in the time domain.

[0053] like Figure 4 As shown, the position tracking error, longitudinal speed tracking error, and heading angle tracking error of the connected vehicle, which briefly increased due to temporary deviation from the reference trajectory during the obstacle avoidance phase, all converged rapidly within a finite time, verifying the synergistic effectiveness of the distributed observer and local model predictive control.

[0054] Nominal control input The safety filter, after online correction using a scalable preset performance safety filter, is applied to the following connected vehicle. The core of the safety filter lies in the integration of a scalable preset performance function with a joint quadratic programming framework of a control Lyapunov function and a control obstacle function. The dynamic boundary of the scalable preset performance function... It is obtained by superimposing three parts: ; in, Basic contraction term, For obstacle-inducing extensions, This is a curvature-induced extension term.

[0055] Basic contraction term ,in As the initial boundary, For steady-state boundary, The decay rate is used to ensure that the tracking error converges to the steady-state boundary without disturbance, in an exponential decay form.

[0056] Obstacle-inducing extensions Activated when the distance between the following connected vehicle and the obstacle is less than the trigger radius: ; in, The maximum extent of expansion induced by the obstacle. To follow connected vehicles With the The distance to the obstacle For the first The radius of the obstacle The obstacle trigger radius, As an indicator function, it indicates that only when... This item is activated at that time, and the closer the distance, the greater the expansion.

[0057] Curvature-induced extension term Dynamically adjust the performance boundary width based on the magnitude of the yaw rate of the connected vehicle: ; in, For curvature gain, For the yaw rate of the connected vehicle, As a curvature saturation parameter, the larger the yaw rate of a connected vehicle, the higher the curvature of the turning curve, and the greater the corresponding performance boundary expansion. When the yaw rate approaches zero, this parameter approaches zero.

[0058] like Figure 5 As shown, the composite error of the following connected vehicle is always within the feasible region of the scalable preset performance function. Even when the performance boundary is dynamically widened due to the obstacle-induced extension term within the obstacle avoidance activation interval, the composite error is still strictly smaller than the dynamic performance boundary, verifying the effectiveness of the scalable preset performance function.

[0059] velocity error component , Construct a mixed error signal: , ; ; in, , and The weights are position and velocity, respectively. is the time constant.

[0060] Constructing candidate Lyapunov functions based on logarithmic barrier functions: ; in, ; The envelope extremum is determined by the dynamic performance boundary.

[0061] The convergence constraints for controlling the Lyapunov function are: ; and For Li Daoshu, These are slack variables.

[0062] To achieve adaptive convergence rate, Based on the convergence rate, It is an adaptive gain coefficient that dynamically adjusts the convergence requirements based on the relative position of the error and the boundary.

[0063] Construct three types of higher-order control obstacle function constraints: for collision avoidance of connected vehicles, candidate functions ,in To follow connected vehicles Two-dimensional position, For the estimated location of the connected vehicle, To maintain the minimum safe distance from connected vehicles; For maintaining communication range, candidate functions ,in The maximum effective communication radius; For obstacle avoidance, candidate functions ,in For the first The location of the obstacle This defines the safety protection radius for connected vehicles. All control obstacle functions uniformly adopt a second-order constraint form. ; in and For Li Daoshu, Let be the convergence rate parameter of the control barrier function, which is greater than zero. The safety filter solves the following quadratic programming problem in each control cycle: ; in, For the corrected control input, For nominal control input, As slack variables, The slack variables are used to penalize weights; the constraints include convergence constraints of the control Lyapunov function, safety constraints of three types of higher-order control barrier functions, and control input saturation constraints. When the safety constraints are not activated, the control Lyapunov function constraints guarantee strict satisfaction of the preset performance; when obstacle avoidance is required, the control Lyapunov function constraints are allowed to be relaxed to a limited extent through slack variables to ensure the priority satisfaction of the hard constraints of the control barrier functions, thereby achieving online coordinated optimization between safety and tracking performance. The quadratic programming form guarantees the online solution efficiency of the safety filter, and the higher-order control barrier function constraints provide a formalized hard safety guarantee through forward invariance theory.

[0064] S104: Each following connected vehicle executes the modified control input to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

[0065] In step S104, the correction control input for each following connected vehicle... It is applied to the chassis actuators of each following connected vehicle to drive each following connected vehicle to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance and obstacle avoidance.

[0066] It solves the technical problems of inaccurate state estimation under time-varying communication topology switching, the contradiction between fixed performance boundaries and safety obstacle avoidance requirements, and the lack of a unified safety-performance coordination architecture. Through the organic integration of distributed observers, local model predictive control and scalable preset performance safety filters, it realizes the integrated control of distributed state estimation, formation maintenance and safety obstacle avoidance.

[0067] To demonstrate the effectiveness of this embodiment, the following simulation verification was performed: In this simulation experiment, a highly maneuverable connected vehicle cluster consisting of three connected vehicles is considered. The first connected vehicle is the lead connected vehicle, and the other two are follower connected vehicles. The lead connected vehicle moves from the starting point to the target point in a two-dimensional plane, while the follower connected vehicles maintain formation control with the lead connected vehicle. Two obstacles are placed in the movement environment of the follower connected vehicles, and one obstacle is placed in the movement environment of the lead connected vehicle. The connected vehicles need to complete the tasks of maintaining formation and avoiding obstacles.

[0068] Specifically, the main system parameters are set as follows: sampling time is 0.02s, total simulation time is 40s, the formation adopts a side-by-side configuration, and the formation offsets for following connected vehicle 1 and following connected vehicle 2 are respectively... rice.

[0069] Figure 2 shows the convoy movement trajectory of connected vehicles in an obstacle environment based on fuzzy rule permissions. It can be observed that the lead connected vehicle and the following connected vehicles can move collaboratively from the starting point to the target point. When approaching obstacles, the convoy trajectory automatically adjusts to avoid them, and the color of the connected vehicle trajectory dynamically changes according to the actual automatic permissions granted by the fuzzy rules. This demonstrates that the fuzzy rule-based human-machine co-driving and sharing control method can achieve coordinated convoy passage and dynamic permission allocation in complex obstacle environments.

[0070] Figure 3 shows the change curve of the adaptive fuzzy rule-based permission allocation for the connected vehicle. It can be observed that the autonomous driving permission weights dynamically change with the output of the fuzzy rules. During periods of large lateral, heading, or longitudinal speed errors (such as when approaching obstacles), the permission increases significantly and approaches the upper limit. Conversely, during periods of smaller errors (such as during smooth straight-line driving), the permission decreases and approaches the lower limit. This permission allocation mechanism based on the three-dimensional Mamdani fuzzy rules can achieve smooth and dynamic adjustment of human-machine control rights according to the real-time tracking error status.

[0071] Figure 4 shows the position tracking error, longitudinal velocity tracking error, and heading angle tracking error curves of the connected vehicle. It can be observed that, except for a brief increase in error due to temporary deviation from the reference trajectory during the obstacle avoidance phase, the position tracking error generally converges to around the 0.1m reference line, and the longitudinal velocity and heading angle errors converge rapidly within a limited range.

[0072] Figure 5 shows the preset performance envelope and composite error curves for followers F1 and F2. It can be observed that the composite error is always within the feasible region of the performance envelope. Even when the performance boundary is dynamically widened due to the obstacle-induced expansion term during the obstacle avoidance activation interval (around 20 s), the composite error is still strictly smaller than the dynamic performance boundary. This indicates that the scalable preset performance function can temporarily widen the tracking boundary during obstacle avoidance and ensure that the error always meets the preset performance constraints.

[0073] Figure 6 shows the timing of communication topology switching and the formation holding error curves. It can be observed that the formation holding error experiences a brief disturbance during topology switching but decays rapidly, and the error remains within a small range under different topology modes. This indicates that the distributed observer and formation control method for switching communication topologies can guarantee the stability and convergence of the formation system under dynamic reconfiguration of the communication link.

[0074] In summary, this embodiment demonstrates that the fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control method can achieve dynamic allocation of human-machine control permissions based on three-dimensional fuzzy rules, distributed exponential convergence estimation of connected vehicle states, safe obstacle avoidance and tracking recovery under scalable preset performance boundaries, platooning configuration maintenance, and multi-connected vehicle cooperative passage in complex dynamic obstacle environments and under time-varying communication topology switching conditions, thus verifying the effectiveness of the method of the present invention.

[0075] Example 2 This embodiment provides a human-machine co-driving connected vehicle platooning cooperative control system based on fuzzy rules, including the following modules: The human-machine co-driving fusion module is configured to: establish a human-machine co-driving control model for the connected vehicle, model the control rights allocation problem between the operator and the autonomous driving system as a weighted fusion of control inputs, solve the optimal control inputs for autonomous driving and driver independently, and determine the fusion weights based on three-dimensional fuzzy rules to obtain the fusion control inputs for the connected vehicle; The distributed state observation module is configured as follows: the lead connected vehicle executes the fusion control input, and each following connected vehicle runs a distributed observer. The distributed observer includes a connected vehicle observation correction term and a neighbor consistency correction term to update the connected vehicle state estimate and obtain the connected vehicle state estimate result. The observer gain design conditions are established based on the common Lyapunov function method to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology. The observer gain design module is configured such that each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced extension term, and a curvature-induced extension term. The execution module is configured to: each following connected vehicle executes the modified control inputs to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

[0076] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0077] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control method, characterized in that, include: A human-machine co-driving control model for a leading connected vehicle is established. The problem of control allocation between the operator and the autonomous driving system is modeled as a weighted fusion of control inputs. The optimal control inputs for autonomous driving and the driver are solved independently, and the fusion weights are determined based on three-dimensional fuzzy rules to obtain the fused control input of the connected vehicle. The lead connected vehicle executes fusion control inputs, and each following connected vehicle runs a distributed observer. The distributed observer includes connected vehicle observation correction terms and neighbor consistency correction terms to update the connected vehicle state estimate and obtain the connected vehicle state estimate result. Furthermore, based on the common Lyapunov function method, the observer gain design conditions are established to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology; Each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced expansion term, and a curvature-induced expansion term. Each following connected vehicle executes the modified control inputs to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

2. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The fusion weights are determined based on three-dimensional fuzzy rules, specifically as follows: First, calculate the lateral position error of the connected vehicle. Heading angle error and longitudinal velocity error And after normalization, it is expressed as: ； in, , and These are the normalized values ​​for lateral position error, heading angle error, and longitudinal velocity error, respectively. This is the upper bound of the lateral position error. This is the upper bound of the heading angle error. This is the upper bound for the longitudinal speed error; if any normalized value reaches or exceeds 1, then autonomous driving will take full control. Secondly, the membership degrees of the small, medium, and large triangles are calculated for the normalized error values, and expressed as follows: ； in, , , These are the membership functions for small, medium, and large triangles, respectively. This represents the normalized value of the lateral position error, the normalized value of the heading angle error, or the normalized value of the longitudinal velocity error. A three-dimensional Mamdani rule base is established, the activation intensity of each rule is calculated through product reasoning, the centroid method is used to defuzzify and obtain the autonomous driving weight coefficients, and the autonomous driving weight coefficients are subjected to amplitude limiting processing.

3. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The fusion control input is represented as follows: ； in, To integrate control inputs, For autonomous driving weighting coefficients, For driver weighting coefficients, This is the optimal control input for autonomous driving. This provides the optimal control input for the driver.

4. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The dynamic equation of the distributed observer is expressed as: ； in, ; In the formula, To follow connected vehicles The constantly updated estimate of the connected vehicle's status. To follow connected vehicles Local estimates of the status of the connected vehicle at any given time. To lead the way in connected vehicles The actual state at any given moment Follow the connected car for your neighbor exist Local estimates of the status of the connected vehicle at all times. The sampling period is For drift term, For comprehensive correction items, To enhance the observation gain of connected vehicles in the navigation network, For neighbor coupling gain, To follow connected vehicles exist The status of the communication link between the vehicle and the navigation network at all times. To follow connected vehicles exist Connected vehicles at all times and following The status of the communication link between them In order to keep up with the total number of connected vehicles, For connected vehicle observation and correction items, This is a neighbor consistency correction item.

5. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The observer gain design conditions are established based on the common Lyapunov function method and are expressed as follows: in, For topology The corresponding Laplace matrix, It is the Lipschitz constant; Under the condition of satisfying the gain design, the estimation error satisfies: ； in, , , It is a constant that depends on the observer gain and network parameters.

6. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The dynamic boundary of the scalable preset performance function is: ； in, Basic contraction term, For obstacle-inducing extensions, This is a curvature-induced extension term.

7. The human-machine co-driving connected vehicle platooning cooperative control method based on fuzzy rules as described in claim 1, characterized in that, The cost function for solving the local MPC optimization problem is: ； in, For formation tracking error, The expected formation position is calculated based on the estimated state of connected vehicles. For the neighbor consensus item, For obstacle avoidance, For communication maintenance items, This is a saturation penalty term.

8. A fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control system, characterized in that, Includes the following modules: The human-machine co-driving fusion module is configured to: establish a human-machine co-driving control model for the connected vehicle, model the control rights allocation problem between the operator and the autonomous driving system as a weighted fusion of control inputs, solve the optimal control inputs for autonomous driving and driver independently, and determine the fusion weights based on three-dimensional fuzzy rules to obtain the fusion control inputs for the connected vehicle; The distributed state observation module is configured as follows: the lead connected vehicle executes the fusion control input, and each following connected vehicle runs a distributed observer. The distributed observer includes a connected vehicle observation correction term and a neighbor consistency correction term to update the connected vehicle state estimate and obtain the connected vehicle state estimate result. Furthermore, based on the common Lyapunov function method, the observer gain design conditions are established to ensure the exponential convergence of the estimation error of the distributed observer under the time-varying switching condition of the communication topology; The observer gain design module is configured such that each following connected vehicle solves the local MPC optimization problem based on the connected vehicle state estimation results, generates nominal control input, and performs online correction through a preset safety filter. The safety filter introduces an extensible preset performance function that includes a basic contraction term, an obstacle-induced extension term, and a curvature-induced extension term. The execution module is configured to: each following connected vehicle executes the modified control inputs to maintain the formation configuration and meet the safety constraints of collision avoidance, communication range maintenance, and obstacle avoidance.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps of the fuzzy rule-based human-machine co-driving connected vehicle platooning cooperative control method as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the fuzzy rule-based human-machine co-driving connected vehicle platooning collaborative control method as described in any one of claims 1-7.

Images