A human-machine shared control method and device based on multi-objective model predictive control

Abstract

The present invention discloses a human-machine shared control method and device based on multi-objective model predictive control. The method includes: constructing a predictive model to implement dynamic trajectory planning based on nonlinear model predictive control; adding an obstacle avoidance penalty term to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain an obstacle avoidance trajectory that follows the plan; determining the time when the vehicle reaches the lane boundary and the time when it collides, and determining a shared control weight factor based on the time when the vehicle reaches the lane boundary and the time when it collides; obtaining the driver's steering command, taking the steering command followed by the driver and the obstacle avoidance trajectory followed by the plan as common optimization objectives, and dynamically adjusting the weights of the two optimization objectives through the shared control weight factor to achieve the purpose of tracking a reference trajectory and matching the driver's steering command when safe, thereby achieving a smooth transition of human-machine control rights. The present invention achieves a smooth transition of human-machine control rights and improves the safety of human-machine shared driving.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving control technology, and in particular to a human-machine shared control method and device based on multi-objective model predictive control. Background Art

[0002] Over the past few decades, intelligent vehicle technology has experienced rapid development. Intelligent vehicle technology is categorized into six levels, from L0 to L5, based on the intelligence level of onboard systems. Currently, advanced driver assistance systems (ADAS) with L2 intelligence are widely deployed in the market, demonstrating their ability to significantly reduce traffic accidents and alleviate driver burdens. However, fully autonomous driving technology remains difficult to achieve in the short term due to technical, legal, and ethical constraints. Consequently, industry and researchers are working to develop higher levels of intelligence to enable more frequent and intensive human-machine collaboration.

[0003] Earlier research has proposed switching control between the driver and the automated system based on the driver's attentiveness and / or the automated system's capabilities. However, some studies have shown that it is actually challenging for the driver to regain control of the vehicle due to issues such as over-trust in the automated system or a lack of situational awareness.

[0004] Therefore, a growing body of research indicates that drivers need to remain in the control loop and cooperate with automated systems to improve driving safety. Shared control, which allows the driver and automated system to steer the vehicle simultaneously by combining their control actions, is considered a viable and promising solution for transitioning from low-level automation to conditional or even high-level automation.

[0005] The core challenge of shared control is the transition or handover of human-machine control authority. In most existing shared control algorithms and system research, the driver's control commands and the intelligent system's control commands are input into the vehicle in parallel, with a simple linear weighting applied via a dynamically changing authority allocation strategy. This approach treats the driver and the intelligent system as two independent entities, combining their commands based solely on a variable weight. This approach fails to account for variations in the driver's control commands, and can easily lead to human-machine conflicts when significant differences in human-machine control commands exist.

[0006] Therefore, it is an urgent problem to propose a human-machine shared control method and device to combine the control instructions of the driver and the intelligent system in a smoother way to ensure driving comfort and safety.

[0007] In this context, the human-machine shared control method and device based on multi-objective model predictive control proposed in the present invention aims to address these limitations and challenges of the existing technology, including:

[0008] (1) Existing shared control systems usually do not take into account the changes in the driver's control instructions. When the difference between human and machine control instructions is large, it is easy to cause human-machine conflict, especially in complex or emergency driving situations.

[0009] (2) Traditional shared control strategies typically use a simple linear weighting method to process control commands from the driver and the automated system. This processing approach may not be able to effectively respond to rapidly changing driving environments in some cases, limiting the system's adaptability and efficiency. Summary of the Invention

[0010] The purpose of the present invention is to provide a human-machine shared control method and device based on multi-objective model predictive control, so as to achieve smooth fusion of driver commands and automatic driving system instructions, and realize smooth transition of human-machine control rights to ensure driving comfort and safety.

[0011] To achieve the above object, according to a first aspect of the present invention, a human-machine shared control method based on multi-objective model predictive control is provided, the method comprising:

[0012] Construct a prediction model to implement dynamic trajectory planning based on nonlinear model predictive control; and add the obstacle avoidance penalty term to the objective function of dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan;

[0013] Determine the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane, and determine a shared control weight factor based on the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane;

[0014] The driver's steering command is obtained, and following the driver's steering command and following the planned obstacle avoidance trajectory are taken as common optimization goals. The weights of these two optimization goals are dynamically adjusted through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering command when safe, thereby achieving a smooth transition of human-machine control rights.

[0015] Furthermore, a prediction model is constructed to implement dynamic trajectory planning based on nonlinear model predictive control. The obstacle avoidance penalty term is added to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan, including:

[0016] A three-degree-of-freedom bicycle model is constructed as a prediction model, and the dynamic model is discretized using the Euler method to obtain the prediction equation for trajectory planning, as shown below:

[0017] ξ t (s+i|s)=Γ[ξ t (s+i-1|s),u t (s+i-1|s)]

[0018] η t =J t ξ t

[0019] i=1,2,...P t ,J t =[0 0 0 0 1]

[0020] Where Γ represents the discrete form of the dynamic model, ξ t (s+i|s) is the predicted vehicle state at time step s+i, ξ t =[v x ,v y ,θ,X,Y] is the vehicle state, where v x is the longitudinal velocity of the vehicle, v y is the lateral velocity of the vehicle, θ is the heading angle, X is the vertical coordinate of the vehicle, and Y is the horizontal coordinate of the vehicle; u t (s+i|s) represents the control input at time step s+i, which is the lateral acceleration γ of the vehicle. y ; J t Extract the output matrix, η t Indicates output; P t is the prediction time domain;

[0021] The goal of dynamic trajectory planning is to reduce the error between the planned trajectory and the global reference trajectory while avoiding obstacles. The following objective function is established:

[0022]

[0023] Where u t is the control input, η t (s+i|s) represents the output at time step s+i, η gr (s+i|s) represents the global reference trajectory in the prediction time domain, Q t and R t are the weight matrices for output and input, respectively, C t To control the time domain, M obs represents the number of obstacles, J obs,j represents the obstacle avoidance penalty for the j-th obstacle.

[0024] Furthermore, the expression of the obstacle avoidance penalty term is as follows:

[0025]

[0026] Where, J obs,j represents the penalty value, v represents the vehicle speed, k obs represents the weight of the obstacle avoidance function, (xc ,y c ) represents the coordinates of the vehicle's center of mass, (x obs,j ,y obs,j ) represents the coordinates of the obstacle, and ζ is a very small number to prevent the denominator from being zero.

[0027] Furthermore, determining the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane boundary, and determining a shared control weight factor based on the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane boundary, includes:

[0028] Determine the vehicle's lane boundary arrival time and collision time; the lane boundary arrival time is the remaining time for the wheels to reach the current lane boundary, and the collision time is the remaining time for the vehicle to collide with the obstacle;

[0029] Determine a high-risk threshold and a low-risk threshold for the time when the vehicle reaches the lane boundary and the time when the vehicle collides, respectively; wherein the high-risk threshold is less than the corresponding low-risk threshold;

[0030] Determine a shared control weight factor based on the time between the vehicle reaching the lane boundary and the collision time, as well as a high risk threshold and a low risk threshold for the time between the vehicle reaching the lane boundary and the collision time; including:

[0031] When there is no collision risk, the shared control weight factor is adjusted according to the time it takes for the vehicle to reach the lane boundary to keep the vehicle in the current lane;

[0032] When the vehicle's arrival time at the lane boundary reaches a low-risk threshold for the vehicle's arrival time at the lane boundary, the control authority shifts from following the driver's steering command to tracking the global reference trajectory;

[0033] When there is a collision risk, the shared control weight factor is adjusted according to the collision time, and the control authority is changed from following the driver's steering instructions to following the planned obstacle avoidance trajectory.

[0034] Furthermore, determining the shared control weight factor also includes:

[0035] Setting thresholds for driver steering commands;

[0036] When the risk of lane departure increases and the driver's steering command continues to exceed this threshold, lane keeping is deemed to need to be deactivated and the driver takes control of the vehicle.

[0037] Furthermore, the formula for determining the shared control weight factor is as follows:

[0038]

[0039] Where σ is the shared control weight factor; TLC and TTC are the time it takes for the vehicle to reach the lane boundary and the time it takes for the vehicle to collide, respectively; TLC min and TTCmin Represents the high risk thresholds of TLC and TTC, TLC max and TTC max represent the low-risk thresholds of TLC and TTC respectively; σ is the driver’s steering instruction, σ lim is the threshold for the driver's steering command.

[0040] Furthermore, the calculation formula for the time it takes for a vehicle to reach the lane boundary is as follows:

[0041]

[0042] Where TLC is the time it takes for the vehicle to reach the lane boundary, v y is the vehicle's lateral speed, y u is the distance from the wheel to the lane boundary, γ y is the lateral acceleration of the vehicle;

[0043] The collision time calculation formula is as follows:

[0044]

[0045] Where TTC is the collision time, (x0, y0) represents the coordinates of the obstacle, (x c ,y c ) represents the coordinates of the vehicle, represents the vector pointing from the center of mass of the vehicle to the center of mass of the obstacle, and v represents the vehicle speed.

[0046] Furthermore, the high-risk threshold of the time when the vehicle reaches the lane boundary is defined as follows:

[0047]

[0048] Where, TLC min The high-risk threshold representing the time when the vehicle reaches the lane boundary, v x is the vehicle longitudinal speed, is the heading angle, μ is the friction force, g is the acceleration due to gravity, t d is the total response time of the actuator and the driver;

[0049] Low risk threshold TLC for the time the vehicle reaches the lane boundary max =2TLC min ;

[0050] The high-risk and low-risk thresholds for collision time are 2 seconds and 4 seconds, respectively;

[0051] Threshold σ of the driver's steering command lim 2 degrees.

[0052] Furthermore, the driver's steering command is obtained, and following the driver's steering command and following the planned obstacle avoidance trajectory are used as common optimization goals. The weights of these two optimization goals are dynamically adjusted through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering command when safe, thereby achieving a smooth transition of human-machine control rights, including:

[0053] Establish the following objective function:

[0054]

[0055] Where η s (s+i|s) represents the output of the prediction model at time step s+i, η ref (s+i|s) represents the reference output at time step s+i, P s and C s Represent the prediction time domain and control time domain respectively, δ d represents the driver's steering command, Q, S, and R represent the weight matrices of following the planned obstacle avoidance trajectory, following the driver's steering command, and control input, respectively; δ f (s|s) represents the front wheel turning angle at time step S, Δδ f (s+i|s) represents the rate of change of the rotation angle at time step s+i, δ f,min and δ f，max Represent the minimum and maximum values ​​of the front wheel turning angle, Δδ f，min and δ f，max denote the minimum and maximum values ​​of the rate of change of the turning angle, respectively; σ is the shared control weight factor. When σ is large, the goal of following the planned obstacle avoidance trajectory is dominant; on the contrary, following the driver's instructions is dominant, that is, the driver dominates the control;

[0056] By converting the objective function, it is transformed into a quadratic programming form in order to solve the optimization problem of the control input sequence; by solving the quadratic programming problem, the control input sequence of each step is obtained, and the first element in the sequence is selected as the actual control increment; finally, the actual control input δ f It is obtained by adding the previous control input and the selected control increment.

[0057] According to a second aspect of the present invention, a human-machine shared control device based on multi-objective model predictive control is provided for implementing any one of the human-machine shared control methods based on multi-objective model predictive control described above, the device comprising:

[0058] The dynamic trajectory planning module is used to build a prediction model to implement dynamic trajectory planning based on nonlinear model predictive control. The obstacle avoidance penalty term is added to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan.

[0059] A scenario assessment module that considers lateral and longitudinal risks is used to determine the time it takes for a vehicle to reach a lane boundary and the time it takes to collide, and to determine the shared control weighting factor based on the time it takes for a vehicle to reach a lane boundary and the time it takes to collide;

[0060] A shared controller module based on multi-objective model predictive control is used to obtain the driver's steering instructions, take following the driver's steering instructions and following the planned obstacle avoidance trajectory as common optimization goals, and dynamically adjust the weights of these two optimization goals through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering instructions when safe, thereby achieving a smooth transition of human-machine control rights.

[0061] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0062] This paper proposes a human-machine shared control framework based on multi-objective model predictive control. Lane keeping and obstacle avoidance objectives are added to the cost function of the MPC shared controller. The lane keeping objective helps ensure that the vehicle always stays in the intended lane, while the obstacle avoidance objective helps the vehicle avoid possible obstacles or dangers. By dynamically adjusting the weights of these two optimization objectives in the MPC cost function, a smooth transition of human-machine control rights is achieved, thereby improving the safety of human-machine shared driving.

[0063] This paper also proposes a scenario-based assessment method that uses time to lane limit (TLC) and time to collision (TTC) as two factors in driving risk assessment, simultaneously considering longitudinal collision risk and lateral lane departure risk. This method assesses risk in real time based on different scenarios and dynamically adjusts control weights to better match driver commands with control system instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A schematic diagram of a human-machine shared control framework based on multi-objective model predictive control provided by an embodiment of the present invention;

[0065] Figure 2 A schematic diagram of a scenario assessment method considering horizontal and vertical risks provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0067] The present invention provides a human-machine shared control method, which combines the control instructions of the driver and the intelligent system in a smoother way to ensure driving comfort and safety.

[0068] Example 1

[0069] This embodiment provides a shared control system design based on multi-objective model predictive control (MPC) to achieve smooth fusion of driver commands and automatic driving system instructions. Figure 1 As shown, the present invention includes a dynamic trajectory planning module based on nonlinear model predictive control (NMPC), a situational assessment method considering lateral and longitudinal risks, and a shared controller module of multi-objective MPC.

[0070] The dynamic trajectory planning module based on nonlinear model predictive control (NMPC) predicts the future state based on the current vehicle state and environmental information using the vehicle prediction model, and plans an adaptive, safe and reliable obstacle avoidance trajectory in real time.

[0071] The situational assessment method considering lateral and longitudinal risks takes into account lateral and longitudinal risks, and can comprehensively assess driving risks according to different driving conditions and driving environments.

[0072] The shared controller module of the multi-objective MPC introduces multiple control objectives such as lane keeping and obstacle avoidance into the cost function. Through the scenario assessment method considering lateral and longitudinal risks, dynamic adjustment of driver input and automation system control weights is achieved, achieving smooth fusion of driver commands and automatic driving system instructions, thereby improving driving comfort and safety.

[0073] The design of a shared control system based on multi-objective model predictive control (MPC) proposed in this embodiment includes the following steps:

[0074] 1. Dynamic trajectory planning based on nonlinear model predictive control (NMPC) requires the construction of a predictive model to accurately describe vehicle dynamic characteristics and environmental influences, and to predict the future vehicle state in real time. Furthermore, by incorporating an obstacle avoidance penalty into the cost function of the NMPC-based trajectory planning module, vehicle trajectory planning is optimized and safety is improved.

[0075] In order to achieve a trade-off between accuracy and complexity, a three-degree-of-freedom bicycle model is established as a prediction model based on the NMPC controller. The dynamic model is discretized using the Euler method to obtain the prediction equation for trajectory planning, as shown in Equation (1):

[0076]

[0077] Among them, Γ represents the discrete form of the dynamic equation, ξ t =[v x v y θX Y] is the vehicle state in the controller, P t is the prediction time domain, ξ t (s+i|s) is the predicted vehicle state at time step s+i. The control input is the vehicle's lateral acceleration γ y ,u t (s+i|s) represents the control input at time step s+i.

[0078] The goal of dynamic trajectory planning is to minimize the error between the planned trajectory and the global reference trajectory while avoiding obstacles. The following objective function is established.

[0079]

[0080] Among them, η gr (s+i|s) represents the global reference trajectory in the prediction time domain, Q t and R t are the weight matrices for output and input respectively, M obs represents the number of obstacles, J obs,j represents the obstacle avoidance penalty term for the j-th obstacle, and its expression is as follows:

[0081]

[0082] Where v represents the vehicle speed, k obs represents the weight of the obstacle avoidance function, (x c ,y c ) represents the coordinates of the vehicle's center of mass, (x obs,j ,y obs,j ) represents the coordinates of the obstacle, and ζ is a very small number to prevent the denominator from being zero.

[0083] 2. A contextual assessment approach that considers both horizontal and vertical risks

[0084] like Figure 2 As shown in the figure, a situational assessment method is designed, which takes into account lateral and longitudinal risks, to achieve dynamic adjustment of driving weights under different driving conditions and driver behaviors.

[0085] In this invention, time to lane boundary (TLC) and time to collision (TTC) are selected as two factors for scenario assessment. TLC is defined as the remaining time for the front wheels to reach the current lane boundary if the vehicle maintains its current lateral acceleration, and its expression is:

[0086]

[0087] Among them, yu It is the distance from the left (or right) front wheel to the left (or right) lane boundary. y It can be estimated through measurable parameters.

[0088] TTC is defined as the time remaining until the vehicle collides with the obstacle if it maintains its current speed. The expression for TTC is as follows:

[0089]

[0090] Among them, (x0, y0) represents the coordinates of the obstacle, (x c ,y c ) represents the coordinates of the vehicle, Represents the vector pointing from the vehicle's center of mass to the obstacle's center of mass.

[0091] When there is no collision risk, the controller's authority is adjusted based on the TLC to keep the vehicle in the current lane. When the TLC reaches a low-risk threshold, the control authority shifts from matching the driver's instructions to tracking the global reference trajectory. When obstacle avoidance becomes a prominent and urgent task, TTC is used as a risk indicator, and the control authority shifts from matching the driver's instructions to tracking the obstacle avoidance trajectory generated by dynamic programming. The present invention sets high-risk and low-risk thresholds for TTC and TLC so that the controller's objectives can be dynamically adjusted according to the corresponding thresholds.

[0092] The high-risk threshold and low-risk threshold of TTC are 2 seconds and 4 seconds respectively, while the high-risk threshold of TLC is defined as follows:

[0093]

[0094] Among them, t d is the total response time of the actuator and the driver. In the present invention, t d = 1 second. In addition, the low risk threshold of TLC is defined as TLC max =2TLC min Considering the conflict between the driver and the automated system in the lane keeping scenario, a threshold σ of the driver’s steering command is introduced when evaluating the lane departure risk. lim When the risk of lane departure increases and the driver input continues to exceed this threshold, it is considered that the lane keeping system needs to be deactivated and the driver takes control of the vehicle. In the present invention, σ is set lim =2°.

[0095] The weight factor σ is obtained as follows:

[0096]

[0097] Among them, TLC min and TTC minRepresents the high risk thresholds of TLC and TTC, TLC max and TTC max Represent the low-risk thresholds of TLC and TTC, respectively.

[0098] 3. Shared controller design for multi-objective MPC

[0099] Following the driver's instructions and following the planned obstacle avoidance trajectory are taken as common optimization goals. By dynamically adjusting the weights of these two optimization goals in the MPC cost function, a smooth transition of human-machine control rights is achieved.

[0100] The control objectives of the MPC shared controller are to track the reference trajectory and match the driver's steering command when it is safe. The smooth fusion of human and machine control commands is achieved by adjusting the weight matrices of the two control objectives.

[0101]

[0102] Among them, η s (s+i|s) represents the output of the prediction model, η ref (s+i|s) represents the reference output, P s and C s denote the prediction time domain and control time domain in the shared controller, δ d represents the driver's current steering command, Q, S, and R represent the weight matrices for trajectory tracking, tracking the driver's command, and control input, respectively. σ is the shared control weight factor, ranging from [0 to 1]. When σ is large, the goal of tracking the reference trajectory dominates; conversely, the driver dominates control.

[0103] By transforming the objective function, it is converted into a quadratic programming (QP) form in order to solve the optimization problem of the control input sequence. By solving the quadratic programming problem, the control input sequence of each step is obtained, and the first element in the sequence is selected as the actual control increment. The final actual control input δ f It is obtained by adding the previous control input and the selected control increment.

[0104] 4. Simulation Verification

[0105] Verified in the CarSim-Matlab / Simulink co-simulation platform, the proposed shared control framework helps the driver successfully complete lane keeping and obstacle avoidance tasks while maintaining vehicle stability. Compared to traditional control authority abrupt transitions, the smoother control authority transition improves vehicle safety and stability and allows the driver to more flexibly participate in vehicle control.

[0106] Example 2

[0107] This embodiment provides a human-machine shared control device based on multi-objective model predictive control for implementing any one of the human-machine shared control methods based on multi-objective model predictive control described above, the device comprising:

[0108] The dynamic trajectory planning module is used to build a prediction model to implement dynamic trajectory planning based on nonlinear model predictive control. The obstacle avoidance penalty term is added to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan.

[0109] A scenario assessment module that considers lateral and longitudinal risks is used to determine the time it takes for a vehicle to reach a lane boundary and the time it takes to collide, and to determine the shared control weighting factor based on the time it takes for a vehicle to reach a lane boundary and the time it takes to collide;

[0110] A shared controller module based on multi-objective model predictive control is used to obtain the driver's steering instructions, take following the driver's steering instructions and following the planned obstacle avoidance trajectory as common optimization goals, and dynamically adjust the weights of these two optimization goals through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering instructions when safe, thereby achieving a smooth transition of human-machine control rights.

[0111] In summary, this paper proposes a shared control framework based on multi-objective model predictive control. This approach achieves smooth fusion of driver commands and autonomous driving system instructions by adding lane keeping and obstacle avoidance control objectives to the cost function of the MPC shared controller.

[0112] Furthermore, the control weights for the driver and the automated system are assigned by adjusting the weights of the objective terms in the cost function. This dynamic weight adjustment is determined by a newly proposed situational assessment method that considers both longitudinal and lateral risks to adapt to different driving conditions and driver behaviors.

[0113] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0114] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0115] It will be easily understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A human-machine shared control method based on multi-objective model predictive control, characterized in that: The method includes: Construct a prediction model to implement dynamic trajectory planning based on nonlinear model predictive control; and add the obstacle avoidance penalty term to the objective function of dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan; Determine the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane, and determine a shared control weight factor based on the time when the vehicle reaches the lane boundary and the time when the vehicle hits the lane; Obtaining the driver's steering command, taking following the driver's steering command and following the planned obstacle avoidance trajectory as common optimization goals, and dynamically adjusting the weights of these two optimization goals through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering command when safe, thereby achieving a smooth transition of human-machine control rights; Among them, the expression of the obstacle avoidance penalty term is as follows: Where, represents the penalty value, Indicates the vehicle speed, represents the weight of the obstacle avoidance function, represents the coordinates of the vehicle's center of mass, represents the coordinates of the obstacle, ζ is a very small number to prevent the denominator from being zero; The calculation formula for the time it takes for a vehicle to reach the lane boundary is as follows: Where TLC is the time it takes for the vehicle to reach the lane boundary, is the vehicle's lateral speed, is the distance from the wheel to the lane boundary, is the lateral acceleration of the vehicle; The collision time calculation formula is as follows: Where TTC is the collision time, represents the coordinates of the obstacle, represents the coordinates of the vehicle, represents the vector pointing from the vehicle's center of mass to the obstacle's center of mass, Indicates the vehicle speed.

2. The human-machine shared control method based on multi-objective model predictive control according to claim 1 is characterized in that: Build a prediction model to realize dynamic trajectory planning based on nonlinear model predictive control; The obstacle avoidance penalty term is added to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan, including: A three-degree-of-freedom bicycle model is constructed as a prediction model, and the dynamic model is discretized using the Euler method to obtain the prediction equation for trajectory planning, as shown below: Where, represents the discrete form of the dynamics model, is the time step The predicted vehicle state at time is the vehicle status, where is the longitudinal velocity of the vehicle, is the lateral velocity of the vehicle, is the heading angle, is the vertical coordinate of the vehicle, is the horizontal coordinate of the vehicle; Represents the time step The control input is the lateral acceleration of the vehicle. ; Extract the matrix for the output quantity, Indicates the output quantity; For the prediction time domain; The goal of dynamic trajectory planning is to reduce the error between the planned trajectory and the global reference trajectory while avoiding obstacles. The following objective function is established: Where, is the control input, Represents the time step The output when represents the global reference trajectory in the prediction time domain, and are the weight matrices for output and input respectively, C t To control the time domain, Indicates the number of obstacles, Indicates the j The obstacle avoidance penalty term is 1 obstacle.

3. The human-machine shared control method based on multi-objective model predictive control according to claim 1, characterized in that: Determine the time it takes for the vehicle to reach the lane boundary and the time it collided with the vehicle, and determine a shared control weight factor based on the time it takes for the vehicle to reach the lane boundary and the time it collided with the vehicle, including: Determine the vehicle's lane boundary arrival time and collision time; the lane boundary arrival time is the remaining time for the wheels to reach the current lane boundary, and the collision time is the remaining time for the vehicle to collide with the obstacle; Determine a high-risk threshold and a low-risk threshold for the time when the vehicle reaches the lane boundary and the time when the vehicle collides, respectively; wherein the high-risk threshold is less than the corresponding low-risk threshold; Determine a shared control weight factor based on the time between the vehicle reaching the lane boundary and the collision time, as well as a high risk threshold and a low risk threshold for the time between the vehicle reaching the lane boundary and the collision time; including: When there is no collision risk, the shared control weight factor is adjusted according to the time it takes for the vehicle to reach the lane boundary to keep the vehicle in the current lane; When the vehicle's arrival time at the lane boundary reaches a low-risk threshold for the vehicle's arrival time at the lane boundary, the control authority shifts from following the driver's steering command to tracking the global reference trajectory; When there is a collision risk, the shared control weight factor is adjusted according to the collision time, and the control authority is changed from following the driver's steering instructions to following the planned obstacle avoidance trajectory.

4. The human-machine shared control method based on multi-objective model predictive control according to claim 3 is characterized in that: Determine the shared control weight factor, also including: Setting thresholds for driver steering commands; When the risk of lane departure increases and the driver's steering command continues to exceed this threshold, lane keeping is deemed to need to be deactivated and the driver takes control of the vehicle.

5. The human-machine shared control method based on multi-objective model predictive control according to claim 4 is characterized in that: The formula for determining the shared control weight factor is as follows: Where σ is the shared control weight factor; TLC and TTC are the time it takes for the vehicle to reach the lane boundary and the time to collision, respectively; and Represent the high-risk thresholds of TLC and TTC, and Represent the low-risk thresholds of TLC and TTC respectively; σ is the driver’s steering instruction, is the threshold for the driver's steering command.

6. The human-machine shared control method based on multi-objective model predictive control according to claim 4 is characterized in that: The high-risk threshold of the time when the vehicle reaches the lane boundary is defined as follows: Where, The high-risk threshold representing the time when the vehicle reaches the lane boundary, is the vehicle longitudinal speed, is the heading angle, μ is the friction force, g is the acceleration due to gravity, is the total response time of the actuator and the driver; Low risk threshold for vehicle arrival time at lane boundary ; The high-risk and low-risk thresholds for collision time are 2 seconds and 4 seconds, respectively; Threshold for driver steering commands 2 degrees.

7. The human-machine shared control method based on multi-objective model predictive control according to claim 1, characterized in that: Obtaining the driver's steering command, taking following the driver's steering command and following the planned obstacle avoidance trajectory as common optimization goals, and dynamically adjusting the weights of these two optimization goals through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering command when safe, thereby achieving a smooth transition of human-machine control rights, including: Establish the following objective function: Where, Represents the prediction model time step The output when Represents the time step The reference output when and represent the prediction time domain and the control time domain respectively, represents the driver's steering command, Q, S, and R represent the weight matrices of following the planned obstacle avoidance trajectory, following the driver's steering command, and control input, respectively; Represents the time step S The front wheel turning angle at Represents the time step The rate of change of the rotation angle, and Represent the minimum and maximum values ​​of the front wheel angle, and denote the minimum and maximum values ​​of the rate of change of the turning angle, respectively; σ is the shared control weight factor. When σ is large, the goal of following the planned obstacle avoidance trajectory is dominant; on the contrary, following the driver's instructions is dominant, that is, the driver dominates the control; By converting the objective function, it is transformed into a quadratic programming form in order to solve the optimization problem of the control input sequence; by solving the quadratic programming problem, the control input sequence of each step is obtained, and the first element in the sequence is selected as the actual control increment; finally, the actual control input It is obtained by adding the previous control input and the selected control increment.

8. A human-machine shared control device based on multi-objective model predictive control for implementing the human-machine shared control method based on multi-objective model predictive control according to any one of claims 1 to 7, characterized in that: The device includes: The dynamic trajectory planning module is used to build a prediction model to implement dynamic trajectory planning based on nonlinear model predictive control. The obstacle avoidance penalty term is added to the objective function of the dynamic trajectory planning based on nonlinear model predictive control to obtain the obstacle avoidance trajectory that follows the plan. A scenario assessment module that considers lateral and longitudinal risks is used to determine the time it takes for a vehicle to reach a lane boundary and the time it takes to collide, and to determine the shared control weighting factor based on the time it takes for a vehicle to reach a lane boundary and the time it takes to collide; A shared controller module based on multi-objective model predictive control is used to obtain the driver's steering instructions, take following the driver's steering instructions and following the planned obstacle avoidance trajectory as common optimization goals, and dynamically adjust the weights of these two optimization goals through the shared control weight factor to achieve the purpose of tracking the reference trajectory and matching the driver's steering instructions when safe, thereby achieving a smooth transition of human-machine control rights.

Images