A humanized controller design method based on model predictive control

Abstract

The present invention discloses a method for designing an anthropomorphic controller based on model predictive control, which relates to the field of intelligent transportation technology and solves the technical problem that the existing autonomous driving technology tends to be homogenized and does not conform to the driver's personalized control preferences. The key points of the technical solution are to design a path tracking controller based on model predictive control, and to use the personalized visual preview-feedback control-proportional gain-neuromuscular delay behavior of different drivers to perform anthropomorphic improvements on the path tracking controller, effectively improving the satisfaction of different drivers with vehicle control. The method can realize the anthropomorphic driving of intelligent vehicles, improve the trust and acceptance of human drivers for intelligent vehicles, and has strong practicality and broad commercial application prospects.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a method for designing an anthropomorphic controller based on model predictive control. Background Art

[0002] Intelligent connected self-driving cars based on electrification, artificial intelligence and 5G networks have become a development trend in the automotive industry. Existing research mainly focuses on improving the driving safety and control accuracy of self-driving cars, ignoring the personalized comfort needs of different passengers for self-driving cars. As consumers and users of vehicles, passengers' subjective feelings and their evaluation of the operability, safety, comfort, economy and other performance of self-driving cars are the core criteria for measuring the quality of intelligent driving technology, and are also important factors in measuring whether intelligent driving cars can be accepted. Therefore, self-driving cars must be able to provide personalized control methods for different driving groups. How to design an anthropomorphic controller that takes into account the control characteristics of different drivers is a problem that needs to be solved in this disclosure. Summary of the invention

[0003] The present application provides an anthropomorphic controller design method based on model predictive control, the technical purpose of which is to provide personalized vehicle control methods for different passenger groups and improve ride comfort.

[0004] The above technical objectives of this application are achieved through the following technical solutions:

[0005] A humanized controller design method based on model predictive control, comprising:

[0006] Establishing a vehicle-road model based on a curvilinear coordinate system and performing linearization processing on the vehicle-road model; wherein the horizontal axis of the curvilinear coordinate system is the target trajectory and the vertical axis is the lateral deviation from the center line of the right lane;

[0007] designing a first path tracking controller based on the vehicle-road model predictive control;

[0008] The first path tracking controller is improved based on the brain and neuromuscular delays in the traditional driver model to obtain a second path tracking controller;

[0009] The parameters of the second path tracking controller are adjusted based on the driver's preview duration and the steering proportional gain to obtain a third path tracking controller.

[0010] The beneficial effect of the present application is that the anthropomorphic controller design method described in the present application takes into account the driver's personalized preview-control-delay-gain behavior, so that the path tracking controller can provide personalized control methods for different drivers, which meets the personalized control preferences of different drivers. It effectively improves the vehicle's driving safety, stability, ride comfort and tracking performance, has strong practicality, and has broad commercial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A flowchart of the method described in this application;

[0012] Figure 2 Schematic diagram of the curvilinear coordinate system designed for this application;

[0013] Figure 3 A schematic diagram of the target path used in this application;

[0014] Figure 4 Schematic diagram of lateral deviation of vehicle trajectory tracking considering different driver characteristics obtained for this application. DETAILED DESCRIPTION

[0015] The technical solution of the present application will be described in detail below with reference to the accompanying drawings.

[0016] like Figure 1 As shown, a humanized controller design method based on model predictive control includes:

[0017] S1: Establishing a vehicle-road model based on a curvilinear coordinate system and linearizing the vehicle-road model for predicting the future motion state of the vehicle in a path tracking controller.

[0018] Using a curvilinear coordinate system (such as Figure 2 The target trajectory is used as the horizontal axis σ-axis of the curved coordinate system, and the lateral deviation from the center line of the right lane is the vertical axis e of the curved coordinate system. y -axis. The lateral deviation between the vehicle and the target path e y and heading deviation e ψ It can be described as:

[0019]

[0020]

[0021] Among them, e y represents the lateral deviation between the vehicle and the target path; e ψ Represents the heading deviation between the vehicle and the target path; V y Indicates the lateral speed of the vehicle; Indicates the yaw rate of the vehicle; V x represents the longitudinal velocity of the vehicle; ψ represents the yaw angle of the vehicle; Indicates the angular velocity of the target path.

[0022] ρ road represents the curvature of the target path, which is a step function that changes with the driving distance σ. It can be considered as a constant in the recursive equation. The driving distance σ on the curved road is expressed as:

[0023]

[0024] A two-degree-of-freedom vehicle model is used to describe the lateral motion and yaw motion of the vehicle. Assume that the front wheel steering angle δ f For a small angle, the longitudinal speed V x Keeping it unchanged, the vehicle model can be described as:

[0025]

[0026]

[0027] Where, m represents the vehicle mass; I z Indicates the vehicle's moment of inertia; F yf Indicates the total lateral force of the front wheel; F yr Represents the total lateral force of the rear wheel; l f Indicates the distance from the vehicle's center of mass to the front axle; l r Represents the distance from the center of mass of the vehicle to the rear axle. Assume that the front wheel slip angle α f and rear wheel slip angle α r are all small angles, then the total lateral force F of the front wheel yf and the total lateral force F of the rear wheel yr It is expressed as:

[0028]

[0029]

[0030] Among them, C f Indicates the front wheel cornering stiffness; C r Represents the rear wheel cornering stiffness; δ f represents the front wheel steering angle of the vehicle. β represents the side slip angle of the vehicle's center of mass, due to the lateral velocity V y Much smaller than the longitudinal velocity V x , β can be expressed as β = V y / V x .

[0031] Combining equations (2) to (8), the vehicle-road model is expressed as:

[0032]

[0033] in, represents the state vector; u = δ f represents the control quantity, and F represents the nonlinear system function.

[0034] The vehicle-road model is linearized and expressed as:

[0035]

[0036] y(t)=Cx(t); (10)

[0037] in, All represent time-varying coefficient matrices; y = [e y , e ψ ] T represents the output vector; C represents the selection matrix.

[0038] S2: Design a first path tracking controller based on the vehicle-road model predictive control.

[0039] Define the prediction time domain N p and control time domain N c , then at the kth moment, the state sequence, output sequence and control sequence are expressed as:

[0040]

[0041] Among them, x(k+1) and u(k) are discretized by equation (9) and expressed as:

[0042]

[0043] in, ΔT represents the discrete time of the system; Δu represents the increment of the control quantity;

[0044] The first path tracking controller is used to track the target path, and its objective function is expressed as:

[0045]

[0046] Among them, Q represents the weight matrix of the state quantity increment; R represents the weight matrix of the control quantity increment;

[0047] Combining equation (11) and equation (12), the first path tracking controller is expressed as:

[0048]

[0049] Among them, u min and u maxThey represent the minimum and maximum limits of the control quantity u respectively; Δu min and Δu max They respectively represent the minimum and maximum limits of the control quantity increment Δu.

[0050] S3: The first path tracking controller is improved based on the brain and neuromuscular delays in the traditional driver model to obtain a second path tracking controller.

[0051] Specifically, in order to reduce the conflict between the driver and the first path tracking controller and improve the driver's satisfaction with the operation of the autonomous driving vehicle, the present application improves the first path tracking controller designed above based on the driver's personalized operation characteristics. The driver's operation behavior includes four links: visual preview, feedback control, proportional gain, and brain and neuromuscular delay. The driver's personalization can be considered as the difference in these four links. The designed first path tracking controller already includes preview-control-gain behavior, but lacks a delay link. The driver's brain and neuromuscular delay can be expressed by the reaction time T d =τ d1 +τ d2 and the differential time constant τ L The present application applies a delay link to the output of the first path tracking controller to characterize the driver's brain and neuromuscular delay, and adds the driver's reaction link to the output of the first path tracking controller in the form of a transfer function δ a , and obtain the second path tracking controller, so that the control quantity output by the second path tracking controller is more in line with the driver's personalized manipulation behavior. Finally, the second path tracking controller is applied to the vehicle steering wheel angle δ af The transfer function is expressed as:

[0052]

[0053] Among them, τ d1 represents brain delay; τ d2 represents neuromuscular delay; τ L Represents the differential time constant.

[0054] S4: adjusting the parameters of the second path tracking controller based on the driver's preview duration and the steering proportional gain to obtain a third path tracking controller.

[0055] Specifically, the driver model parameters are used to modify the parameters of the second path tracking controller to better consider the driver's personalized control characteristics in the controller. Since the prediction time domain in the model predictive control is similar to the driver's preview behavior and has the same function, both reflect the ability to predict and control the future state of the vehicle. Therefore, the present application establishes the prediction time domain and the preview time length T by adjusting the discrete step length while keeping the number of prediction time domain steps unchanged. p to better simulate the driver's personalized preview behavior.

[0056] In order to take into account both the controller control accuracy and the personalized driver's control needs, this application sets the prediction time domain to half of the preview time length to effectively improve the trajectory tracking accuracy and the driver's driving experience. Set the prediction time domain step number of the second path tracking controller to N p , then the discrete step length t of the third path tracking controller step It is expressed as:

[0057]

[0058] Among them, T p Indicates preview duration.

[0059] The driver's gain behavior in the driver model changes from the expected to the proportional gain K p and compensation steering proportional gain K c Two proportional gains, K p Used to reduce lateral deviation from the target path, K c Used to reduce deviation from road curvature. K p and K c The function of the weight matrix Q in the path tracking controller is consistent. Therefore, the present application corrects the weight matrix of the second path tracking controller by proportional gain, and obtains the weight matrix of the third path tracking controller as:

[0060]

[0061] Among them, α and β are adjustable gain parameters, which are used to characterize the influence level of the driver's gain parameters. As the driver's preview time is longer, the driver's steering behavior will become gentler, and the convergence speed of the lateral deviation will be slower. Therefore, this application selects Q l =K p / T p Represents the weight of the lateral deviation in the weight matrix Q. The corresponding gain parameter α is l / T p .

[0062] As a specific embodiment, after step S4, an optimal control sequence is calculated according to the third path tracking controller, and a front wheel steering angle control instruction of the vehicle is output according to the optimal control sequence.

[0063] The initial state in the prediction time domain is the state of the vehicle at the previous moment. Combining equations (15) to (17), equation (14) is solved to obtain the increment sequence ΔU of the optimal control sequence: * , expressed as:

[0064] ΔU * (k)=[Δu * (k|k), Δu * (k+1|k),…,Δu * (k+N c -1|k)]; (18);

[0065] According to ΔU * The optimal control sequence is obtained, and the front wheel steering angle control command of the vehicle is output according to the optimal control sequence. The front wheel steering angle control command output at time k is expressed as:

[0066] u(k|k)=u(k-1|k-1)+Δu * (k|k); (19).

[0067] The new control variable u(k|k) is output to the vehicle to achieve path tracking. At the next time step k+1, the updated state from the vehicle is used to solve a new optimization problem within the prediction range, and finally the target path is tracked.

[0068] The following path tracking simulation test involving novice drivers and experienced drivers is designed to verify the anthropomorphic controller design method of the above embodiment through straight roads and U-turn roads with large curvature. Figure 3 As shown in the figure, the vehicle first passes through a 100m straight road, then enters a semicircular dangerous road with a curvature of 0.01, and finally another 100m straight road. The vehicle speed is set to 20m / s. The tracking effect obtained by inputting different driver control characteristic parameters (as shown in Table 1-2) is as follows: Figure 4 shown.

[0069] Table 1: Parameter settings for novice drivers

[0070] parameter Numeric parameter Numeric <![CDATA[τ L (--)]]> 0.28 <![CDATA[K p (--)]]> 1.7 <![CDATA[R g (--)]]> 1 / 16.7 <![CDATA[K c (--)]]> 1.7 <![CDATA[T p (s)]]> 1.6 <![CDATA[N p (--)]]> 10 <![CDATA[T d (s)]]> 0.26 <![CDATA[N c (--)]]> 10 <![CDATA[τ d1 (s)]]> 0.11 ΔT(s) 0.08 <![CDATA[τ d2 (s)]]> 0.15 α 1 / 1.6 β 40 <![CDATA[γ1,γ2]]> 0.73，0.12

[0071] Table 2: Parameter settings for skilled drivers

[0072] parameter Numeric parameter Numeric <![CDATA[τ L (--)]]> 0.28 <![CDATA[K p (--)]]> 3.2 <![CDATA[R g (--)]]> 1 / 16.7 <![CDATA[K c (--)]]> 2.8 <![CDATA[T p (s)]]> 1.2 <![CDATA[N p (--)]]> 10 <![CDATA[T d ( s )]]> 0.14 <![CDATA[N c (--)]]> 10 <![CDATA[τ dl (s)]]> 0.06 ΔT(s) 0.06 <![CDATA[τ d2 (s)]]> 0.08 α 1 / 1.2 β 40 <![CDATA[γ1,γ2]]> 0.55，0.07

[0073] contrast Figure 4From the two trajectories in Figure 1, it can be found that the path tracking controller that considers the driver's personalized control characteristics can provide different track tracking control behaviors for different drivers when driving on a curve. Different drivers' personalized factors such as visual preview, brain and neuromuscular delay, and steering proportional gain have a great influence on the control effect of the controller. The influence of visual preview is mainly reflected in the advance steering. Since the novice driver has a longer preview distance, the path tracking controller for the novice driver starts to pre-steer earlier before entering the curve, starting at about 93m, while the path tracking controller for the skilled driver starts to pre-steer at 95m.

[0074] The impact of brain and neuromuscular reaction delay is mainly reflected in the steering lag and the time to enter steady state. The prediction time domain of the path tracking controller for novice drivers is 0.2s longer than that of skilled drivers. When the vehicle speed is 20m / s, the novice driver should start steering 4m earlier than the skilled driver. However, the reaction time of the novice driver is 0.12s slower than that of the skilled driver. Therefore, the path tracking controller of the novice driver is only about 2m ahead of the path tracking controller of the skilled driver. At the same time, the reaction delay will also affect the time it takes for the vehicle to enter the steady-state steering stage on a road with large curvature (the tracking error with the target road remains unchanged). The path tracking controller for skilled drivers has a faster control response and enters the steady-state steering stage in advance at 130m, while the path tracking tracker for the novice enters the steady state at 140m.

[0075] The influence of the steering proportional gain is mainly reflected in the lateral deviation. Since the steering proportional gain of the skilled driver is larger, the tracking error of the path tracking controller for the skilled driver is smaller, which is 0.06m, while the tracking error of the path tracking controller for the novice driver is 0.13m. Therefore, the anthropomorphic controller design method proposed in this application can meet the manipulation preferences of different drivers and further improve the driving experience of the driver.

[0076] The above are exemplary embodiments of the present application, and the protection scope of the present application is defined by the claims and their equivalents. Those skilled in the art should understand that the present application is not limited by the above examples, and the above examples and descriptions are only for illustrating the principles of the present invention. The present invention may have various changes and improvements without departing from the spirit and scope of the present invention, and these changes and improvements fall within the scope of the present invention to be protected. The protection scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A humanized controller design method based on model predictive control, characterized in that: include: Establishing a vehicle-road model based on a curvilinear coordinate system and performing linearization processing on the vehicle-road model; wherein the horizontal axis of the curvilinear coordinate system is the target trajectory and the vertical axis is the lateral deviation from the center line of the right lane; designing a first path tracking controller based on the vehicle-road model predictive control; The first path tracking controller is improved based on the brain and neuromuscular delays in the traditional driver model to obtain a second path tracking controller; Adjusting the parameters of the second path tracking controller based on the driver's preview duration and the steering proportional gain to obtain a third path tracking controller; Wherein, the vehicle-road model is expressed as: in, represents the state vector; u = δ f represents the control quantity, F represents the nonlinear system function; e y represents the lateral deviation between the vehicle and the target path; e ψ Represents the heading deviation between the vehicle and the target path; V y Indicates the lateral speed of the vehicle; represents the yaw rate of the vehicle; σ represents the distance traveled on a curved road; δ f Indicates the front wheel steering angle of the vehicle; Where m represents the vehicle mass; V x represents the longitudinal velocity of the vehicle; ψ represents the yaw angle of the vehicle; represents the heading angular velocity of the target path; ρ road Represents the curvature of the target path; I z Indicates the vehicle's moment of inertia; F yf Indicates the total lateral force of the front wheel; F yr Represents the total lateral force of the rear wheel; C f Indicates the front wheel cornering stiffness; C r Represents the rear wheel cornering stiffness; α f represents the front wheel slip angle; α r Represents the rear wheel slip angle; l f Indicates the distance from the vehicle's center of mass to the front axle; l r represents the distance from the vehicle's center of mass to the rear axle; β represents the vehicle's center of mass side slip angle, and β=V y / V x ; The vehicle-road model is linearized and expressed as: y(t)=Cx(t); (10) in, All represent time-varying coefficient matrices; y = [e y ,e ψ ] T represents the output vector; C represents the selection matrix; The design of a first path tracking controller based on the vehicle-road model predictive control comprises: Define the prediction time domain N p and control time domain N c , then at the kth moment, the state sequence, output sequence and control sequence are expressed as: Among them, x(k+1) and u(k) are discretized by equation (9) and expressed as: in, ΔT represents the discrete time of the system; Δu represents the increment of the control quantity; The first path tracking controller is used to track the target path, and its objective function is expressed as: Among them, Q represents the weight matrix of the state quantity increment; R represents the weight matrix of the control quantity increment; Combining equation (11) and equation (12), the first path tracking controller is expressed as: Among them, u min and u max They represent the minimum and maximum limits of the control quantity u respectively; Δu min and Δu max They represent the minimum and maximum limits of the control quantity increment Δu respectively; The first path tracking controller is improved based on the brain and neuromuscular delay in the traditional driver model to obtain a second path tracking controller, including: The reaction time of the brain and neuromuscular delay is added to the output δ of the first path tracking controller through the transfer function a , the second path tracking controller is obtained, and the transfer function is expressed as: Among them, δ af represents the steering wheel angle applied to the vehicle by the second path tracking controller; τ d1 represents brain delay; τ d2 represents neuromuscular delay; τ L represents the differential time constant.

2. The method according to claim 1, characterized in that The method of adjusting the parameters of the second path tracking controller based on the preview duration and the steering proportional gain of the driver to obtain a third path tracking controller includes: Set the prediction time domain step number of the second path tracking controller to N p , then the discrete step length t of the third path tracking controller step Expressed as: Among them, T p Indicates preview duration; The weight matrix of the second path tracking controller is corrected by the proportional gain, and the weight matrix of the third path tracking controller is obtained as follows: Where α and β are adjustable gain parameters; K p Indicates the expected steering proportional gain; K c Indicates the compensation steering proportional gain; Q l =K p / T p Represents the weight of the lateral deviation in the weight matrix Q. The corresponding gain parameter α is l / T p .

Images