A trajectory tracking system and method for four-wheel independent steering vehicles on icy and snowy roads.

Abstract

This invention discloses a trajectory tracking system and method for a four-wheel independent steering vehicle on icy and snowy roads. The system includes: collecting icy and snowy road surface information data and vehicle driving state parameter data during vehicle operation to obtain the adhesion coefficients of the four wheels of the four-wheel independent steering vehicle; estimating the lateral stiffness of the four tires of the four-wheel independent steering vehicle; establishing an actual system model to determine the input and output quantities of trajectory tracking control; and combining the optimal nominal system model control law and the pipeline control law to form the final control input of the actual system model, thereby achieving trajectory tracking of the four-wheel independent steering vehicle on icy and snowy roads. This invention, by combining the optimal nominal system model control law and the pipeline control law, enables the four-wheel independent steering vehicle to achieve trajectory tracking control under the uncertainty of uneven snow depth on the road surface.

Description

Technical Field

[0001] This invention belongs to the field of automotive drive-by-wire chassis technology, specifically relating to a trajectory tracking system and method for a four-wheel independent steering vehicle on icy and snowy roads. Background Technology

[0002] Four-wheel independent steering (4WIS) vehicles can independently control the angle of each wheel via electrical signals, providing a new and optimal steering platform for intelligent driving. Vehicle trajectory tracking control is not only crucial for improving vehicle safety and passenger comfort, but also helps maintain vehicle stability, enabling the vehicle to adaptively adjust to dynamic changes in the driving environment. With the increasing prevalence of autonomous driving and advanced driver assistance systems (ADAS), four-wheel independent steering vehicles have become the mainstream steering platform, a natural trend. Compared to traditional steering vehicles, four-wheel independent steering vehicles can individually control the angle of each wheel, maximizing the precision of vehicle trajectory control and adapting to dynamic changes in various driving environments, thus improving trajectory tracking accuracy.

[0003] Under icy and snowy road conditions, the trajectory tracking performance of four-wheel independent steering vehicles deteriorates due to uncertainties in road conditions, including uncertainties in the adhesion coefficients of the four wheels and the unevenness of snow depth. Therefore, it is urgent to develop trajectory tracking control solutions for four-wheel independent steering vehicles on icy and snowy roads.

[0004] Chinese invention patent application CN202310480148.9, entitled "Vehicle Trajectory Tracking Control System Based on Parameter Uncertainty and Yaw Stability," proposes a method that combines commonly used low-cost onboard sensors with a mathematical model to estimate the vehicle's yaw rate and longitudinal velocity. Applying yaw stability control to the lateral LQR controller significantly improves path tracking accuracy; however, this patent only considers its own modeling uncertainty, ignoring the uncertainty of the external environment, thus increasing trajectory tracking error. Chinese invention patent application CN201710146567.3, entitled "An Unmanned Vehicle Extreme Dynamics Trajectory Tracking Control System," proposes an unmanned vehicle extreme dynamics trajectory tracking control system. This control system includes a sensor module, a velocity file solving module, and a computational control module. For trajectory tracking problems with known paths, it can realize the extreme driving behavior of unmanned vehicles, enabling them to complete the trajectory tracking process at the fastest speed; however, this patent only considers the limits of its own model, without considering the driving environment, increasing the danger of vehicle operation.

[0005] Existing trajectory tracking control methods often neglect the influence of road surface conditions in the driving environment, especially on icy and snowy roads. Due to the structural characteristics of four-wheel independent steering vehicles, each wheel is an individual control unit. Therefore, icy and snowy roads cause uncertainty in the adhesion coefficient of the four wheels, resulting in uncertainty in the lateral stiffness of the four tires, as well as uncertainty caused by uneven snow depth on the road surface, all of which affect the accuracy of vehicle trajectory tracking. Therefore, trajectory tracking methods that address the uncertainties of road surface conditions on icy and snowy roads have great application prospects. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the purpose of this invention is to provide a trajectory tracking system and method for four-wheel independent steering vehicles on icy and snowy roads, so as to solve the problem that the prior art ignores the driving environment of the vehicle on icy and snowy roads, resulting in uncertainty in the lateral stiffness of the four-wheel independent steering tires and uncertainty caused by the unevenness of the snow depth on the road surface, thus making it impossible to accurately track the trajectory of four-wheel independent steering vehicles.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] The present invention provides a trajectory tracking system for a four-wheel independent steering vehicle on icy and snowy roads, comprising: a four-wheel tire lateral stiffness estimation unit and a trajectory tracking control unit;

[0009] The four-wheel tire lateral stiffness estimation unit includes: an ice and snow road environment data acquisition module, a vehicle driving state parameter acquisition module, and an estimation module;

[0010] The ice and snow road surface environment data acquisition module is used to collect ice and snow road surface information data and obtain the adhesion coefficient of the four wheels of the four-wheel independent steering vehicle.

[0011] The vehicle driving status parameter acquisition module is used to collect vehicle driving status parameter data.

[0012] The estimation module is used to estimate the lateral stiffness of the four tires using the adhesion coefficients of the four wheels and vehicle driving state parameter data.

[0013] The trajectory tracking control unit includes: a nominal system module, an actual system module, a pipeline construction module, and a control module;

[0014] The nominal system module is used to establish a nominal system model based on the estimation results of the lateral stiffness of the four-wheel tires and output the nominal system model control law;

[0015] The actual system module is used to establish an actual system model based on the nominal system model and the uncertainties caused by the unevenness of snow depth on the road surface;

[0016] The pipeline construction module is used to generate a pipeline centered on the trajectory of the nominal system model and taking into account the uncertainty of the icy and snowy road conditions. This ensures that the trajectory of the actual system model remains within the pipeline despite the uncertainty caused by the uneven snow depth on the road surface, and outputs the pipeline control law.

[0017] The control module is used to generate the final control input of the actual system model based on the nominal system model control law and pipeline control law, thereby minimizing the trajectory tracking error.

[0018] Furthermore, the icy and snowy road surface environment data acquisition module is a road detection sensor.

[0019] Furthermore, the vehicle driving state parameter acquisition module includes: a lateral acceleration sensor, a yaw rate acceleration sensor, a wheel rotation angle sensor, a longitudinal velocity sensor, a center of gravity sideslip angle sensor, and a yaw rate sensor.

[0020] Furthermore, the vehicle driving state parameter data includes: vehicle lateral acceleration, yaw acceleration, wheel angle, longitudinal velocity, center of gravity sideslip angle, and yaw rate, wherein the wheel angle includes the left front wheel angle, left rear wheel angle, right front wheel angle, and right rear wheel angle.

[0021] Furthermore, the nominal system module establishes a nominal system model, and based on the established nominal system model, establishes a cost function with the goal of minimizing the tracking error, obtains the optimal nominal system model control sequence, takes the first term of the sequence, and outputs the optimal nominal system model control law.

[0022] Furthermore, the actual system module establishes an actual system model. Compared with the nominal system model, the actual system model includes the uncertainty of uneven snow depth on the road surface. The input and output quantities of trajectory tracking control are determined through the actual system model.

[0023] Furthermore, the pipeline construction module is used to define the uncertainty of icy and snowy road conditions and generate pipelines, and outputs pipeline control laws.

[0024] Furthermore, the definition of uncertainty in the icy and snowy road surface conditions describes the possible range of deviations from the nominal system model, representing the uncertainty of uneven snow depth on the road surface.

[0025] Furthermore, the generation of the pipeline represents the allowable deviation range centered on the trajectory of the nominal system model.

[0026] Furthermore, the pipeline construction module outputs a pipeline control law to ensure that the trajectory of the actual system model remains within the generated pipeline despite the uncertainty caused by the uneven depth of snow accumulation on the road surface.

[0027] Furthermore, the control module combines the optimal nominal system model control law and pipeline control law to form the final control input of the actual system model, which is used for trajectory tracking of four-wheel independent steering vehicles on icy and snowy roads.

[0028] The present invention provides a trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads, based on the above system, comprising the following steps:

[0029] 1) Collect information data on icy and snowy road surfaces and vehicle driving status parameters during vehicle operation, and use the information data on icy and snowy road surfaces to obtain the adhesion coefficients of the four wheels of the four-wheel independent steering vehicle.

[0030] 2) The adhesion coefficients of the four wheels and the vehicle driving state parameter data are used as inputs to a bidirectional long short-term memory neural network to estimate the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle.

[0031] 3) Establish a nominal system model based on the estimation results of the four-wheel tire lateral stiffness, output the optimal nominal system model control law, establish an actual system model based on the nominal system model and the uncertainty of the unevenness of snow depth on the road surface during vehicle driving, and determine the input and output quantities of trajectory tracking control.

[0032] 4) Using the trajectory of the nominal system model as the center, a pipeline is generated in combination with the uncertainty of the icy and snowy road conditions, and the pipeline control law is output. The optimal nominal system model control law and the pipeline control law are combined to form the final control input of the actual system model, so as to realize the trajectory tracking of the four-wheel independent steering vehicle on the icy and snowy road.

[0033] Furthermore, the vehicle driving state parameter data in step 1) includes: vehicle lateral acceleration, yaw acceleration, wheel angle, longitudinal velocity, center of gravity sideslip angle, and yaw rate, wherein the wheel angle includes the left front wheel angle, left rear wheel angle, right front wheel angle, and right rear wheel angle.

[0034] Furthermore, in step 1), the road detection sensor takes pictures to obtain image information of the road ahead, and the image information is used to compare the road surface conditions under different adhesion coefficients to obtain the adhesion coefficients of the four wheels.

[0035] Furthermore, the bidirectional long short-term memory neural network in step 2) includes: an input gate, a forget gate, and an output gate; the input gate includes a forward input gate and a backward input gate;

[0036] The forward input gate combines the immediate impact of the current icy and snowy road conditions on the lateral stiffness of the four-wheel tires with the impact of the icy and snowy road surface on the lateral stiffness of the four-wheel tires over a longer period of time, adding or deleting relevant information on the changes in the lateral stiffness of the four-wheel tires. The reverse input gate is added to obtain the adhesion coefficient or driving state parameter data that was missed under the icy and snowy road conditions. The forget gate is used to control the deletion of redundant adhesion coefficient or driving state parameter data of the four wheels. Finally, the output gate outputs the data.

[0037] Furthermore, the specific steps for estimating the lateral stiffness of the four-wheel tires in step 2) are as follows:

[0038] 21) When the icy and snowy road conditions change during the vehicle's operation, the forget gate forgets the adhesion coefficients of the four wheels or the vehicle's driving state parameter data from the previous moment, re-enters the current adhesion coefficients and driving state parameter data of the four wheels, and updates the output of the forget gate.

[0039] 22) The forward input gate uses the adhesion coefficients of the four wheels to adjust the influence weights of the adhesion coefficients and driving state parameter data of the four wheels, and obtains the adhesion coefficient data or driving state parameter data that were missed on the icy and snowy road surface through the reverse input gate to update the input of the bidirectional long short-term memory neural network.

[0040] 23) Based on the input of the forget gate and the bidirectional long short-term memory neural network, update the adhesion coefficient and driving state parameter data accumulation of the four wheels for the current time-to-time four-wheel tire lateral stiffness estimation;

[0041] 24) Update the current output of the output gate, and combine the current adhesion coefficient and driving state parameter data of the four wheels to output the memory and processing status of the four-wheel tire lateral stiffness at the current moment.

[0042] 25) The bidirectional long short-term memory neural network outputs the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle.

[0043] Further, in step 21), the lateral acceleration a of the vehicle is selected. y yaw acceleration Left

[0044] Front wheel steering angle δ lf Left rear wheel steering angle δ lr Right front wheel steering angle δ rf Right rear wheel steering angle δ rr Longitudinal velocity v x Side slip angle β, yaw rate ω, and left front wheel adhesion coefficient u lf Left rear wheel adhesion coefficient u lr Right front wheel adhesion coefficient u rf and the right rear wheel adhesion coefficient u rrThe input, used as the input to the bidirectional long short-term memory neural network, is expressed as follows:

[0045]

[0046] The output of the forget gate is represented as:

[0047] f (t) =σ(W f h (t-1) +U f x0 (t) +b f )

[0048] Among them, f (t) The output of the forget gate is σ, where σ is the activation function and h is the output of the forget gate. (t-1) This refers to the memory and processing status of the adhesion coefficient data and driving state parameter data of the four wheels at time t-1, x0 (t) W is the input to the forward input gate at time t. f U f b is the weighting coefficient for the forget gate. f This is the bias value for the forget gate.

[0049] Furthermore, in step 22), the adhesion coefficient data or driving state parameter data that were missed on the icy and snowy road surface are obtained through the reverse input gate. (t) The input to the bidirectional long short-term memory neural network at time t is expressed as follows:

[0050] x (t) =x0 (t) +x n (t)

[0051] Where, x n (t) represents the missing adhesion coefficient or driving state parameter data, which is the input of the reverse input gate at time t.

[0052] Furthermore, the input to the bidirectional long short-term memory neural network updated in step 22) comprises two parts: the first part is the output i through the activation function σ. (t) The second part outputs a temporary estimate of the lateral stiffness 'a' of the four-wheel tires through the activation function tanh. (t) The input representation of a bidirectional long short-term memory neural network is as follows:

[0053]

[0054] Among them, i (t) This represents the first part of the input, a. (t) W represents a temporary estimate of the lateral stiffness of the four-wheel tires. i U is the weighting factor for the adhesion coefficient of the four wheels. iThe weighting coefficients for driving state parameter data are adjusted by W. i and U i Update input, W a U a b represents the weighting coefficients for feature extraction. i b is the input bias value. a This is the bias value for feature extraction.

[0055] Furthermore, in step 23), the accumulated data of the adhesion coefficient and driving state parameters of the four wheels are updated. (t) C (t) Determined by the forget gate and the input, the expression is as follows:

[0056] C (t) =C (t-1) ⊙f (t) +i (t) ⊙a (t)

[0057] Where ⊙ represents the Hadamard product.

[0058] Furthermore, in step 24), the current output O of the output gate is updated. (t) The state h is determined by the memory and processing of the adhesion coefficient data and driving state parameter data of the four wheels at time t-1. (t-1) The input x of the bidirectional long short-term memory neural network at time t (t) The output of the output gate is obtained through the activation function σ and is expressed as follows:

[0059] O (t) =σ(W o h (t-1) +U o x (t) +b o )

[0060] Among them, W o U o b represents the weighting coefficient of the output gate. o This is the bias value for the output gate.

[0061] Furthermore, in step 24), the memory and processing state h of the current moment's four wheel adhesion coefficient data and driving state parameter data is output. (t) h (t) It is determined by the product of two parts, the first part being the output O of the output gate obtained through the activation function σ. (t) The second part is C obtained through the activation function tanh. (t) , is represented as:

[0062] h (t) =O (t)⊙tanh(C t ).

[0063] Furthermore, the output of the four-wheel independent steering vehicle's four-wheel tire lateral stiffness in step 25) is expressed as:

[0064] Output: {C lf C lr C rf C rr}=σ(Vh (t) +c)

[0065] Where V is the weight matrix connecting the adhesion coefficient data and driving state parameter data of the four wheels, representing the memory, processing state, and output; c is the bias term added to the weighted sum; C lf For the lateral stiffness of the left front wheel, C lr For the lateral stiffness of the left rear wheel, C rf For the right front wheel lateral stiffness, C rr This refers to the lateral stiffness of the right rear wheel.

[0066] Furthermore, the specific steps of step 3) are as follows:

[0067] 31) Establish a nominal system model based on the estimation results of the four-wheel tire lateral stiffness. By generating the cost function, obtain the optimal nominal system model control sequence. Take the first term of the sequence and output the optimal nominal system model control law.

[0068] 32) Based on the nominal system model and the uncertainty of uneven snow depth on the road surface, establish the actual system model and determine the input and output quantities of trajectory tracking control.

[0069] Furthermore, the nominal system model in step 31) is as follows:

[0070]

[0071] in, The trajectory of the nominal system model. The deviation between the nominal system model output and the actual system model output. The deviation between the nominal system model input and the actual system model input is represented by A, which is the state matrix containing the lateral stiffness of the four tires, and B is the input matrix containing the lateral stiffness of the four tires.

[0072] Further, the cost function in step 31) is defined as:

[0073]

[0074] Among them, the objective function constraint is U is a constraint on the function. For additional mandatory constraints, the constant N p and N c Indicates the prediction range and control range; k represents time. and These are weighted diagonal matrices representing the state variables and input variables, respectively. For the state dimension, The input dimension is defined by the relaxation factor ε and the weighting coefficient κ, which are used to balance constraints and avoid infeasible optimization problems caused by constraints. When the constraints are exceeded, the relaxation variable ε becomes positive and expands the input constraints of the nominal system model. The feasible range; otherwise, the slack variable ε is set to 0.

[0075] Furthermore, the optimal nominal system model control law in step 31) is expressed as:

[0076]

[0077] Among them, U * (k) is the optimal control sequence for the nominal system model. The optimal nominal system model control law is given by the corresponding optimal nominal system model trajectory. The cost function is J * (k).

[0078] Furthermore, the actual system model in step 32) is represented as follows:

[0079]

[0080] Where μ is the input quantity for trajectory tracking control, μ = (δ lf ,δ lr ,δ rf ,δ rr ) T , For the output quantity of trajectory tracking control, Let be the heading angle of the four-wheel independent steering vehicle, Y be the lateral position of the four-wheel independent steering vehicle, and v be the yaw angle. x Let ω be the longitudinal velocity of the four-wheel independent steering vehicle, and ω be the yaw rate of the four-wheel independent steering vehicle. This indicates the uncertainty caused by the uneven depth of snow accumulation on the road surface.

[0081] Furthermore, the specific steps of step 4) are as follows:

[0082] 41) Define the uncertainty caused by uneven snow depth on the road surface;

[0083] 42) Based on the uncertainty, generate a pipeline centered on the trajectory of the nominal system model;

[0084] 43) Design the pipeline control law, and use the output of the pipeline control law to keep the trajectory of the actual system model within the generated pipeline;

[0085] 44) Combining the optimal nominal system model control law and pipeline control law, the final control input of the actual system model is formed.

[0086] Furthermore, the uncertainty arising from the uneven snow depth on the road surface in step 41) The expression is as follows:

[0087]

[0088] Where ∈(k) is the error between the nominal system model and the actual system model, and η(k) represents the pipeline control law.

[0089] Furthermore, in step 42), the pipe Ω centered on the trajectory of the nominal system model... tube The expression is as follows:

[0090]

[0091] Among them, C e It is a constant matrix. This represents the maximum range of uncertainty in the unevenness of snow depth on the road surface.

[0092] Furthermore, the pipeline control law expression in step 43) is as follows:

[0093]

[0094] Where η(k) is the pipeline control law; t represents the current time. λ represents the uncertainty of uneven snow depth on the road surface; λ represents the pipeline parameters, which are determined by deviations from the nominal system model.

[0095] Furthermore, the final control input expression of the actual system model in step 44) is:

[0096]

[0097] in, Let η(k) be the optimal nominal system model control law, and η(k) be the pipeline control law. This serves as the final control input for the actual system model.

[0098] The beneficial effects of this invention are:

[0099] This invention addresses the unique characteristic of four-wheel independent steering vehicles where wheel states are independent. It estimates the adhesion coefficients of all four wheels and uses a long short-term memory (LSTM) neural network with a forgetting gate to accurately determine changes in road surface information. Furthermore, by adding a bidirectional LSTM neural network, it acquires adhesion coefficient or driving state parameter data that may have been missed on icy or snowy roads. This prevents the forgetting gate from forgetting retained road surface information, ensuring accurate capture of adhesion coefficient information on icy or snowy roads. It also prevents the omission of estimated road surface information due to the uncertainty of adhesion coefficients for all four wheels, effectively improving the estimation results of four-wheel tire lateral stiffness.

[0100] This invention establishes a nominal system model based on the estimation results of the four-wheel tire lateral stiffness, and obtains the optimal nominal system model control law, enabling four-wheel independent steering vehicles to achieve accurate trajectory tracking under the uncertainty of the adhesion coefficient of the four wheels on icy and snowy roads; it constructs a pipeline to minimize the difference between the trajectory of the actual system model and the trajectory of the nominal system model, helping the actual system model to keep up with the nominal system model, which not only achieves robustness, but also effectively improves the trajectory tracking accuracy.

[0101] This invention combines the optimal nominal system model control law and pipeline control law to enable four-wheel independent steering vehicles to achieve trajectory tracking control under uncertainties caused by uneven snow depth on the road surface. Attached Figure Description

[0102] Figure 1 This is a structural diagram of the system of the present invention.

[0103] Figure 2 This is a flowchart of the method of the present invention.

[0104] Figure 3 This is a flowchart for estimating the lateral stiffness of four-wheeled tires in this invention. Detailed Implementation

[0105] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0106] Reference Figure 1 As shown, the present invention provides a trajectory tracking system for a four-wheel independent steering vehicle on icy and snowy roads, comprising: a four-wheel tire lateral stiffness estimation unit and a trajectory tracking control unit;

[0107] The four-wheel tire lateral stiffness estimation unit includes: an ice and snow road environment data acquisition module, a vehicle driving state parameter acquisition module and an estimation module; the ice and snow road environment data acquisition module is a road detection sensor.

[0108] The ice and snow road surface environment data acquisition module is used to collect ice and snow road surface information data to obtain the adhesion coefficient of the four wheels of the four-wheel independent steering vehicle.

[0109] The vehicle driving status parameter acquisition module is used to collect vehicle driving status parameter data; the vehicle driving status parameter acquisition module includes: a lateral acceleration sensor, a yaw angle acceleration sensor, a wheel rotation angle sensor, a longitudinal velocity sensor, a center of gravity sideslip angle sensor, and a yaw rate sensor;

[0110] The vehicle driving state parameter data includes: vehicle lateral acceleration, yaw acceleration, wheel angle, longitudinal velocity, center of gravity sideslip angle, and yaw rate, wherein the wheel angle includes the left front wheel angle, left rear wheel angle, right front wheel angle, and right rear wheel angle.

[0111] The estimation module is used to estimate the lateral stiffness of the four tires using the adhesion coefficients of the four wheels and vehicle driving state parameter data.

[0112] The trajectory tracking control unit includes: a nominal system module, an actual system module, a pipeline construction module, and a control module;

[0113] The nominal system module is used to establish a nominal system model based on the estimation results of the lateral stiffness of the four-wheel tires and output the nominal system model control law;

[0114] The actual system module is used to establish an actual system model based on the nominal system model and the uncertainties caused by the unevenness of snow depth on the road surface;

[0115] The pipeline construction module is used to generate a pipeline centered on the trajectory of the nominal system model and taking into account the uncertainty of the icy and snowy road conditions. This ensures that the trajectory of the actual system model remains within the pipeline despite the uncertainty caused by the uneven snow depth on the road surface, and outputs the pipeline control law.

[0116] The control module is used to generate the final control input of the actual system model based on the nominal system model control law and pipeline control law, thereby minimizing the trajectory tracking error.

[0117] The nominal system module establishes a nominal system model, and based on the established nominal system model, establishes a cost function with the goal of minimizing the tracking error, obtains the optimal nominal system model control sequence, takes the first term of the sequence, and outputs the optimal nominal system model control law.

[0118] The actual system module establishes an actual system model, which, compared to the nominal system model, includes the uncertainty of uneven snow depth on the road surface. The actual system model is used to determine the input and output quantities of trajectory tracking control.

[0119] The pipeline construction module is used to define the uncertainty of icy and snowy road conditions and generate pipelines, and outputs pipeline control laws.

[0120] The uncertainty definition of the icy and snowy road surface conditions describes the possible range of deviations from the nominal system model, representing the uncertainty of the unevenness of snow depth on the road surface.

[0121] The generation of the pipeline represents the allowable deviation range centered on the trajectory of the nominal system model.

[0122] The pipeline construction module outputs a pipeline control law to ensure that the trajectory of the actual system model remains within the generated pipeline despite the uncertainty caused by the uneven depth of snow accumulation on the road surface.

[0123] The control module combines the optimal nominal system model control law and pipeline control law to form the final control input of the actual system model, which is used for trajectory tracking of four-wheel independent steering vehicles on icy and snowy roads.

[0124] Reference Figure 2 As shown, the present invention provides a trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads, based on the above system, with the following steps:

[0125] 1) Collect information data on icy and snowy road surfaces and vehicle driving status parameters during vehicle operation, and use the information data on icy and snowy road surfaces to obtain the adhesion coefficients of the four wheels of the four-wheel independent steering vehicle.

[0126] Vehicle driving status parameter data include: vehicle lateral acceleration, yaw acceleration, wheel angle, longitudinal velocity, center of gravity sideslip angle, and yaw rate, where the wheel angle includes the left front wheel angle, left rear wheel angle, right front wheel angle, and right rear wheel angle.

[0127] In step 1), the road detection sensor takes pictures to obtain image information of the road ahead, and the image information is used to compare the road surface conditions under different adhesion coefficients to obtain the adhesion coefficients of the four wheels.

[0128] 2) The adhesion coefficients of the four wheels and the vehicle driving state parameter data are used as inputs to a bidirectional long short-term memory neural network to estimate the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle.

[0129] The bidirectional long short-term memory neural network includes: an input gate, a forget gate, and an output gate; the input gate includes a forward input gate and a backward input gate.

[0130] The forward input gate combines the immediate impact of the current icy and snowy road conditions on the lateral stiffness of the four-wheel tires with the impact of the icy and snowy road surface on the lateral stiffness of the four-wheel tires over a longer period of time, adding or deleting relevant information on the changes in the lateral stiffness of the four-wheel tires. The reverse input gate is added to obtain the adhesion coefficient or driving state parameter data that was missed under the icy and snowy road conditions. The forget gate is used to control the deletion of redundant adhesion coefficient or driving state parameter data of the four wheels. Finally, the output gate outputs the data.

[0131] Reference Figure 3 As shown, the specific steps for estimating the lateral stiffness of the four-wheel tires in step 2) are as follows:

[0132] 21) When the icy and snowy road conditions change during the vehicle's operation, the forget gate forgets the adhesion coefficients of the four wheels or the vehicle's driving state parameter data from the previous moment, re-enters the current adhesion coefficients and driving state parameter data of the four wheels, and updates the output of the forget gate.

[0133] 22) The forward input gate uses the adhesion coefficients of the four wheels to adjust the influence weights of the adhesion coefficients and driving state parameter data of the four wheels, and obtains the adhesion coefficient data or driving state parameter data that were missed on the icy and snowy road surface through the reverse input gate to update the input of the bidirectional long short-term memory neural network.

[0134] 23) Based on the input of the forget gate and the bidirectional long short-term memory neural network, update the adhesion coefficient and driving state parameter data accumulation of the four wheels for the current time-to-time four-wheel tire lateral stiffness estimation;

[0135] 24) Update the current output of the output gate, and combine the current adhesion coefficient and driving state parameter data of the four wheels to output the memory and processing status of the four-wheel tire lateral stiffness at the current moment.

[0136] 25) The bidirectional long short-term memory neural network outputs the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle.

[0137] Specifically, in step 21), the lateral acceleration a of the vehicle is selected. y yaw acceleration Left front wheel steering angle δ lf Left rear wheel steering angle δ lr Right front wheel steering angle δ rf Right rear wheel steering angle δ rr Longitudinal velocity v x Side slip angle β, yaw rate ω, and left front wheel adhesion coefficient u lf Left rear wheel adhesion coefficient u lr Right front wheel adhesion coefficient u rf and the right rear wheel adhesion coefficient u rrThe input, used as the input to the bidirectional long short-term memory neural network, is expressed as follows:

[0138]

[0139] The output of the forget gate is represented as:

[0140] f (t) =σ(W f h (t-1) +U f x0 (t) +b f )

[0141] Among them, f (t) The output of the forget gate is σ, where σ is the activation function and h is the output of the forget gate. (t-1) This refers to the memory and processing status of the adhesion coefficient data and driving state parameter data of the four wheels at time t-1, x0 (t) W is the input to the forward input gate at time t. f U f b is the weighting coefficient for the forget gate. f This is the bias value for the forget gate.

[0142] Specifically, in step 22), the adhesion coefficient data or driving state parameter data missed on the icy and snowy road surface are obtained through the reverse input gate. (t) The input to the bidirectional long short-term memory neural network at time t is expressed as follows:

[0143] x (t) =x0 (t) +x n (t)

[0144] Where, x n (t) represents the missing adhesion coefficient or driving state parameter data, which is the input of the reverse input gate at time t.

[0145] Specifically, the input to the bidirectional long short-term memory neural network in step 22) consists of two parts: the first part is the output i through the activation function σ. (t) The second part outputs a temporary estimate of the lateral stiffness 'a' of the four-wheel tires through the activation function tanh. (t) The input representation of a bidirectional long short-term memory neural network is as follows:

[0146]

[0147] Among them, i (t) This represents the first part of the input, a. (t) W represents a temporary estimate of the lateral stiffness of the four-wheel tires. i U is the weighting factor for the adhesion coefficient of the four wheels. iThe weighting coefficients for driving state parameter data are adjusted by W. i and U i Update input, W a U a b represents the weighting coefficients for feature extraction. i b is the input bias value. a This is the bias value for feature extraction.

[0148] Specifically, in step 23), the accumulated data of the adhesion coefficient and driving state parameters of the four wheels are updated. (t) C (t) Determined by the inputs of the forget gate and the bidirectional long short-term memory neural network, the expression is as follows:

[0149] C (t) =C (t-1) ⊙f (t) +i (t) ⊙a (t)

[0150] Where ⊙ represents the Hadamard product.

[0151] Specifically, in step 24), the current output O of the output gate is updated. (t) The state h is determined by the memory and processing of the adhesion coefficient data and driving state parameter data of the four wheels at time t-1. (t-1) and the input x at time t (t) The output of the output gate is obtained through the activation function σ and is expressed as follows:

[0152] O (t) =σ(W o h (t-1) +U o x (t) +b o )

[0153] Among them, W o U o b represents the weighting coefficient of the output gate. o This is the bias value for the output gate.

[0154] In step 24), the memory and processing state h of the adhesion coefficient data and driving state parameter data of the four wheels at the current moment is output. (t) h (t) It is determined by the product of two parts, the first part being the output O of the output gate obtained through the activation function σ. (t) The second part is C obtained through the activation function tanh. (t) , is represented as:

[0155] h (t) =O (t) ⊙tanh(Ct ).

[0156] Specifically, the output of the four-wheel independent steering vehicle's four-wheel tire lateral stiffness in step 25) is expressed as follows:

[0157] Output: {C lf C lr C rf C rr}=σ(Vh (t) +c)

[0158] Where V is the weight matrix connecting the adhesion coefficient data and driving state parameter data of the four wheels, representing the memory, processing state, and output; c is the bias term added to the weighted sum; C lf For the lateral stiffness of the left front wheel, C lr For the left rear wheel lateral stiffness, C rf For the right front wheel lateral stiffness, C rr This refers to the lateral stiffness of the right rear wheel.

[0159] 3) Establish a nominal system model based on the estimated results of the four-wheel tire lateral stiffness, output the optimal nominal system model control law, establish an actual system model based on the nominal system model and the uncertainty of uneven snow depth on the road surface during vehicle driving, and determine the input and output quantities of trajectory tracking control.

[0160] The specific steps of step 3) are as follows:

[0161] 31) Establish a nominal system model based on the estimation results of the four-wheel tire lateral stiffness. By generating the cost function, obtain the optimal nominal system model control sequence. Take the first term of the sequence and output the optimal nominal system model control law.

[0162] 32) Based on the nominal system model and the uncertainty of uneven snow depth on the road surface, establish the actual system model and determine the input and output quantities of trajectory tracking control.

[0163] Specifically, the nominal system model in step 31) is as follows:

[0164]

[0165] in, The trajectory of the nominal system model, The deviation between the nominal system model output and the actual system model output. The deviation between the nominal system model input and the actual system model input is represented by A, which is the state matrix containing the lateral stiffness of the four tires, and B is the input matrix containing the lateral stiffness of the four tires.

[0166] Specifically, the cost function in step 31) is defined as:

[0167]

[0168] Among them, the objective function constraint is U is a constraint on the function. For additional mandatory constraints, the constant N p and N c This represents the prediction range and control range, where k represents time. and These are weighted diagonal matrices representing the state variables and input variables, respectively. For the state dimension, The input dimension is defined by the relaxation factor ε and the weighting coefficient κ, which are used to balance constraints and avoid infeasible optimization problems caused by constraints. When the constraints are exceeded, the relaxation variable ε becomes positive and expands the input constraints of the nominal system model. The feasible range; otherwise, the slack variable ε is set to 0.

[0169] Specifically, the optimal nominal system model control law in step 31) is expressed as:

[0170]

[0171] Among them, U * (k) is the optimal control sequence for the nominal system model. The optimal nominal system model control law is given by the corresponding optimal nominal system model trajectory. The cost function is J * (k).

[0172] Specifically, the actual system model in step 32) is represented as follows:

[0173]

[0174] Where μ is the input quantity for trajectory tracking control, μ = ( δ lf ,δ lr ,δ rf ,δ rr ) T , For the output quantity of trajectory tracking control, Let be the heading angle of the four-wheel independent steering vehicle, Y be the lateral position of the four-wheel independent steering vehicle, and v be the yaw angle. x Let ω be the longitudinal velocity of the four-wheel independent steering vehicle, and ω be the yaw rate of the four-wheel independent steering vehicle. This indicates the uncertainty caused by the uneven depth of snow accumulation on the road surface.

[0175] 4) Using the trajectory of the nominal system model as the center, a pipeline is generated in combination with the uncertainty of the icy and snowy road conditions, and the pipeline control law is output; the optimal nominal system model control law and the pipeline control law are combined to form the final control input of the actual system model, so as to realize the trajectory tracking of the four-wheel independent steering vehicle on the icy and snowy road.

[0176] The specific steps of step 4) are as follows:

[0177] 41) Define the uncertainty caused by uneven snow depth on the road surface;

[0178] 42) Based on the uncertainty, generate a pipeline centered on the trajectory of the nominal system model;

[0179] 43) Design the pipeline control law, and use the output of the pipeline control law to keep the trajectory of the actual system model within the generated pipeline;

[0180] 44) Combining the optimal nominal system model control law and pipeline control law, the final control input of the actual system model is formed.

[0181] Specifically, the uncertainty caused by the uneven snow depth on the road surface in step 41) The expression is as follows:

[0182]

[0183] Where ∈(k) is the error between the nominal system model and the actual system model, and η(k) represents the pipeline control law.

[0184] Specifically, in step 42), the pipe Ω centered on the trajectory of the nominal system model. tube The expression is as follows:

[0185]

[0186] Among them, C e It is a constant matrix. This represents the maximum range of uncertainty in the unevenness of snow depth on the road surface.

[0187] Specifically, the pipeline control law expression in step 43) is as follows:

[0188]

[0189] Where η(k) is the pipeline control law, and t represents the current time. λ represents the uncertainty of uneven snow depth on the road surface, and λ represents the pipeline parameters, which are determined by deviations from the nominal system model.

[0190] Specifically, the final control input expression of the actual system model in step 44) is:

[0191]

[0192] in, Let η(k) be the optimal nominal system model control law, and η(k) be the pipeline control law. This serves as the final control input for the actual system model.

[0193] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A trajectory tracking system for a four-wheel independent steering vehicle on icy and snowy roads, characterized in that, include: Four-wheel tire lateral stiffness estimation unit and trajectory tracking control unit; The four-wheel tire lateral stiffness estimation unit includes: an ice and snow road environment data acquisition module, a vehicle driving state parameter acquisition module, and an estimation module; The ice and snow road surface environment data acquisition module is used to collect ice and snow road surface information data to obtain the adhesion coefficient of the four wheels of the four-wheel independent steering vehicle. The vehicle driving status parameter acquisition module is used to collect vehicle driving status parameter data. The estimation module is used to estimate the lateral stiffness of the four tires using the adhesion coefficients of the four wheels and vehicle driving state parameter data. The trajectory tracking control unit includes: a nominal system module, an actual system module, a pipeline construction module, and a control module; The nominal system module is used to establish a nominal system model based on the estimation results of the lateral stiffness of the four-wheel tires and output the nominal system model control law; The actual system module is used to establish an actual system model based on the nominal system model and the uncertainties caused by the unevenness of snow depth on the road surface; The pipeline construction module is used to generate a pipeline centered on the trajectory of the nominal system model and taking into account the uncertainty of the icy and snowy road conditions. This ensures that the trajectory of the actual system model remains within the pipeline despite the uncertainty caused by the uneven snow depth on the road surface, and outputs the pipeline control law. The control module is used to generate the final control input of the actual system model based on the nominal system model control law and pipeline control law, thereby minimizing the trajectory tracking error.

2. The trajectory tracking system for a four-wheel independent steering vehicle on icy and snowy roads according to claim 1, characterized in that, The nominal system module establishes a nominal system model, and based on the established nominal system model, establishes a cost function with the goal of minimizing the tracking error, obtains the optimal nominal system model control sequence, takes the first term of the sequence, and outputs the optimal nominal system model control law.

3. The trajectory tracking system for a four-wheel independent steering vehicle on icy and snowy roads according to claim 1, characterized in that, The actual system module establishes an actual system model. Compared with the nominal system model, the actual system model includes the uncertainty of uneven snow depth on the road surface. The input and output quantities of trajectory tracking control are determined through the actual system model.

4. A trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads, based on the system described in any one of claims 1-3, characterized in that, The steps are as follows: 1) Collect information data on icy and snowy road surfaces and vehicle driving status parameters during vehicle operation, and use the information data on icy and snowy road surfaces to obtain the adhesion coefficients of the four wheels of the four-wheel independent steering vehicle. 2) The adhesion coefficients of the four wheels and the vehicle driving state parameter data are used as inputs to a bidirectional long short-term memory neural network to estimate the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle. 3) Establish a nominal system model based on the estimation results of the four-wheel tire lateral stiffness, output the optimal nominal system model control law, establish an actual system model based on the nominal system model and the uncertainty of the unevenness of snow depth on the road surface during vehicle driving, and determine the input and output quantities of trajectory tracking control. 4) Using the trajectory of the nominal system model as the center, a pipeline is generated in combination with the uncertainty of the icy and snowy road conditions, and the pipeline control law is output. The optimal nominal system model control law and the pipeline control law are combined to form the final control input of the actual system model, so as to realize the trajectory tracking of the four-wheel independent steering vehicle on the icy and snowy road.

5. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 4, characterized in that, In step 1), the road detection sensor takes pictures to obtain image information of the road ahead, and the image information is used to compare the road surface conditions under different adhesion coefficients to obtain the adhesion coefficients of the four wheels.

6. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 4, characterized in that, The bidirectional long short-term memory neural network in step 2) includes: an input gate, a forget gate, and an output gate; the input gate includes a forward input gate and a backward input gate. The forward input gate combines the immediate impact of the current icy and snowy road conditions on the lateral stiffness of the four-wheel tires with the impact of the icy and snowy road surface on the lateral stiffness of the four-wheel tires over a longer period of time, adding or deleting relevant information on the changes in the lateral stiffness of the four-wheel tires. The reverse input gate is added to obtain the adhesion coefficient or driving state parameter data that was missed under the icy and snowy road conditions. The forget gate is used to control the deletion of redundant adhesion coefficient or driving state parameter data of the four wheels. Finally, the output gate outputs the data.

7. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 6, characterized in that, The specific steps for estimating the lateral stiffness of the four-wheel tires in step 2) are as follows: 21) When the icy and snowy road conditions change during the vehicle's operation, the forget gate forgets the adhesion coefficients of the four wheels or the vehicle's driving state parameter data from the previous moment, re-enters the current adhesion coefficients and driving state parameter data of the four wheels, and updates the output of the forget gate. 22) The forward input gate uses the adhesion coefficients of the four wheels to adjust the influence weights of the adhesion coefficients and driving state parameter data of the four wheels, and obtains the adhesion coefficient data or driving state parameter data that were missed on the icy and snowy road surface through the reverse input gate to update the input of the bidirectional long short-term memory neural network. 23) Based on the input of the forget gate and the bidirectional long short-term memory neural network, update the adhesion coefficient and driving state parameter data accumulation of the four wheels for the current time-to-time four-wheel tire lateral stiffness estimation; 24) Update the current output of the output gate, and combine the current adhesion coefficient and driving state parameter data of the four wheels to output the memory and processing status of the four-wheel tire lateral stiffness at the current moment. 25) The bidirectional long short-term memory neural network outputs the lateral stiffness of the four-wheel tires of a four-wheel independent steering vehicle.

8. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 4, characterized in that, The specific steps of step 3) are as follows: 31) Establish a nominal system model based on the estimation results of the four-wheel tire lateral stiffness. By generating the cost function, obtain the optimal nominal system model control sequence. Take the first term of the sequence and output the optimal nominal system model control law. 32) Based on the nominal system model and the uncertainty of uneven snow depth on the road surface, establish the actual system model and determine the input and output quantities of trajectory tracking control.

9. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 8, characterized in that, The nominal system model in step 31) is as follows: in, The trajectory of the nominal system model, The deviation between the nominal system model output and the actual system model output. The deviation between the nominal system model input and the actual system model input is represented by A, which is the state matrix containing the lateral stiffness of the four tires, and B is the input matrix containing the lateral stiffness of the four tires. The cost function in step 31) is defined as follows: Among them, the objective function constraint is U is a constraint on the function. For additional mandatory constraints, the constant N p and N c This represents the prediction range and control range, where k represents time. and These are weighted diagonal matrices representing the state variables and input variables, respectively. For the state dimension, The input dimension is defined by the relaxation factor ε and the weighting coefficient κ, which are used to balance constraints and avoid infeasible optimization problems caused by constraints. When the constraints are exceeded, the relaxation variable ε becomes positive and expands the input constraints of the nominal system model. The feasible range; otherwise, the slack variable ε is set to 0; The optimal nominal system model control law in step 31) is expressed as: Among them, U * (k) is the optimal control sequence for the nominal system model. The optimal nominal system model control law is given by the corresponding optimal nominal system model trajectory. The cost function is J * (k); The actual system model in step 32) is represented as follows: Where μ is the input quantity for trajectory tracking control, μ = ( δ lf ,δ lr ,δ rf ,δrr) T , For the output quantity of trajectory tracking control, Let be the heading angle of the four-wheel independent steering vehicle, Y be the lateral position of the four-wheel independent steering vehicle, and v be the yaw angle. x ω is the longitudinal velocity of the four-wheel independent steering vehicle, and ω is the yaw rate of the four-wheel independent steering vehicle. This indicates the uncertainty caused by the uneven depth of snow accumulation on the road surface.

10. The trajectory tracking method for a four-wheel independent steering vehicle on icy and snowy roads according to claim 4, characterized in that, The specific steps of step 4) are as follows: 41) Define the uncertainty caused by uneven snow depth on the road surface; 42) Based on the uncertainty, generate a pipeline centered on the trajectory of the nominal system model; 43) Design the pipeline control law, and use the output of the pipeline control law to keep the trajectory of the actual system model within the generated pipeline; 44) Combining the optimal nominal system model control law and pipeline control law, the final control input of the actual system model is formed.

Images