Real-time tracking data guided learning control method for unmanned autonomous vehicle

Abstract

The present application relates to the technical field of control, in particular to a real-time tracking data guided learning control method for unmanned autonomous vehicle, comprising the following steps: determining a desired target trajectory of the unmanned autonomous vehicle; determining nonlinear time-varying dynamic characteristics of a corresponding system of the unmanned autonomous vehicle; determining a motion control parameter of the unmanned autonomous vehicle, and initializing the motion control parameter; obtaining a tracking trajectory of the unmanned autonomous vehicle; performing instruction filtering on the desired target trajectory based on an attraction rule; updating the current motion control parameter by using data guided learning based on the instruction filtered desired target trajectory, the tracking trajectory and the current motion control parameter; performing iterative updating until completing complete real-time tracking of the desired target trajectory of the unmanned autonomous vehicle; and the present application can accurately track the unmanned autonomous vehicle with time-varying nonlinear dynamic motion characteristics in real time.

Description

Technical Field

[0001] This invention relates to the field of control technology, specifically to a real-time tracking data-guided learning control method for unmanned autonomous vehicles. Background Technology

[0002] Currently, the real-time tracking and control problem of autonomous vehicles has received widespread attention in practical applications. With the increasing complexity of autonomous vehicle structures and the rising demands for autonomous control, designing high-precision control algorithms for autonomous vehicles using intelligent control methods has become a trend. For the tracking and control objective of autonomous vehicles, most intelligent control methods can only complete the steady-state tracking task. While control methods such as finite / fixed-time control or finite-time predetermined performance / predetermined performance control strive to improve transient performance or accuracy in repetitive tasks, they may not fully meet the dual requirements of control accuracy and control effort. These methods prioritize achieving specific performance criteria or convergence within a defined time range, which is valuable, but may be insufficient in handling the dynamic balance between accuracy and the expenditure of control effort.

[0003] Finite-time / fixed-time control methods are effective in certain situations, but they are not without limitations. They cannot guarantee a uniform frequency across different system components, nor do they possess the ability to ensure the entire system operates without delay within a finite time. Including fractional exponent terms in the controller can lead to numerous singularity problems during repeated differentiation, resulting in complex solutions detrimental to stability and control system design. Furthermore, the abundance of fractional exponent terms complicates parameter selection. Finite-time predetermined performance / predetermined performance control also presents challenges. This approach places high demands on the system output and tracking signal. If the system cannot consistently remain within the limits of the performance function, it becomes unresponsive. The requirement to meet the performance function limits from the outset, coupled with the requirement for high control gain, can make many systems unable to withstand large transient drives, and it is fragile because any sudden disturbance causing the limited term to exceed the performance function boundaries can cause the entire control method to fail. The existence of predefined function constraints further complicates parameter selection, especially in practical systems where fixed parameters cannot be arbitrarily changed. Moreover, the discontinuities in piecewise performance functions hinder smoothness, making it difficult to derive and prove the continuity of Lyapunov functions. In the analysis of continuous systems, it is difficult to rigorously prove piecewise Lyapunov functions. Summary of the Invention

[0004] In view of the above problems, the present invention provides a real-time tracking data-guided learning control method for unmanned autonomous vehicles, which solves the problem of real-time tracking of unmanned autonomous vehicles with time-varying nonlinear dynamic motion characteristics in the prior art.

[0005] On the one hand, this invention provides a real-time tracking data-guided learning control method for unmanned autonomous vehicles, comprising the following steps:

[0006] Step S1: Determine the desired target trajectory of the unmanned autonomous vehicle; determine the nonlinear time-varying dynamic characteristics of the unmanned autonomous vehicle; determine the motion control parameters of the unmanned autonomous vehicle, and initialize the motion control parameters.

[0007] Step S2: Based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, control the unmanned autonomous vehicle to track the desired target trajectory and obtain the tracking trajectory of the unmanned autonomous vehicle;

[0008] Step S3: Perform instruction filtering on the desired target trajectory based on the attraction rule; based on the instruction-filtered desired target trajectory, the tracking trajectory, and the current motion control parameters, update the current motion control parameters using data-guided learning;

[0009] Step S4: Return to step S2 and perform iterative updates until the desired target trajectory of the unmanned autonomous vehicle is fully tracked in real time.

[0010] Preferably, determining the desired target trajectory of the unmanned autonomous vehicle in step S1 specifically involves: setting y d (t) represents the expected target trajectory of the unmanned autonomous vehicle, where t is a positive integer representing the running time;

[0011] The nonlinear time-varying dynamic characteristics mentioned in step S1 are as follows:

[0012]

[0013]

[0014] y k =x 1,k

[0015] Where, x i,k Let i be the i-th state of the unmanned autonomous vehicle in the k-th iteration. For state x i,k The derivative with respect to time, where i is the ordinal number of the state variable, n is the total number of state variables, and k∈Z. + Z represents the number of iterations. + U is the set of positive integers. k This is the control input for the system in the k-th iteration; f represents the state variables of the system from order 1 to i; i , Let y be the time-varying system function for the i-th state of the unmanned autonomous vehicle, ∈ be an additional time-varying system function. kLet x be the position state of the system in the k-th iteration. i,k , u k , f i , ∈、y k All of them are time variables related to the running time t.

[0016] Preferably, the motion control parameters include: actual motion state x i,k Target motion state θ i,k Control and adjust parameter c i Control input u k Time-varying system function estimator Estimated parameter δ i,k estimator System nonlinear parameter ζ n,k Set the initial values ​​of all parameters in the motion control parameters to 0 to complete the initialization.

[0017] Preferably, step S3 specifically includes: step S3-1, performing command filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the command filtering value; step S3-2, updating the learning estimation law parameters, including: obtaining the actual motion state from the current position, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state; step S3-3, calculating and updating the control input of the unmanned autonomous vehicle based on the command filtering value and motion control parameters of the unmanned autonomous vehicle.

[0018] Preferably, step S3-1 specifically includes: obtaining the target motion state θ from the desired target trajectory. i,k ; Perform filtering operation on the target motion state: Set the attraction rule and adjust the filtering error using the following set of equations:

[0019]

[0020]

[0021]

[0022] Where, θ i,k It is the target value of the state variable. ω is the derivative of the target value of the state variable, used as the instruction filter value. i α is the filter bandwidth. i,k For the function corresponding to the virtual controller, c i Let η be the gain of the i-th order controller. i,kLet be the state variable of the filter error compensation signal of order i. Let be the derivative of the state variable of the filter error compensation signal of order i.

[0023] Preferably, step S3-2 specifically includes: obtaining the actual motion state x from the current position. i,k The learning law is calculated using the following data-guided approach. Perform iterative updates:

[0024]

[0025]

[0026] in, For the learning law projection operator of the time-varying system function ∈ value, l i For adaptive parameters, κ i,k The calculation method is κ i,k =z i,k -η i,k where i = 1, ..., n, To estimate the parameter δ i,k The learning law projection operator, ζ i,k The nonlinear parameters of the system are calculated as follows: Where i = 1, ..., n, and j is the summation index. Let x be the summation weight of the j-th state variable in the k-th iteration. j,k Let ρ be the j-th state variable in the k-th iteration, tanh be the hyperbolic tangent function, and ρ be the tangent. k is the scaling factor for the k-th iteration.

[0027] Preferably, step S3-3 specifically includes: determining the actual motion state x i,k and the target motion state θ i,k Calculate the tracking error z i,k Then, based on the instruction filter value of the unmanned autonomous vehicle... The tracking error z i,k The control adjustment parameter c i Actual motion state x i,k The updated time-varying system function estimate obtained in step S3-2 Updated estimated parameter δ n,k estimator and the system nonlinear parameter ζ i,k Perform calculations to update the control input u of the unmanned autonomous vehicle. k .

[0028] Preferably, the updated control input u of the unmanned autonomous vehicle k The calculation method is as follows:

[0029]

[0030] Among them, z n,k =x n,k -θ n,k Let z be the error of the system variables in the nth state during the kth iteration. When n=1, z 1,k =y k -y d Let c be the position error variable of the system in the k-th iteration. n For the nth control parameter, Let be the estimate of the value of the time-varying system function ∈ under the k-th iteration. The estimated parameter δ for the nth state in the kth iteration n,k The estimate.

[0031] On one hand, the present invention provides a real-time tracking data-guided learning control system for unmanned autonomous vehicles, characterized in that it includes:

[0032] The module determines the desired target trajectory of the unmanned autonomous vehicle; determines the nonlinear time-varying dynamic characteristics of the unmanned autonomous vehicle; determines the motion control parameters of the unmanned autonomous vehicle, and initializes the motion control parameters.

[0033] The acquisition module, based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, controls the unmanned autonomous vehicle to perform tracking for the entire cycle in the current iteration, and acquires the tracking trajectory of the unmanned autonomous vehicle.

[0034] The adjustment module performs instruction filtering on the desired target trajectory based on the attraction rule; based on the instruction-filtered desired target trajectory, the tracking trajectory, and the current motion control parameters, it updates the current motion control parameters using guided learning;

[0035] The update module is used to perform iterative updates until the desired target trajectory of the unmanned autonomous vehicle is fully tracked in real time.

[0036] Preferably, the adjustment module includes:

[0037] The instruction filtering unit is used to perform instruction filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the instruction filtering value;

[0038] The data-guided learning unit is used to update the learning estimation law parameters, including: obtaining the actual motion state from the tracking trajectory, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state;

[0039] The control input calculation unit is used to calculate and update the control input of the unmanned autonomous vehicle based on the instruction filter value and motion control parameters of the unmanned autonomous vehicle.

[0040] Compared with the prior art, the present invention has at least the following beneficial effects:

[0041] (1) The control strategy of this invention can adaptively adjust the control input of the unmanned autonomous vehicle based on the real-time feedback of tracking error and system state, thereby improving the system's adaptability to uncertainties and external disturbances. This enables the unmanned autonomous vehicle to maintain stable trajectory tracking performance in various complex environments.

[0042] (2) This invention employs data-guided learning control, which effectively identifies and compensates for dynamic errors occurring during trajectory tracking. The system learns from data in each iteration, reducing tracking errors and continuously optimizing control inputs to finely adjust the vehicle's path, thereby more accurately tracking the predetermined trajectory and ultimately achieving fully real-time tracking. Unlike current popular finite-time and predefined-performance control technologies, this invention effectively avoids various limitations on system and parameter adjustments, thus providing a more flexible and effective solution.

[0043] (3) By using an instruction filter based on attraction rules, the present invention avoids directly calculating higher-order partial derivatives, so that the error will not gradually increase during the traction process, and simplifies the design of the control law, which can improve the calculation efficiency and make the control strategy easier to implement and debug. Attached Figure Description

[0044] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0045] Figure 1 A flowchart of a real-time tracking data-guided learning control method for unmanned autonomous vehicles provided in an embodiment of the present invention;

[0046] Figure 2 This is a structural diagram of a real-time tracking data-guided learning control system for unmanned autonomous vehicles provided in an embodiment of the present invention.

[0047] Figure 3 Another structural diagram of the real-time tracking data-guided learning control system for unmanned autonomous vehicles provided in an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram illustrating the dynamic evolution of the position and expected target trajectory of an unmanned autonomous vehicle along the time axis before and after data learning, as provided in an embodiment of the present invention. Detailed Implementation

[0049] To better understand the above-described objectives, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other. Furthermore, the present invention can be implemented in other ways different from those described herein; therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0050] To illustrate the effectiveness of the method proposed in this invention, the following detailed description of the above technical solution is provided through a specific embodiment, such as... Figure 1 As shown, a real-time tracking data-guided learning control method for unmanned autonomous vehicles is disclosed, and the specific implementation steps are as follows:

[0051] Step S1: Determine the desired target trajectory of the unmanned autonomous vehicle; determine the nonlinear time-varying dynamic characteristics of the unmanned autonomous vehicle; determine the motion control parameters of the unmanned autonomous vehicle, and initialize the motion control parameters.

[0052] The desired target trajectory represents the reference trajectory that the autonomous vehicle ultimately hopes to follow. The desired target trajectory is a pre-set value, which is used as input information preset by the human based on the tracking requirements.

[0053] In some embodiments, the preset method for the desired target trajectory may include methods such as manually setting function values, map-based path planning, dynamic obstacle avoidance algorithm to generate trajectory, smooth path to generate trajectory considering vehicle dynamics constraints, or trajectory planning based on model predictive control (MPC). By presetting the desired target trajectory, the motion characteristics of the unmanned autonomous vehicle can be reflected, providing a reference for subsequent target trajectory tracking.

[0054] In some embodiments, y is set d (t) represents the expected target trajectory of the unmanned autonomous vehicle, where t∈1,2,…

[0055] The unmanned autonomous vehicle system in this embodiment has the following time-varying dynamic real-time tracking characteristics:

[0056]

[0057]

[0058] y k (t)=x 1,k (t) t∈0,1,2,...

[0059] Where, x i,k (t) represents the i-th state of the unmanned autonomous system at time t in the k-th iteration. For state xi,k (t) is the derivative with respect to time, where i represents the ordinal number of the state variable, n represents the total number of state variables, t is the running time, and k∈Z. + Z represents the number of iterations. + Represents the set of positive integers, u k (t) represents the control input of the system at time t in the kth iteration; f represents the system state variables from order 1 to i; i (t), Let y be the time-varying system function corresponding to the i-th state of the autonomous vehicle at time t, and let ∈(t) be an additional time-varying system function corresponding to time t. k Let (t) represent the position state of the system at time t in the k-th iteration. Without loss of generality, let the initial state of the autonomous vehicle be...

[0060] The time-varying dynamic characteristics of the unmanned autonomous vehicle system targeted by this invention were determined through the above method. The model consists of a set of differential equations, where for the i-th state variable x... i,k (t), its time derivative The (i+1)th state variable x i+1,k (t), nonlinear function f i and time-varying disturbance terms Composition. Among them, f i (t) is a function of the current state and all previous states, and It is a time-dependent perturbation term; for the last state variable x n,k (t), its time derivative By control input u k (t), nonlinear function f n and time-varying disturbance terms Composition. This model allows for a detailed description of the variable relationships that govern the dynamic characteristics of the system as they change over time.

[0061] The signals that an autonomous vehicle system can obtain are information that the system can directly measure or observe. In some embodiments, these signals may include physical state information such as the vehicle's position, speed, and acceleration. These signals can be obtained through sensors (such as GPS receivers, radar, cameras, etc.) and used as input data for the control system.

[0062] The process of obtaining the corresponding state feedback signal can be described as processing and transforming the original signal to obtain the system state, including the actual motion state x. i,k Target motion state θ i,k Control and adjust parameter c i Used to determine how the system state affects the control input, and thus derive the control input u.k The control input directly affects the movement of the unmanned autonomous vehicle, adjusting its position.

[0063] For the motion control parameters of unmanned autonomous vehicles, the motion control parameters of unmanned autonomous vehicles determined by this invention include parameters reflecting the system's motion state: actual motion state x i,k Target motion state θ i,k The parameters that reflect the real-time feedback of motion by the control system: control adjustment parameter c i Control input u k The iterative parameters reflecting the data-guided learning process: time-varying system function estimator Estimated parameters estimators The nonlinear parameter ζ of the system after smooth approximation n,k In this step, these parameters are initialized. Subsequent steps will iteratively update these parameters until real-time tracking of the target trajectory is completed.

[0064] In some embodiments, the initial values ​​of each parameter in the above motion control parameters are set to 0 to complete the initialization of these parameters.

[0065] Step S2: Based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, control the unmanned autonomous vehicle to track the desired target trajectory and obtain the tracking trajectory of the unmanned autonomous vehicle.

[0066] In some embodiments, the position of the unmanned autonomous vehicle is adjusted according to the nonlinear time-varying dynamic characteristics and the control input in the current motion control parameters to complete the tracking of the entire cycle in the current iteration and obtain the tracking trajectory of the unmanned autonomous vehicle.

[0067] In some embodiments, the current position of an autonomous vehicle can be obtained by installing sensor devices on the vehicle. These sensors may include GPS receivers, inertial measurement units, and optical flow sensors, providing detailed data about the vehicle's position.

[0068] In some embodiments, after acquiring the real-time location information of the autonomous vehicle, the tracking error for the current iteration can be calculated. The real-time location data is compared with a predetermined path to determine the deviation between the actual location of the autonomous vehicle and its intended location. This deviation is taken as the position tracking error.

[0069] In some embodiments, the position tracking error can be estimated using simple geometric or algebraic operations. For example, the tracking error can be estimated by calculating the Euclidean distance between the current position of the autonomous vehicle and the nearest point on the predetermined path. Furthermore, considering that the autonomous vehicle may deviate from the predetermined path due to various external factors (such as changes in wind speed, terrain undulations, etc.), more complex algorithms can be used to optimize the estimation of the tracking error, thereby improving the navigation accuracy and reliability of the autonomous vehicle.

[0070] In some embodiments, the position tracking error z in the k-th iteration 1,k The calculation method is as follows: z 1,k =y k -y d The tracking error is used for calculations in subsequent steps.

[0071] Step S3: Perform instruction filtering on the desired target trajectory based on the attraction rule; based on the instruction-filtered desired target trajectory, the tracking trajectory, and the current motion control parameters, update the current motion control parameters using data-guided learning;

[0072] In specific implementation, in the control method provided in the embodiments of the present invention, step S3 specifically includes the following steps:

[0073] Step S3-1: Perform instruction filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the instruction filter value;

[0074] To avoid the problem of exploding computational complexity of higher-order partial derivatives, this invention introduces instruction filtering operations. in The calculation represents the instruction filtering operation introduced to avoid the problem of the explosion of computational complexity of higher-order partial derivatives. The instruction filtering operation will have dynamic errors.

[0075] To compensate for the error of this filter, an attractive force rule is designed and adjusted using the following set of equations:

[0076]

[0077]

[0078]

[0079] Where, θ i,k It is the target value of the state variable. ω is the derivative of the target value of the state variable, used as the instruction filter value. i α is the filter bandwidth. i,k For the function corresponding to the virtual controller, ci Let η be the gain of the i-th order controller. i,k Let be the state variable of the filter error compensation signal of order i. Let be the derivative of the state variable of the filter error compensation signal of order i.

[0080] The attraction rule is used to design the filter error so that its change during dynamic processes tends to decrease, similar to a traction mechanism. Without the attraction rule, filter error cannot be effectively limited, and convergence analysis of the filter error can only be obtained after the design is completed. In this calculation, a virtual controller with parameter α is introduced. i,k In a nonlinear backstepping design, there are n-1 virtual controllers. The control values ​​of the real controller are finally obtained through the cascading design of coordinate transformation.

[0081] In some embodiments, the individual controller gains c1,...,c i These together constitute the control and adjustment parameters. By adjusting the controller state and gain in the above manner, the error of the unmanned vehicle during its movement is prevented from gradually increasing and can be effectively controlled and reduced.

[0082] Step S3-2: Update the learning estimation law parameters, including: obtaining the actual motion state from the tracking trajectory of the current iteration, and updating the time-varying system function estimator, the estimated parameter estimator, and the system nonlinear parameters based on the actual motion state;

[0083] This invention is based on time-varying system function estimators. Estimated parameters estimators Guided adaptive adjustment to calculate the vehicle's control input u k In updating u k First of all, Iterative updates will be performed.

[0084] The data-guided learning control learning estimation law is:

[0085]

[0086]

[0087] in, For the learning law projection operator of the time-varying system function ∈ value, l i For adaptive parameters, κ i,k The calculation method is κ i,k =z i,k -η i,k Where i = 1, ..., n, represents the compensated coordinate transformation obtained by coordinate transformation and attraction rules. To estimate the parameter δ i,k The learning law projection operator, the system nonlinear parameter ζ i,k The calculation method is as follows: Where i = 1, ..., n represents the nonlinear function of the system after approximation smoothing using the hyperbolic tangent function, and j is the summation index. Let x be the summation weight of the j-th state variable in the k-th iteration. j,k Let ρ be the j-th state variable in the k-th iteration, tanh be the hyperbolic tangent function, and ρ be the tangent. k It is the scaling factor for the k-th iteration, used to adjust the rate of change of the inflection point of the smooth function.

[0088] By using the above method, the parameters of the current iteration are updated based on the previous estimates. and By conducting data-guided learning, the control system of unmanned autonomous vehicles can achieve adaptive adjustments to better cope with the impact of environmental changes and internal system dynamics.

[0089] Step S3-3: Calculate and update the control input of the unmanned autonomous vehicle based on the command filter value and motion control parameters.

[0090] The calculation process is as follows: the tracking error is calculated from the actual motion state and the target motion state; then, the control input of the unmanned autonomous vehicle is updated based on the command filter value of the unmanned autonomous vehicle, the tracking error, the control adjustment parameters, the actual motion state, the updated time-varying system function estimate obtained in step S3-2, the updated estimate of the estimated parameters, and the updated system nonlinear parameters.

[0091] Based on the system matrix corresponding to the unmanned autonomous vehicle, its corresponding coordinate transformation z is obtained. 1,k =y k -y d ,zi,k=xi,k-θi,k。 where, the coordinate transformation z1,k=yk-yd is the position error variable of system t in the k-th iteration, and the coordinate transformation z i,k =x i,k -θ i,k Let be the system position error variable in the i-th state and k-th iteration.

[0092] In some embodiments, the system matrix is ​​determined based on the dynamics and control system model of the unmanned autonomous vehicle, and includes various parameters and state variables of the vehicle's motion. Coordinate transformation z 1,k =y k -y d This represents the difference between the vehicle's actual position and the target position, while z... i,k =xi,k -θ i,k This represents the difference between the actual and target values ​​of other state variables of the vehicle (such as speed, acceleration, etc.). This coordinate transformation is for the convenience of subsequent error analysis and control strategy design.

[0093] The control input for the autonomous vehicle in this iteration is determined by the following formula:

[0094]

[0095] Among them, z n,k =x n,k -θ n,k Let z be the error of the system variables in the nth state during the kth iteration. When n=1, z 1,k =y k -y d Let c be the position error variable of the system in the k-th iteration. n This is the nth gain coefficient, used to adjust the control strength. Let be the estimate of the value of the time-varying system function ∈ under the k-th iteration. The estimated parameter δ for the nth state in the kth iteration n,k The estimated quantity. It should be noted that for z in control design... i,k ,x i,k , θ n,k The values ​​of these functions are time variables related to t, and t is omitted in the representation.

[0096] Step S4: Return to step S2 for the next iteration until complete real-time tracking of the target trajectory is achieved, i.e., lim k→∞ y k (t)=y d (t), t∈{1,2,…}.

[0097] It should be noted that in the control method provided in the embodiments of the present invention, the execution order of each step in step S1 is not fixed and is not limited here.

[0098] The real-time tracking iterative learning control system and method for time-varying unmanned autonomous vehicles provided in this embodiment of the invention can overcome the influence of the time-varying dynamics of unmanned autonomous vehicles, enabling unmanned autonomous vehicles with unknown nonlinear time-varying characteristics to track trackable target trajectories in real time at all moments within a finite time period, and has extremely high tracking accuracy.

[0099] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0100] The following is a simulation case of a real-time high-precision tracking iterative learning control system and method for unmanned autonomous vehicles under disturbance, provided by this invention. Consider a time-varying unmanned autonomous vehicle with the following nonlinear dynamics:

[0101]

[0102]

[0103] Among them, the system parameters are

[0104] ∈=1+0.2sin(t)

[0105] f1(x 1,k )=cos(t)·x 1,k ,

[0106] f2(x 1,k ,x 2,k )=sin(t)[x 1,k ·cos(x 1,k )+x 2,k ·sin(x 2,k )].

[0107] f1 and f2 represent the nonlinear components of the system, introducing nonlinear dynamic behavior, and ∈ = 1 + 0.2sin(t) represents the time-varying disturbance intensity of the system. and It is an additional nonlinear function, further increasing the complexity of the system.

[0108] Considering an operating cycle of T=10, and selecting the initial state of the unmanned autonomous vehicle as... and the initial input is Perform the first iteration k=0.

[0109] Step S10: Determine the expected target trajectory of the unmanned autonomous vehicle as y. d (t)=sin 3 (t), t∈{1,2,…,10}. This step defines the expected path that the autonomous vehicle needs to track within a certain time period, from 1 to 10 seconds.

[0110] Step S20: Obtain the system matrix corresponding to the unmanned autonomous vehicle, and find that the corresponding system function satisfies the following relationship:

[0111]

[0112]

[0113] Where λ1 and λ2 are constants, representing the tightness of the inequality.

[0114] Step S30: Obtain the position of the unmanned autonomous vehicle at the current iteration k. And calculate the tracking error in the current iteration.

[0115] Step S40: Update the control input of the autonomous vehicle for the next iteration based on the control input, tracking error, and selected control adjustment parameters under the current iteration, and control the autonomous vehicle to execute the next iteration process. Specifically, this includes the following steps:

[0116] Step S401: Determine the control input of the unmanned autonomous vehicle based on the control input and tracking error under the current iteration. The calculation method for the control input in this step is the same as the aforementioned method.

[0117] Step S402: Based on the control input data of the unmanned autonomous vehicle under the current iteration, the learning estimation law is adjusted according to the tracking error to obtain the learning estimation law for the next iteration. The controller is updated, and its own position is adjusted to ultimately achieve real-time tracking of the expected target trajectory, i.e., lim k→∞ y k (t)=y d (t), t∈{1,2,…,10}.

[0118] Step S50: Return to step S30 and run multiple iterations until real-time tracking of the target trajectory is completed.

[0119] To further illustrate the effectiveness of the real-time tracking iterative learning control system and method for time-varying unmanned autonomous vehicles provided by this invention, Figure 4 The figure illustrates the dynamic evolution of the autonomous vehicle's position and the expected target trajectory along the time axis after 10 iterations. It demonstrates that the autonomous vehicle's position output tracks the expected target trajectory in real time at any given moment with extremely high tracking accuracy. Therefore, the real-time high-precision tracking iterative learning control system and method for time-varying autonomous vehicles provided by this invention can overcome the influence of the time-varying dynamics of autonomous vehicles, enabling autonomous vehicles with unknown nonlinear dynamics to track trackable target trajectories in real time at all moments within a finite time period with extremely high tracking accuracy.

[0120] This invention also discloses a real-time high-precision tracking iterative learning control system for unmanned autonomous vehicles under interference, such as... Figure 2 As shown, it includes: Determine module 1, Obtain module 2, Adjust module 3, and Update module 4, specifically including:

[0121] Module 1 is used to determine the desired target trajectory of the unmanned autonomous vehicle; determine the nonlinear time-varying dynamic characteristics of the unmanned autonomous vehicle; determine the motion control parameters of the unmanned autonomous vehicle, and initialize the motion control parameters.

[0122] The acquisition module 2 is used to control the unmanned autonomous vehicle to track the desired target trajectory based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, and to acquire the tracking trajectory of the unmanned autonomous vehicle.

[0123] Adjustment module 3 is used to perform instruction filtering on the desired target trajectory based on the attraction rule; and to update the current motion control parameters using data-guided learning based on the instruction-filtered desired target trajectory, the tracking trajectory, and the current motion control parameters.

[0124] Update module 4 is used for iterative updates until the desired target trajectory of the unmanned autonomous vehicle is fully tracked in real time.

[0125] In some embodiments, such as Figure 3 As shown, adjustment module 3 specifically includes:

[0126] Command filtering unit 31 is used to perform command filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the command filtering value;

[0127] The data-guided learning unit 32 is used to update the learning estimation law parameters, including: obtaining the actual motion state from the current position, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state;

[0128] The control input calculation unit 33 is used to calculate and update the control input of the unmanned autonomous vehicle based on the instruction filter value and motion control parameters of the unmanned autonomous vehicle.

[0129] The real-time tracking data-guided learning control system for time-varying unmanned autonomous vehicles provided in this embodiment of the invention can overcome the influence of time-varying dynamics of unmanned autonomous vehicles, enabling time-varying unmanned autonomous vehicles with unknown nonlinear dynamics to track trackable target trajectories in real time at all moments within a finite time interval, and has extremely high tracking accuracy.

[0130] While the specific embodiments of the present invention depict actions or steps in a particular order, this should be understood as requiring such actions or steps to be performed in the shown specific order or sequential order, or requiring all illustrated actions or steps to be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations. The above descriptions are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

[0131] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A real-time tracking data-guided learning control method for unmanned autonomous vehicles, characterized in that, Includes the following steps: Step S1: Determine the desired target trajectory of the unmanned autonomous vehicle; determine the nonlinear time-varying dynamic characteristics of the unmanned autonomous vehicle; determine the motion control parameters of the unmanned autonomous vehicle, and initialize the motion control parameters. Step S2: Based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, control the unmanned autonomous vehicle to track the desired target trajectory and obtain the tracking trajectory of the unmanned autonomous vehicle; Step S3: Perform instruction filtering on the desired target trajectory based on the attraction rule; Based on the desired target trajectory after instruction filtering, the tracking trajectory, and the current motion control parameters, data-guided learning is used to update the current motion control parameters; Step S4: Return to step S2 and perform iterative updates until the desired target trajectory of the unmanned autonomous vehicle is fully tracked in real time. Step S3 specifically includes: Step S3-1: Perform instruction filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the instruction filter value; Step S3-2, updating the learning estimation law parameters, includes: obtaining the actual motion state from the tracking trajectory, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state; Step S3-3: Calculate and update the control input of the unmanned autonomous vehicle based on the command filter value and motion control parameters. Step S3-1 specifically includes: Obtain the target motion state from the desired target trajectory ; Perform a filtering operation on the target's motion state: , Set the attraction rule and adjust the filtering error using the following set of equations: in, It is the target value of the state variable. It is the derivative of the target value of the state variable, used as the instruction filter value. For the filter bandwidth, For the function corresponding to the virtual controller, For the gain of the i-th order controller, Let be the state variable of the filter error compensation signal of order i. Let be the derivative of the state variables of the filter error compensation signal of order i; Step S3-2 specifically includes: Obtain the actual motion state from the current position ; The following data-guided learning law calculation is used to calculate the pair. , Perform iterative updates: in, For time-varying system functions The learning law of projection operator, For adaptive parameters, The calculation method is as follows ,in , For estimating parameters The learning law projection operator, The nonlinear parameters of the system after smooth approximation are calculated as follows: ,in Where j is the summation index, Let be the summation weight of the j-th state variable in the k-th iteration. Let be the j-th state variable in the k-th iteration, and tanh be the hyperbolic tangent function. is the scaling factor for the k-th iteration.

2. The real-time tracking data-guided learning control method for unmanned autonomous vehicles according to claim 1, characterized in that, The specific steps in step S1 for determining the desired target trajectory of the unmanned autonomous vehicle are: setting... The expected target trajectory of the unmanned autonomous vehicle, among which A positive integer representing the running time; The nonlinear time-varying dynamic characteristics mentioned in step S1 are as follows: in, For unmanned autonomous vehicles in the first k The corresponding iteration in the nth iteration One state, For state The derivative with respect to time, Let be the ordinal number of the state variable. The total number of state variables. For the number of iterations, It is the set of positive integers. For the system in the first k Control input in the next iteration; Let represent the system state variables from order 1 to 1i; , For the first unmanned autonomous vehicle A time-varying system function with 1 state. For additional time-varying system functions, For unmanned autonomous vehicles in the first k The position state in the next iteration, the 、 、 、 、 、 、 、 All of them are time variables related to the running time t.

3. The real-time tracking data-guided learning control method for unmanned autonomous vehicles according to claim 2, characterized in that: The motion control parameters include: actual motion state. Target motion state Control and adjust parameters Control input Time-varying system function estimator Estimated parameters estimator System nonlinear parameters Set the initial values ​​of all parameters in the motion control parameters to 0 to complete the initialization.

4. The real-time tracking data-guided learning control method for unmanned autonomous vehicles according to claim 3, characterized in that, Step S3-3 specifically includes: From the actual state of motion and target motion state Calculate tracking error Then, based on the instruction filter value of the unmanned autonomous vehicle... The tracking error The control adjustment parameters Actual motion state The updated time-varying system function estimate obtained in step S3-2 Updated estimated parameters estimator and system nonlinear parameters Perform calculations to update the control inputs of the unmanned autonomous vehicle. .

5. The real-time tracking data-guided learning control method for unmanned autonomous vehicles according to claim 4, characterized in that, The updated control input for unmanned autonomous vehicles The calculation method is as follows: in, For the first The state of the first k The error of the system variables in the next iteration, when n=1, For the first k The position error variable of the system in the next iteration For the first One control adjustment parameter, For the first k Time-varying system function under the next iteration An estimate of the value For the first The state of the first k Estimated parameters in the next iteration The estimate.

6. A real-time tracking data-guided learning control system for unmanned autonomous vehicles, characterized in that, include: The module determines the desired target trajectory of the unmanned autonomous vehicle; Determine the nonlinear time-varying dynamic characteristics of unmanned autonomous vehicles; Determine the motion control parameters of the unmanned autonomous vehicle and initialize the motion control parameters; The acquisition module, based on the nonlinear time-varying dynamic characteristics and the current motion control parameters, controls the unmanned autonomous vehicle to track the desired target trajectory and acquires the tracking trajectory of the unmanned autonomous vehicle. The adjustment module performs instruction filtering on the desired target trajectory based on the attraction rule; Based on the desired target trajectory after instruction filtering, the tracking trajectory, and the current motion control parameters, guided learning is used to update the current motion control parameters; The update module is used to perform iterative updates until the desired target trajectory of the unmanned autonomous vehicle is fully tracked in real time. The processing steps of the adjustment module specifically include: Step S3-1: Perform instruction filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the instruction filter value; Step S3-2, updating the learning estimation law parameters, includes: obtaining the actual motion state from the tracking trajectory, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state; Step S3-3: Calculate and update the control input of the unmanned autonomous vehicle based on the command filter value and motion control parameters. Step S3-1 specifically includes: Obtain the target motion state from the desired target trajectory ; Perform a filtering operation on the target's motion state: , Set the attraction rule and adjust the filtering error using the following set of equations: in, It is the target value of the state variable. It is the derivative of the target value of the state variable, used as the instruction filter value. For the filter bandwidth, For the function corresponding to the virtual controller, For the gain of the i-th order controller, Let be the state variable of the filter error compensation signal of order i. Let be the derivative of the state variables of the filter error compensation signal of order i; Step S3-2 specifically includes: Obtain the actual motion state from the current position ; The following data-guided learning law calculation is used to calculate the pair. , Perform iterative updates: in, For time-varying system functions The learning law of projection operator, For adaptive parameters, The calculation method is as follows ,in , For estimating parameters The learning law projection operator, The nonlinear parameters of the system after smooth approximation are calculated as follows: ,in Where j is the summation index, Let be the summation weight of the j-th state variable in the k-th iteration. Let be the j-th state variable in the k-th iteration, and tanh be the hyperbolic tangent function. is the scaling factor for the k-th iteration.

7. A real-time tracking data-guided learning control system for unmanned autonomous vehicles according to claim 6, characterized in that, The adjustment module includes: The instruction filtering unit is used to perform instruction filtering on the desired target trajectory based on the attraction rule, including: obtaining the target motion state from the desired target trajectory, filtering the target motion state, and obtaining the instruction filtering value; The data-guided learning unit is used to update the learning estimation law parameters, including: obtaining the actual motion state from the tracking trajectory, and updating the time-varying system function estimator and the estimator of the estimation parameters based on the actual motion state; The control input calculation unit is used to calculate and update the control input of the unmanned autonomous vehicle based on the instruction filter value and motion control parameters of the unmanned autonomous vehicle.

Images