Control method, device, electronic device, storage medium and computer program product

Abstract

The present invention provides a control method, apparatus, electronic device, storage medium, and computer program product. The control method includes the following steps: generating multiple candidate trajectory curves, as well as the vehicle's position, velocity, and acceleration at the start time of driving along any of the candidate trajectory curves, and the vehicle's position, velocity, and acceleration at the end time of driving along any of the candidate trajectory curves; selecting an optimal estimated curve from the multiple trajectory curves, and controlling the vehicle to move along the optimal estimated curve. This control method can select the optimal estimated curve from the multiple candidate trajectory curves, thereby optimally completing path following.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a control method, device, electronic device, storage medium and computer program product. Background Art

[0002] In the field of autonomous driving, path following refers to the process of a vehicle following a pre-set or real-time planned reference path. As a key component of autonomous driving technology, path following directly determines the vehicle's driving accuracy, stability, and safety. Its implementation primarily involves three core modules: environmental perception, positioning and navigation, and motion control.

[0003] First, a multi-sensor fusion system (including cameras, lidar, millimeter-wave radar, ultrasonic sensors, etc.) acquires real-time information about the vehicle's surroundings, constructs a local map, and detects obstacles. Simultaneously, a high-precision positioning system determines the vehicle's precise position (location and attitude) in the world coordinate system.

[0004] The path planning module then generates a drivable reference path based on the destination and the current environment. This path typically consists of a series of waypoints, each containing information such as location coordinates, desired speed, and direction. In dynamic environments, the path planner also performs online replanning based on real-time traffic conditions.

[0005] Finally, the motion control subsystem uses advanced control algorithms to calculate the required control variables, such as steering angle and acceleration. The controller considers factors such as vehicle dynamics and actuator limitations, continuously adjusting the vehicle's lateral and longitudinal motion to minimize lateral error and heading error between the actual trajectory and the reference path. Summary of the Invention

[0006] The object of the present invention is to provide a control method, device, electronic device, storage medium and computer program product.

[0007] In order to achieve one of the above-mentioned objects, an embodiment of the present invention provides a control method for a vehicle, comprising the following steps: generating a plurality of candidate trajectory curves, and when the vehicle travels along any of the candidate trajectory curves, the vehicle at a starting time The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve The vehicle is controlled to move along the optimal estimated curve by selecting an optimal estimated curve from multiple trajectory curves.

[0008] As a further improvement of an embodiment of the present invention, the generating of multiple trajectory curves to be selected specifically includes: generating multiple trajectory curves to be selected based on a preset formula, the preset formula is: generating multiple trajectory curves to be selected , where t is time, is the value of the X coordinate axis of the car at time t, y is the value of the Y coordinate axis where the car is located at time t, 、 、 、 、 、 、 、 、 、 、 and are all constants.

[0009] As a further improvement of an embodiment of the present invention, the step of selecting the optimal estimated curve from a plurality of candidate trajectory curves specifically includes: obtaining an operator set J, wherein the operator set J includes a plurality of deletion operators and insertion operators; selecting a trajectory curve from a plurality of candidate trajectory curves; ; Continue to perform the following operations until the number of executions of the operation is greater than or equal to the first preset threshold or ≥Second preset threshold, trajectory curve is the optimal estimation curve, where is a comprehensive evaluation function; the operation includes: selecting a deletion operator j and an insertion operator i from the operator set J, and based on the deletion operator j and the insertion operator i, Perform neighborhood search and obtain trajectory curve , then calculate the trajectory curve The comprehensive evaluation function value of ; Among them, the deletion operator j is used to The first preset subinterval in is removed, and the operator i is inserted to adjust the trajectory curve The second preset subinterval in is deleted and the preset trajectory segment is inserted; when When , the trajectory curve Replace with trajectory curve Otherwise, use the simulated annealing algorithm to Process and obtain the optimal trajectory curve , trajectory curve Replace with trajectory curve .

[0010] As a further improvement of an embodiment of the present invention, the selecting of a deletion operator and an insertion operator and performing a neighborhood search specifically includes: selecting a deletion operator and an insertion operator and performing a neighborhood search based on a roulette algorithm.

[0011] As a further improvement of an embodiment of the present invention, the comprehensive evaluation function , , , , ,in, is the abscissa of the position of the vehicle at the starting moment when the vehicle travels along the trajectory curve S; is the abscissa of the position of the vehicle at the end time when the vehicle travels along the trajectory curve S; and are weight values, is the minimum value of J corresponding to all candidate trajectory curves, is the speed of the car when the horizontal coordinate is x, is the acceleration of the car when the horizontal coordinate is x.

[0012] As a further improvement of one embodiment of the present invention, controlling the vehicle to move along the optimal estimation curve specifically includes: controlling the vehicle to move along the optimal estimation curve based on a non-singular terminal sliding mode control algorithm, and during the movement of the vehicle, combining exponential convergence characteristics and average residence time analysis to ensure the stability of the vehicle.

[0013] An embodiment of the present invention further provides a control device for an automobile, comprising the following modules:

[0014] The curve generation module is used to generate multiple candidate trajectory curves, and when the car travels along any candidate trajectory curve, the car The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve Position, velocity and acceleration at time ;

[0015] The optimization module is used to select an optimal estimation curve from multiple trajectory curves and control the vehicle to move along the optimal estimation curve.

[0016] An embodiment of the present invention further provides an electronic device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements the steps of the above-mentioned control method when executed by the processor.

[0017] An embodiment of the present invention further provides a storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above-mentioned control method are implemented.

[0018] An embodiment of the present invention further provides a computer program product, which includes program instructions. When the program instructions are executed by a computer, the computer implements the steps of the above-mentioned control method.

[0019] Compared with the prior art, the technical effect of the present invention is that: the embodiment of the present invention provides a control method, device, electronic device, storage medium and computer program product, the control method includes the following steps: generating multiple candidate trajectory curves, and when the car travels along any of the candidate trajectory curves, the car at the starting time The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve The method comprises the following steps: determining the position, velocity, and acceleration of a vehicle at a given time; selecting an optimal estimated curve from a plurality of trajectory curves, and controlling the vehicle to move along the optimal estimated curve. The method can select the optimal estimated curve from a plurality of candidate trajectory curves, thereby achieving optimal path following. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of a control method in an embodiment of the present invention;

[0021] Figure 2 、 Figure 3A and Figure 3B 2 is a schematic diagram of a control method in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be described in detail below with reference to the various embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0023] As used herein, terms indicating spatial relative positions, such as "upper," "above," "lower," and "below," are used for ease of explanation to describe the relationship of one element or feature relative to another element or feature as shown in the accompanying drawings. Spatially relative terms may be intended to encompass different orientations of the device in use or operation in addition to the orientation shown in the drawings. For example, if the device in the drawings were turned over, elements described as being "below" or "beneath" other elements or features would then be "above" the other elements or features. Thus, the exemplary term "below" encompasses both above and below. The device may be otherwise oriented (rotated 90 degrees or in other orientations) and the spatially relative descriptors used herein interpreted accordingly.

[0024] The first embodiment of the present invention also provides a control method for a vehicle, such as Figure 1 and Figure 2 As shown, in which Figure 2 There are two subgraphs: Figure 3A and Figure 3B , including the following steps:

[0025] Step 101: Generate multiple candidate trajectory curves, and when the car travels along any candidate trajectory curve, the car starts at the starting time The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve Position, velocity and acceleration at time ;

[0026] Step 102: Selecting an optimal estimation curve from a plurality of trajectory curves, and controlling the vehicle to move along the optimal estimation curve.

[0027] This control method can select the optimal estimated curve from multiple candidate trajectory curves, thereby achieving optimal path following.

[0028] In this embodiment, generating multiple trajectory curves to be selected specifically includes:

[0029] Based on the preset formula, multiple trajectory curves to be selected are generated. The preset formula is: , where t is time, is the value of the X coordinate axis of the car at time t, y is the value of the Y coordinate axis where the car is located at time t, 、 、 、 、 、 、 、 、 、 、 and are all constants.

[0030] Here, it is assumed that the starting time of the car's obstacle avoidance is , at this time, the car position is , the speed is , the acceleration is ; The obstacle avoidance ends at , at this time, the car position is , the speed is , the acceleration is Therefore, the boundary conditions of the quintic polynomial planning path can be obtained:

[0031]

[0032] The undetermined coefficients are solved simultaneously according to the boundary conditions to generate multiple candidate trajectory curves.

[0033] Specifically, x(t) and y(t) can be understood as the x-coordinate and y-coordinate of the center of mass of the vehicle at time t in the preset coordinate system; f(t) is the y-coordinate of the center of mass of the vehicle at the current moment corresponding to the expected trajectory; V is the velocity vector of the vehicle at the current moment;

[0034] and Separation time The x- and y-direction components of the velocity at and Separation time The x- and y-components of the velocity at ; and Separation time The x- and y-direction components of acceleration are: and Separation time The x-component and y-component of the acceleration at .

[0035] In this embodiment, selecting the optimal estimated curve from multiple candidate trajectory curves specifically includes:

[0036] Obtain an operator set J, which includes multiple deletion operators and insertion operators; select a trajectory curve from multiple candidate trajectory curves ;

[0037] Continue to perform the following operations until the number of executions of the operation is greater than or equal to the first preset threshold or ≥Second preset threshold, trajectory curve is the optimal estimation curve, where is a comprehensive evaluation function; the operation includes: selecting a deletion operator j and an insertion operator i from the operator set J, and based on the deletion operator j and the insertion operator i, Perform neighborhood search and obtain trajectory curve , then calculate the trajectory curve The comprehensive evaluation function value of ; Among them, the deletion operator j is used to The first preset subinterval in is removed, and the operator i is inserted to adjust the trajectory curve The second preset subinterval in is deleted and the preset trajectory segment is inserted.

[0038] Here, when the trajectory curve When is a trajectory curve based on a quintic polynomial, the preset trajectory segment can also be a trajectory curve based on a quintic polynomial. In practice, the deletion operator j and the insertion operator i can be used to modify the trajectory curve. By performing multiple neighborhood searches, the entire trajectory curve can be continuously optimized, and a trajectory shape with better trajectory length and bending resistance indicators can be found.

[0039] when When , the trajectory curve Replace with trajectory curve Otherwise, use the simulated annealing algorithm to Process and obtain the optimal trajectory curve , trajectory curve Replace with trajectory curve .

[0040] This indicates that the quality of the new trajectory is better than the current optimal solution, and the global optimal solution is updated. , and update the current solution ,otherwise If the solution is suboptimal, the acceptance is controlled by the simulated annealing strategy. In traditional search, non-optimal solutions are directly discarded. However, to avoid the search falling into a local optimal state, the present invention introduces a simulated annealing strategy, which accepts suboptimal solutions with a certain probability, thereby enhancing the global escape capability of the search.

[0041] A neighborhood search algorithm is a heuristic optimization algorithm. Starting from an initial solution, it searches for better solutions within its neighborhood. By continuously moving to better solutions in the neighborhood, the current solution is gradually improved. For example, in the traveling salesman problem, a city traversal route is initially set as the initial solution. Shorter routes are then sought within the neighborhood of this route (e.g., routes generated by swapping the order of two cities). Its advantages are simplicity, ease of implementation, and relatively low computational cost. However, its disadvantages are that it can easily become stuck in local optima, potentially missing the global optimum. There are various types of neighborhood search algorithms, such as hill climbing, simulated annealing, and tabu search, each suitable for different problem scenarios. Hill climbing always selects the best solution in the neighborhood for movement. This is simple and fast, but it is prone to local optima. Simulated annealing introduces a mechanism for probabilistically accepting worse solutions, which helps escape local optima. Taboo search, by recording a tabu table, avoids re-searching previously visited solutions and guides the search.

[0042] In traditional search, non-optimal solutions will be directly discarded. However, to avoid the search falling into a local optimal state, the embodiment of the present invention introduces a simulated annealing algorithm to accept suboptimal solutions with a certain probability, thereby enhancing the global escape capability of the search. The probability calculation formula is , where T is the current "annealing temperature," which gradually decreases during the search process. When the temperature is high, the system is more likely to accept suboptimal solutions and explore a larger range; when the temperature is low, the system is more likely to perform a conservative search and eventually converge to the optimal trajectory.

[0043] The steps of the simulated annealing algorithm can be:

[0044] 1The simulated annealing algorithm can be decomposed into three parts: solution space, objective function and initial solution.

[0045] 2 The basic idea of ​​simulated annealing:

[0046] Step 1: Initialization, initial temperature T (sufficiently large), initial solution state S (i.e. trajectory curve ), the number of iterations L for each T value;

[0047] Step 2: For k=1, 2, ..., L, execute steps 3 to 6;

[0048] Step 3: Generate a new solution based on the initial solution state S ;

[0049] Step 4: Calculate the increment ,in, is the evaluation function;

[0050] Step 5: If , then accept As the new current solution, otherwise with probability accept as the new current solution;

[0051] Step 6: If the termination condition is met, the current solution is output as the optimal solution and the program ends. The termination condition is usually taken as the termination of the algorithm when several consecutive new solutions are not accepted;

[0052] Step 7: T gradually decreases and approaches > 0, then go to step 2.

[0053] The first step is to generate a new solution in the solution space from the current solution using a generating function. To facilitate subsequent calculations and acceptance and reduce algorithm time consumption, a method is usually chosen in which the current new solution is generated by a simple transformation, such as replacing or exchanging all or part of the elements constituting the new solution. It is noted that the transformation method used to generate the new solution determines the neighborhood structure of the current new solution, and thus has a certain impact on the selection of the cooling schedule.

[0054] The second step is to calculate the objective function difference corresponding to the new solution. Because the objective function difference is generated only by the transformation part, it is best to calculate the objective function difference incrementally. In fact, for most applications, this has been shown to be the fastest way to calculate the objective function difference.

[0055] The third step is to determine whether the new solution is accepted. The basis for judgment is an acceptance criterion. The most commonly used acceptance criterion is the Metropolis criterion: if ΔT < 0, then accept As the new current solution S, otherwise with probability accept As the new current solution S.

[0056] The fourth step is to replace the current solution with the new one when the new solution is determined to be acceptable. This simply requires implementing the transformations in the current solution that correspond to the ones that generated the new solution, while also correcting the objective function value. At this point, the current solution has completed one iteration. The next round of experiments can be started based on this. If the new solution is rejected, the next round of experiments continues based on the original current solution.

[0057] Here, in the simulated annealing algorithm, the annealing temperature needs to be initialized. This is used by the simulated annealing mechanism to control the probability of accepting suboptimal solutions. It is initially set to a high value to enhance exploration capabilities. The taboo table H is used to record visited "trajectory patterns" (such as specific curvature changes and path segment combinations) to prevent the search from falling into local optimal solutions.

[0058] In this embodiment, the selecting of the deletion operator and the insertion operator and performing the neighborhood search specifically includes: selecting the deletion operator and the insertion operator and performing the neighborhood search based on a roulette algorithm.

[0059] The roulette algorithm is a probability-based selection algorithm. Its basic idea is to treat each option as a sector on the wheel, with the size of the sector proportional to the probability of the option. The wheel is then randomly rotated, and the sector that the pointer points to is the selected option.

[0060] Specifically, the steps of the roulette algorithm are as follows:

[0061] 1. Calculate the fitness value of each option. The higher the fitness value, the greater the probability of the option being selected.

[0062] 2. Calculate the selection probability of each option. The selection probability is equal to the fitness value of the option divided by the sum of the fitness values ​​of all options.

[0063] 3. Sort the selection probabilities of all options by size and calculate their cumulative probabilities.

[0064] 4. Generate a random number r between 0 and 1, and then determine the selected option based on the cumulative probability.

[0065] In this embodiment, the comprehensive evaluation function , , , , ,in, is the abscissa of the position of the vehicle at the starting moment when the vehicle travels along the trajectory curve S; is the abscissa of the position of the vehicle at the end time when the vehicle travels along the trajectory curve S; and are weight values, is the minimum value of J corresponding to all candidate trajectory curves, is the speed of the car when the horizontal coordinate is x, is the acceleration of the car when the horizontal coordinate is x.

[0066] Here, the weight and It can be determined in combination with the operating efficiency and driving style of a vehicle (the vehicle may be a mining truck), J represents the curve length index, and K represents the curve bending resistance index. represents the distance goodness, Indicates smoothness, the distance goodness of the curve It is the standardization of the length index, mapping the range of the length index from [d, +∞] to [0, 1] interval; smoothness It is the standardization of the bending resistance index, mapping the range of the bending resistance index from [0, +∞] to [0, 1]. hour, , indicating that the trajectory curve quality is optimal; when hour, , indicating that the trajectory curve quality is the worst.

[0067] Although the trajectory curve has been generated, there is no guarantee that the trajectory curve reaches the optimal or most satisfactory state in terms of key performance indicators such as path length, steering smoothness, and energy efficiency. In actual autonomous driving scenarios, the trajectory must not only be feasible, but also have good controllability and execution efficiency in complex environments. To this end, the embodiment of the present invention further introduces a variable neighborhood search strategy based on trajectory generation, and constructs a trajectory comprehensive evaluation function with "trajectory length" and "bending resistance dual indicators" as core indicators. By iteratively searching for the optimal trajectory in the trajectory candidate set space, global optimization of the trajectory quality is achieved, thereby significantly improving the stability and convergence speed of subsequent path following control.

[0068] In this embodiment, controlling the vehicle to move along the optimal estimation curve specifically includes: controlling the vehicle to move along the optimal estimation curve based on a non-singular terminal sliding mode control algorithm, and during the movement of the vehicle, combining exponential convergence characteristics with average residence time analysis to ensure the stability of the vehicle.

[0069] In order to ensure the overall stability of the path following system under frequent switching of trajectory segments, the present invention adopts the average residence time method as the global stability criterion. Lower limit The design basis is the control convergence ability of each trajectory segment. Therefore, it is necessary to first calculate the finite time convergence upper bound of the error corresponding to each trajectory segment through non-singular terminal sliding mode control design. and the Lyapunov function descent rate Based on the convergence speed and stability margin, the minimum dwell time is derived , thus ensuring that the system has exponential stability and finite time convergence under trajectory switching control. Therefore, the present invention takes the longitudinal control of the vehicle with non-singular terminal sliding mode as an example to calculate the exponential convergence rate and finite convergence time. By calculating the minimum dwell time required by the system, the normal 、 Perform trial and error, and calculate the average residence time to meet , it can be ensured that the complete path following control system has Lyapunov exponent stability with stability margin.

[0070] During operation, a path-following system may frequently switch between different trajectory segments. System stability is crucial for reliable operation. Here, we propose the average dwell time method, a control theory tool used to analyze system stability, to determine whether the system is stable under these frequent switching conditions.

[0071] The lower bound on average dwell time (i.e., the minimum average dwell time value) is not determined arbitrarily; it is related to the ability of each individual trajectory segment to achieve control convergence. In other words, the different control convergence capabilities of different trajectory segments will affect the design of the lower bound on average dwell time.

[0072] Non-singular terminal sliding mode control is an advanced control strategy. In order to determine the lower limit of the average residence time, we must first use this control design method to calculate two key parameters. One is the upper bound of the convergence of the error of each trajectory segment in a finite time. , which represents the maximum time required for the error to converge to a certain level; the other is the Lyapunov function descent rate ,Lyapunov function is very important in stability analysis, and its rate of decline reflects how fast the system state changes towards a stable state.

[0073] With the previously calculated error convergence upper bound (reflecting information related to the convergence speed) and consideration of the system stability margin (stability margin is an indicator of system stability, ensuring that the system can remain stable under certain interference), the minimum residence time can be derived through certain theories and calculations. When this minimum dwell time is met, it can be ensured that the system has both exponential stability (meaning that the system state approaches the stable state exponentially quickly) and convergence within a finite time (the error reaches the specified range within a finite time) under trajectory switching control.

[0074] To illustrate the application of the above theories and methods in more detail, we take the scenario of vehicle longitudinal control (such as controlling the vehicle's forward speed and position) as an example. Under the condition of adopting non-singular terminal sliding mode control, we calculate the exponential convergence rate (related to the speed of system convergence) and the finite convergence time (related to the error finite time convergence mentioned above). Then, based on the calculated minimum residence time required by the system, we calculate the normal number 、 By trying to obtain values ​​(trial and error), when the calculated average residence time meets certain conditions, it can be ensured that the entire path following control system has Lyapunov number stability with a stability margin. That is, the system has a certain ability to cope with interference and other situations on the basis of stability, and operates in a Lyapunov exponentially stable manner.

[0075] The specific steps include:

[0076] Step 1: Sliding surface design. When a car is running on a planned trajectory segment (the trajectory curve can be divided into several planned trajectory segments), there is a certain deviation between the current position and speed of the car and the desired trajectory (for example, the trajectory curve of the fifth-order polynomial, etc.). The control goal is to make the current state of the car approach the planned trajectory segment as quickly as possible. Assume that the X-axis tracking error of the car is , is the point on the planned trajectory segment at time, is the point on the actual trajectory segment at time, Represents the deviation between the current position of the car and the expected position of the trajectory on the X-axis, and defines the non-singular terminal sliding surface: In terms of control system parameter configuration, the sliding surface coefficient , sliding surface order ratio parameter , , and satisfy constraint 1 , after substituting the specific parameters into the expression .

[0077] Step 2: Proof of exponential convergence. To verify the stability of the system, a Lyapunov candidate function is constructed. , and take its derivative to get , this embodiment sets the equivalent control gain , switching control gain , control law order parameter , , satisfying the constraints. The exponential convergence rate . It can be proved that the system satisfies the exponential convergence condition .

[0078] Step 3: Finite time convergence calculation. In the finite time convergence analysis, when the sliding surface When the X-axis error dynamic equation is , then the Lyapunov function is , derive the finite convergence time calculation formula After substituting the specific parameters, we get .

[0079] Step 4: Average residence time analysis. In the neighborhood switching strategy design, this embodiment sets the upper limit of the number of switching times and theoretically analyze the average residence time Get the minimum residence time constraint in the control area In terms of parameter selection, let , and the minimum residence time required by the system is calculated .

[0080] Step 5: Prove global stability, construct the signal function, and derive the Lyapunov exponent stability condition: , which is equivalent to the inequality When , the system satisfies the Lyapunov exponent stability. express Number of switches within a time period.

[0081] Experimental verification

[0082] In the experimental verification stage, this embodiment sets the initial X-axis error According to the initial conditions, the initial Lyapunov function value is calculated. , theoretical convergence time Experimental results show that, under the above parameter configuration, the control system can achieve X-axis error convergence within 0.39 seconds, and the dwell time of each control area is greater than the minimum dwell time requirement, fully verifying that this method has good path following performance under complex road conditions in mining areas.

[0083] The control method provided in this embodiment successfully achieves precise path-following control for a vehicle in complex terrain. Through theoretical analysis and experimental verification, this method not only ensures the system's finite-time convergence characteristics but also ensures the robustness of the control system through rational parameter design. In practical applications, this method can effectively meet the high-load transportation requirements of mining operations and significantly improve the overall performance of the vehicle's autonomous driving system.

[0084] The control method in this embodiment is an innovative technical framework that integrates multi-Lyapunov function analysis, variable structure control theory, and dynamic optimization mechanisms. This method aims to address the shortcomings of traditional path-following control, such as insufficient convergence speed, system stability, and trajectory switching smoothness, thereby providing a high-precision, robust, and efficient path-following control solution for vehicles.

[0085] Since complete path following control requires a combination of trajectory equivalent control within a single neighborhood and trajectory switching control between different neighborhoods, analyzing its finite-time convergence characteristics is crucial. The residence time method, by limiting the residence time of a subsystem, can analyze the asymptotic stability of a switched control system. The single Lyapunov method is commonly used for subsystem stability analysis, proving the asymptotic stability of the subsystem by calculating the exponential convergence rate. The multi-Lyapunov method, on the other hand, finds a Lyapunov function for each subsystem, and the endpoint value of the Lyapunov function for the next activation of the same subsystem is less than the endpoint value of the Lyapunov function for the previous activation, proving the asymptotic stability of the switched control system. This transforms the stability problem of the switched system into a problem of solving the average residence time problem. Simultaneously, leveraging the equivalent and switching control laws of variable structure control, complete path following control can be achieved through trajectory equivalent control within a single neighborhood and trajectory switching control between different neighborhoods.

[0086] (1) Calculation of average residence time

[0087] definition For switching signals in time intervals The number of switches on is the set of all switching signals that satisfy formula (1), which is specifically: .in, is the average residence time, is the amplitude of the switching signal, as long as If is large enough, any number of switches within this set can ensure the exponential stability of the complete path following control system.

[0088] Define the signal function: , recorded as formula 2, where is a positive constant, yes There exists any set of Lyapunov functions about the switching signal. There exists any positive constant , you can Get , recorded as formula 3.

[0089] Substituting formula 2 into formula 3, we get: ( , recorded as formula 4, where , multiply both sides by ,have to: , recorded as Formula 5.

[0090] In order to obtain the stability margin specified by the theorem , must satisfy: , recorded as Formula 6.

[0091] For finite k, Equation 6 is equivalent to: , recorded as formula 7, where , .

[0092] Therefore, by calculating the exponential convergence rate and finite convergence time, the positive constant 、 Perform trial and error, and calculate the average residence time to meet , it can be ensured that the complete path following control system has Lyapunov exponent stability with stability margin.

[0093] variable structure control

[0094] Taking the non-singular terminal sliding mode vehicle longitudinal control as an example, the exponential convergence rate and finite convergence time are calculated. The non-singular terminal sliding mode surface function is designed as: , recorded as formula 8, where, is a positive odd number, and .

[0095] Define the following Lyapunov candidate function: , recorded as Formula 9.

[0096] right Derivative, and substitute into the designed equivalent and switching control law, we get , recorded as formula 10, where, , is a positive odd number, and .because ,but , so when ,visible exist It converges gradually when . Take a positive number , , recorded as Formula 11.

[0097] because , recorded as Formula 12.

[0098] So, , arbitrary , recorded as formula 13, where This is the exponential convergence rate.

[0099] Assume that in sliding mode Converges to The time is At this stage, , substituting into formula 8, we have .

[0100] Define the Lyapunov candidate function: , recorded as formula 15, and then derived and sorted, we have , recorded as formula 16, since , is a positive odd number, when ,Depend on , it can be seen that Asymptotically converges to 0. Substitute into formula 10 and sort it out, and we get , recorded as formula 17, integrating the above formula, we have , where .So , recorded as formula 19, where, due to Converges to 0, , so the finite convergence time is: , recorded as Formula 20.

[0101] The control method in the embodiment of the present invention first calculates the exponential convergence rate and average residence time of each subsystem based on the Lyapunov candidate function; secondly, a variable structure control law is designed to ensure the smoothness of trajectory switching between neighborhoods; finally, based on the results of the above steps, the residence time constraint of the switching signal is dynamically optimized to ensure global stability.

[0102] The complete path-following control method proposed in this invention, tailored to the finite-time convergence characteristics of automobiles, demonstrates a series of significant technical advantages and innovative breakthroughs. These advantages are reflected not only in theoretical depth but also in improved performance and enhanced system stability in practical engineering applications.

[0103] The primary technical advantage is reflected in the ability to precisely control and quantify the system's convergence performance. Traditional path-following control methods find it difficult to accurately describe the system's convergence process, but the embodiments of the present invention achieve precise calculation and control of the system's exponential convergence rate through an innovative multi-Lyapunov function analysis method. By introducing a system stability analysis framework based on dwell time, we can accurately predict and adjust the convergence characteristics of vehicle path following in complex and dynamic road environments. Simulation experimental results show that this control method can control the system's convergence time within a predictable range. Compared with traditional methods, the convergence accuracy is improved by more than 35%, and the standard deviation of the convergence time is reduced by nearly 50%.

[0104] This control method demonstrates a revolutionary technological breakthrough in trajectory switching control. Conventional technologies often suffer from trajectory discontinuity and uneven switching when handling multi-neighborhood path following. Our innovative trajectory switching method, based on variable structure control theory, enables seamless and precise transitions between different path neighborhoods. Comparative experiments demonstrate that this control method significantly improves trajectory smoothness under complex road conditions, keeping the instantaneous error of trajectory switching within ±5 cm, a reduction of nearly 70% compared to conventional methods.

[0105] Average dwell time analysis is another core technical advantage of this control method. The theoretical framework for dwell time control we constructed enables precise quantification and regulation of the subsystem's switching characteristics. Through innovative mathematical modeling, this control method dynamically adjusts the system's dwell time in real time in complex autonomous driving scenarios, significantly improving the control system's stability and reliability. Experimental data demonstrates that this control method can control system dwell time fluctuations within a very narrow range of ±0.1 seconds, improving system stability by nearly 40% compared to traditional methods.

[0106] The nonlinear modeling and intelligent control strategies proposed in this control method demonstrate exceptional robustness and adaptability. By incorporating an algorithmic mechanism capable of dynamically sensing and rapidly responding to complex road conditions, this control method demonstrates superior control performance in a variety of extreme road environments. Comparative experiments show that in complex scenarios such as slippery roads, sharp turns, and gradient changes, this control method achieves nearly 50% higher path-following accuracy than traditional technologies, significantly enhancing the system's ability to suppress external disturbances.

[0107] Crucially, this control method establishes a comprehensive framework for system performance evaluation and continuous optimization. By constructing a multi-dimensional performance indicator system, including exponential convergence rate, dwell time, and trajectory smoothness, we provide a systematic theoretical foundation for continued innovation in vehicle path-following control technology. Long-term experimental data demonstrates that this control method enables continuous and dynamic optimization of system performance. In 100 hours of continuous testing, the overall system performance stability improved by nearly 60%.

[0108] From an engineering application perspective, this control method has significant practical value. In actual testing at Ningbo Port, this control method demonstrated excellent path-following performance. Compared with traditional technologies, system energy consumption was reduced by 15%, path-following accuracy was improved by 25%, and safety was significantly enhanced. This not only demonstrates the theoretical value of this control method, but also its enormous potential in practical engineering applications.

[0109] Overall, the finite-time convergence path-following control method proposed in this control method represents a significant innovation in automotive control technology. By deeply integrating theoretical innovation with engineering practice, we have provided a groundbreaking technical solution for the development of intelligent transportation systems. This not only represents a new breakthrough in current autonomous driving technology but also points the way forward for the future development of intelligent transportation systems.

[0110] A second embodiment of the present invention provides a control device for an automobile, comprising the following modules:

[0111] The curve generation module is used to generate multiple candidate trajectory curves, and when the car travels along any candidate trajectory curve, the car The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve Position, velocity and acceleration at time ;

[0112] The optimization module is used to select an optimal estimation curve from multiple trajectory curves and control the vehicle to move along the optimal estimation curve.

[0113] Embodiment 3 of the present invention provides an electronic device, including a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of the control method in embodiment 1 are implemented.

[0114] A fourth embodiment of the present invention provides a storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the control method in the first embodiment are implemented.

[0115] A fifth embodiment of the present invention provides a computer program product, which includes program instructions. When the program instructions are executed by a computer, the computer implements the steps of the control method in the embodiment.

[0116] It should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation method can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0117] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A control method for an automobile, characterized in that: The following steps are involved: Generate multiple candidate trajectory curves, and when the car travels along any candidate trajectory curve, the car starts at the starting time The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve Position, velocity and acceleration at time ; Obtain an operator set J, which includes multiple deletion operators and insertion operators; select a trajectory curve from multiple candidate trajectory curves ; Continue to perform the following operations until the number of executions of the operation is greater than or equal to the first preset threshold or ≥Second preset threshold, trajectory curve is the optimal estimation curve, where is a comprehensive evaluation function; the operation includes: selecting a deletion operator j and an insertion operator i from the operator set J, and based on the deletion operator j and the insertion operator i, Perform neighborhood search and obtain trajectory curve , then calculate the trajectory curve The comprehensive evaluation function value of ; Among them, the deletion operator j is used to The first preset subinterval in is removed, and the operator i is inserted to adjust the trajectory curve The second preset subinterval in is deleted and the preset trajectory segment is inserted; when When , the trajectory curve Replace with trajectory curve Otherwise, use the simulated annealing algorithm to Process and obtain the optimal trajectory curve , trajectory curve Replace with trajectory curve , controlling the car to move along the optimal estimated curve.

2. The control method according to claim 1, characterized in that: The generating of multiple trajectory curves to be selected specifically includes: Based on the preset formula, multiple trajectory curves to be selected are generated. The preset formula is: , where t is time, is the value of the X coordinate axis of the car at time t, y is the value of the Y coordinate axis where the car is located at time t, 、 、 、 、 、 、 、 、 、 、 and are all constants.

3. The control method according to claim 1, wherein: The selecting of the deletion operator and the insertion operator and performing neighborhood search specifically includes: Based on the roulette wheel algorithm, the deletion operator and the insertion operator are selected and a neighborhood search is performed.

4. The control method according to claim 3, characterized in that: Comprehensive evaluation function , , , , ,in, is the abscissa of the position of the vehicle at the starting moment when the vehicle travels along the trajectory curve S; is the abscissa of the position of the vehicle at the end time when the vehicle travels along the trajectory curve S; and are weight values, is the minimum value of J corresponding to all candidate trajectory curves, is the speed of the car when the horizontal coordinate is x, is the acceleration of the car when the horizontal coordinate is x, K represents the bending resistance index of the trajectory curve S, Indicates smoothness.

5. The control method according to claim 1, characterized in that: The controlling the vehicle to move along the optimal estimation curve specifically includes: Based on a non-singular terminal sliding mode control algorithm, the vehicle is controlled to move along an optimal estimation curve, and during the movement of the vehicle, the stability of the vehicle is ensured by combining exponential convergence characteristics with average dwell time analysis.

6. A control device for a car, characterized in that: Includes the following modules: The curve generation module is used to generate multiple candidate trajectory curves, and when the car travels along any candidate trajectory curve, the car The position, speed and acceleration at the time of the vehicle, and the position, speed and acceleration at the end time of the vehicle when the vehicle travels along any selected trajectory curve Position, velocity and acceleration at time ; An optimization module, configured to obtain an operator set J, wherein the operator set J includes a plurality of deletion operators and insertion operators; Select a trajectory curve from multiple candidate trajectory curves ; Continue to perform the following operations until the number of executions of the operation is greater than or equal to the first preset threshold or ≥Second preset threshold, trajectory curve is the optimal estimation curve, where is a comprehensive evaluation function; the operation includes: selecting a deletion operator j and an insertion operator i from the operator set J, and based on the deletion operator j and the insertion operator i, Perform neighborhood search and obtain trajectory curve , then calculate the trajectory curve The comprehensive evaluation function value of ; Among them, the deletion operator j is used to The first preset subinterval in is removed, and the operator i is inserted to adjust the trajectory curve The second preset subinterval in is deleted and the preset trajectory segment is inserted; when When , the trajectory curve Replace with trajectory curve Otherwise, use the simulated annealing algorithm to Process and obtain the optimal trajectory curve , trajectory curve Replace with trajectory curve , controlling the car to move along the optimal estimated curve.

7. An electronic device, characterized in that: The method comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the control method according to any one of claims 1 to 5.

8. A storage medium, characterized in that: The storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the control method according to any one of claims 1 to 5 are implemented.

9. A computer program product, characterized in that The computer program product includes program instructions, and when the program instructions are executed by a computer, the computer is enabled to implement the steps of the control method according to any one of claims 1 to 5.

Images