Parking control method and system, automatic driving equipment and readable storage medium
By constructing a parking planning model based on the distance and speed parameters of the planned path in automatic parking, the problem of large sensor calibration data volume is solved, and efficient and comfortable parking control is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-15
- Publication Date
- 2026-03-10
AI Technical Summary
In existing technologies, the large amount of parameter calibration data obtained by sensors during automatic parking leads to low parking efficiency.
By determining distance and speed parameters based on the planned parking path, and inputting them into a pre-built parking planning model, parking control parameters are obtained to control the vehicle to travel along the parking path.
The number of sensor types and the amount of calibration data have been reduced, shortening the processing time of the vehicle's infotainment system and improving parking efficiency and comfort.
Smart Images
Figure CN121626104A_ABST
Abstract
Description
[0001] This application claims priority to Chinese patent application filed on August 30, 2024, application number 202411216902.9, entitled "Parking control method, system, autonomous driving device and readable storage medium", the entire contents of which are incorporated herein by reference. Technical Field
[0002] Embodiments of this application relate to the field of vehicle control technology, and in particular to a parking control method, system, autonomous driving device, and readable storage medium. Background Technology
[0003] Automated parking is a common automatic control technology in current autonomous driving systems. During automated parking, the vehicle speed is mostly below 3 kph. Due to the low speed, the smoothness of the speed and subtle changes in vehicle acceleration are easily perceived by the driver. Currently, various types of sensors are typically used to collect information about the vehicle's surrounding environment, and then a controller manages the vehicle's movement in different environments. To effectively reproduce the information in the scene, the sensors need to be specifically calibrated for various parameters in the environment. This results in a large amount of calibration data, leading to excessive processing time for the vehicle's infotainment system and low parking efficiency. Summary of the Invention
[0004] The purpose of this application is to provide a parking control method, system, autonomous driving device, and readable storage medium to solve the technical problems of large amount of sensor parameter calibration data for the environment and low parking efficiency in the prior art.
[0005] To address the aforementioned technical problems, embodiments of this application disclose the following technical solutions:
[0006] Firstly, a parking control method is provided, including:
[0007] Based on the planned parking path, distance parameters and speed parameters are determined. The distance parameters are used to characterize the remaining distance of the parking path during the parking process, and the speed parameters are used to characterize the vehicle speed during the parking process.
[0008] The distance and speed parameters are input into a pre-built parking planning model to obtain parking control parameters;
[0009] The vehicle is controlled to travel along the parking path based on the parking control parameters.
[0010] Secondly, a parking control system is provided, the system comprising:
[0011] The parameter acquisition module is configured to determine distance parameters and speed parameters based on the planned parking path. The distance parameters are used to characterize the remaining distance of the parking path during the parking process, and the speed parameters are used to characterize the vehicle speed during the parking process.
[0012] A parameter input module is configured to input the distance parameter and the speed parameter into a pre-built parking planning model to obtain parking control parameters;
[0013] A vehicle control module configured to control a vehicle to travel along the parking path based on the parking control parameters.
[0014] Thirdly, a parking control system is provided, including:
[0015] Memory, used to store programs;
[0016] A processor for executing a program stored in the memory;
[0017] When the program stored in the memory is executed, the processor performs the parking control method as described in any one of the first aspects.
[0018] Fourthly, an autonomous driving device is provided, the autonomous driving device including a parking control system as described in the second or third aspect.
[0019] Fifthly, a computer-readable storage medium is provided that stores instructions for execution by a computing device, wherein the computing device, when executing the instructions, implements the parking control method as described in any one of the first aspects.
[0020] One of the above technical solutions has the following advantages or beneficial effects:
[0021] This application provides a parking control method, including: determining distance parameters and speed parameters based on a planned parking path, where the distance parameters characterize the remaining distance of the parking path during the parking process, and the speed parameters characterize the vehicle speed during the parking process; inputting the distance parameters and speed parameters into a pre-built parking planning model to obtain parking control parameters; and controlling the vehicle to travel along the parking path based on the parking control parameters. The parking control method provided in this application uses fewer types of sensors, does not require specific calibration of various parameters in the environment, has a small amount of calibration data, short processing time for the vehicle's infotainment system, and results in high parking efficiency and high comfort.
[0022] This application also provides a parking control system, comprising: a parameter acquisition module configured to determine distance and speed parameters based on a planned parking path, wherein the distance parameter represents the remaining distance of the parking path during parking, and the speed parameter represents the vehicle speed during parking; a parameter input module configured to input the distance and speed parameters into a pre-built parking planning model to obtain parking control parameters; and a vehicle control module configured to control the vehicle to travel along the parking path based on the parking control parameters. The parking control system provided in this application uses fewer types of sensors, does not require specific calibration of various parameters in the environment, has a small amount of calibration data, short processing time of the vehicle system, and offers high parking efficiency and comfort. Attached Figure Description
[0023] The technical solution and other beneficial effects of this application will become apparent from the following detailed description of specific embodiments in conjunction with the accompanying drawings.
[0024] Figure 1 This is a schematic flowchart of the parking control method provided in the embodiments of this application;
[0025] Figure 2 A schematic diagram illustrating the construction of the parking planning model provided in this application embodiment;
[0026] Figure 3 A schematic diagram of distance variation curves provided for embodiments of this application;
[0027] Figure 4 A schematic diagram of the velocity change curve provided in the embodiments of this application;
[0028] Figure 5 This is a schematic diagram of acceleration variation curves provided in an embodiment of this application;
[0029] Figure 6 This is a schematic diagram of the acceleration variation curve provided in the embodiments of this application;
[0030] Figure 7 This is a schematic diagram of a parking route planning application scenario provided in an embodiment of this application;
[0031] Figure 8 This application provides schematic diagrams illustrating parking route planning application scenarios for some embodiments.
[0032] Figure 9 A schematic diagram of a parking route planning application scenario provided in yet another embodiment of this application;
[0033] Figure 10 This is a schematic diagram of a parking braking trajectory provided in one embodiment of this application;
[0034] Figure 11 A schematic diagram of a parking braking trajectory provided in another embodiment of this application;
[0035] Figure 12 This is a schematic diagram of a parking braking trajectory provided in yet another embodiment of this application;
[0036] Figure 13 A schematic diagram of the architecture of the primary planning model provided in the embodiments of this application;
[0037] Figure 14 This is a schematic diagram of the data processing architecture for vehicle parking planning provided in an embodiment of this application;
[0038] Figure 15 This is a schematic diagram of a module for preprocessing vehicle data provided in an embodiment of this application;
[0039] Figure 16 This is a schematic diagram of a cyclic slice of vehicle data provided in an embodiment of this application;
[0040] Figure 17 A schematic diagram showing the comparison between the actual driving torque and the predicted driving torque of a vehicle provided in an embodiment of this application;
[0041] Figure 18 A schematic diagram of the acceleration variation curve of a vehicle provided in an embodiment of this application;
[0042] Figure 19 A schematic diagram showing the changes in speed, acceleration, jerk, and driving torque over time during the vehicle braking process provided in this application embodiment;
[0043] Figure 20 A schematic diagram illustrating the application scenario of the parking planning model provided in this application on a vehicle;
[0044] Figure 21 A block diagram of a parking control system provided in an embodiment of this application;
[0045] Figure 22 This is a schematic diagram of the control connection of the parking control system module provided in an embodiment of this application;
[0046] Figure 23 A schematic diagram of the algorithm control logic of the parking control system provided in the embodiments of this application.
[0047] The attached figures are labeled as follows:
[0048] 100 - Vehicle, 200 - Parking route, 210 - Target route, 300 - Target parking location. Detailed Implementation
[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0050] Parking control is a common technology in autonomous driving. During parking, extremely low speeds are typically used to control the vehicle's reversing and adjust its posture. In these ultra-low speed scenarios, vehicles are more prone to loss of control; for example, maintaining speed is difficult in response to changes in road surface and environment, and vehicle acceleration is easily affected by the performance limits of actuators, resulting in large fluctuations. In the entire longitudinal motion control of parking, not only speed accuracy but also distance accuracy must be considered. Compared to speed accuracy, distance accuracy is more critical for parking and plays a vital role in a successful parking maneuver. For a single parking maneuver, the required distance control accuracy is ±10cm.
[0051] Parking control typically employs controller-based solutions and motion planning solutions. In controller-based solutions, the control variable is often a tracking of the target variable, making it difficult to control complex systems with multiple critical states. It also struggles to address constraints and achieve dynamic smoothing of the output, resulting in low comfort during parking control. In contrast, motion planning solutions generally begin by discretizing the system. Then, through system modeling, the physical relationship between the parking motion's speed and distance is abstracted, establishing the system's state-space equations. Finally, the control problem is transformed into an optimization problem to find the optimal control variable that meets specific constraints.
[0052] Those skilled in the art have noted that when using a controller for parking control, although the controller can be implemented with lower resource consumption and a simpler structure at the algorithm level, speed and distance are decoupled, meaning that different controllers are used to complete the control, making it difficult to achieve the control objective simultaneously for both. Furthermore, achieving the same controllable performance in different scenarios requires targeted calibration, resulting in a large number of calibration parameters, a heavy calibration workload, and difficulties in balancing performance.
[0053] The specific implementation methods of this application are illustrated below through examples:
[0054] like Figure 1 As shown in the figure, this application provides a parking control method, including:
[0055] S1: Determine distance and speed parameters based on the planned parking path. The distance parameter is used to characterize the remaining distance of the parking path during the parking process, and the speed parameter is used to characterize the vehicle speed during the parking process.
[0056] Specifically, in this embodiment, the parking path planning method includes: determining the vehicle's current position, the target parking position, and the positions of obstacles. The specific method includes: the vehicle's infotainment system acquiring the vehicle's current position information through the vehicle's sensors or positioning system, such as using a global positioning system, a BeiDou satellite navigation system, or onboard sensors; determining the target parking position based on parking space information captured by a fisheye camera or the parking position input by the driver; and detecting surrounding obstacles, including other vehicles, pedestrians, trees, bollards, and walls, using a fisheye camera, millimeter-wave radar, or infrared sensors to acquire their position information. Then, based on the current position, the target parking position, and the obstacle positions, a parking path is planned: based on the information obtained above, a path planning algorithm is used to determine the optimal parking path. Commonly used path planning algorithms include: shortest path algorithms (e.g., Dijkstra's algorithm, A* algorithm), used to find the shortest path between two points; minimum spanning tree algorithms (e.g., Prim's algorithm, Kruskal's algorithm), used to find the minimum spanning tree in a connected graph, suitable for situations where all nodes need to be covered; bidirectional search algorithms, which search simultaneously from the starting point and the ending point until the two search paths meet, reducing the time and space complexity of the search; genetic algorithms, which simulate the biological evolution process, searching for the optimal solution through operations such as selection, crossover, and mutation; and ant colony algorithms, which simulate the behavior of ants searching for food, searching for the optimal path through the transmission and updating of pheromones.
[0057] In this embodiment, the distance parameters include a first distance between the current vehicle and the target parking position and a second distance between the vehicle and the target parking position during movement; the speed parameters include the initial speed of the current vehicle and the target speed of the vehicle along the parking path. The target speed of the vehicle along the parking path is determined based on the first distance; the target speed is positively correlated with the first distance.
[0058] Specifically, after determining the parking path, the path is segmented to obtain multiple segmented paths. The Euclidean distance algorithm is then used to obtain the value of each segmented path, and these values are summed to obtain the total distance of the parking path. Based on the distance, the target parking position, and the positions of obstacles, the vehicle speed during the parking process is determined. Generally, the greater the distance between the vehicle's current position and the target parking position, and the farther the obstacle is from the parking path, the higher the vehicle's parking speed; conversely, the smaller the distance between the vehicle's current position and the target parking position, or the closer the obstacle is to the parking path, the lower the vehicle's parking speed. It is worth noting that when the movement trajectory of an obstacle intersects with the parking path, the vehicle's parking speed is reduced. During the parking process, the above steps are repeated in real time to dynamically calculate the distance between the vehicle and the target parking position, thereby achieving real-time control of the vehicle's speed.
[0059] Understandably, parking path planning methods can adjust the parking path of a vehicle based on its current location and target parking location, offering a degree of flexibility. By detecting obstacle locations, parking path planning can avoid collisions with obstacles, improving parking safety, accuracy, and efficiency. Furthermore, by determining the vehicle's remaining path and speed during the parking process based on the parking path, real-time vehicle control can be achieved, thereby enhancing the comfort and safety of the parking process.
[0060] S2: Input the distance and speed parameters into the pre-built parking planning model to obtain the parking control parameters.
[0061] Specifically, in this embodiment, the distance parameters include a first distance between the current vehicle and the target parking position and a second distance between the vehicle and the target parking position during the movement; the speed parameters include the initial speed of the current vehicle and the target speed of the vehicle along the parking path; the parking control parameters include the acceleration of the vehicle at each time point; the first distance, the second distance, the initial speed, and the target speed are input into the parking planning model to obtain the acceleration of the vehicle at each time point.
[0062] Specifically, such as Figure 2 As shown in the embodiments of this application, the method for constructing a parking planning model includes:
[0063] S201: Construct a motion planning function based on vehicle motion parameters. The motion planning function is configured to output multiple planning results based on distance and speed parameters.
[0064] First, polynomial functions are constructed based on vehicle motion parameters, including travel distance, travel speed, acceleration, and jerk. According to the parking path, the relationship between speed and distance during the longitudinal parking process is represented by a continuous curve, such as... Figure 3 As shown, the vertical axis represents parking speed, and the horizontal axis represents distance. During parking, the vehicle's speed accelerates from 0, reaches the target speed, and then stops accelerating. Near the target parking position, the speed gradually decreases, undergoes minor adjustments, and finally comes to a stop at 0. The greater the vehicle's speed, the greater the distance it travels. Therefore, a 5th-order polynomial can be used as the distance function to describe the distance *s* traveled. The 4th-order function, after reducing the distance function, is the speed function *v*. Similarly, the 3rd-order function, after reducing the speed function, is the acceleration function *a*, and the 2nd-order function, after reducing the acceleration function, is the jerk function *j*. The specific formulas are as follows:
[0065]
[0066] In the above equation, the unknowns include c3, c4, c5, t, given the initial conditions (s0, v0, a0, j0) and the terminal conditions (s... e ,v e ,a e ,j e The initial conditions are the initial distance s0, initial velocity v0, initial acceleration a0, and initial jerk j0 of the vehicle when entering the parking path, and the final condition is the displacement distance s of the vehicle when it reaches a stable speed. e Target speed v e The final acceleration a e and terminal accelerometer j e By combining the initial and final conditions, the coefficient expressions of the unknown coefficients c3, c4, and c5 with respect to the time parameter t can be obtained:
[0067]
[0068] After simplifying the function, a time plan t is determined based on the parking path. The time plan t represents the time it takes for the vehicle to complete the parking path. Different time plans t are set: First, the time domain is limited to a closed interval [1s, 15s] for finite discrete selection. Each time the time domain is set, eight different time parameters t∈{t|t0,t1,t2,t3,t4,t5,t6,t7} are set, where t0~t7 represent the time required for the vehicle to traverse the parking path. Therefore, substituting t0~t7 into the above formula yields eight different control methods. Simultaneously, the unknown coefficients c3, c4, and c5 can be calculated. By filtering different control methods, an optimal t is found. * As the final governing equation.
[0069] During initialization, the value of the time schedule t will be set to...
[0070] If it's not initialization, meaning the vehicle stops after traveling a certain distance, then replanning is required. Subsequent planning needs to be carried out based on a normal distribution within the remaining time of the previous planning, primarily selecting in the positive direction. Considering the continuity between the current and previous planning, a portion of the negative direction selection also needs to be retained. The specific expression is:
[0071]
[0072] In the above formula, μ is the remaining time of the previous plan, σ is the standard deviation of the time plan, and π is the mathematical constant pi.
[0073] Based on the remaining time from the previous plan and the set time constraints, the rules that the subsequent time planning t0 to t7 must satisfy are shown in Table 1 below:
[0074] Table 1
[0075]
[0076] In the table above, t p It is the remaining time from the last plan, t c It is the calculation time of the vehicle's infotainment system, t min This is the time for minimum planning; generally, t min =1. Based on the time planning rules in Table 1, the subsequent path is planned to obtain 8 different time control methods related to t0 to t7. Then, the optimal t is selected from the 8 time control methods. * The governing equations are derived as part of the time planning process.
[0077] It is worth noting that, in order to prevent the accelerometer j from overflowing when it is derived to 0, the time schedule t is further constrained, resulting in the constrained time t. limitThe specific formulas include:
[0078]
[0079] t max =min(t) limit ,15s);
[0080] In the above formula, t max The optional time cannot exceed 15 seconds.
[0081] S202: Construct constraint functions based on constraints, and select planning results that meet the constraints from multiple planning results through the constraint functions.
[0082] Specifically, from a physical and comfort perspective, the vehicle motion parameters are designed to meet the following constraints: maximum parking speed, maximum acceleration, minimum acceleration, maximum jerk, and minimum jerk.
[0083]
[0084] In the above formula, velocity represents the vehicle's parking speed; vMax is the maximum permissible parking speed; Acceleration represents the vehicle's parking acceleration; aMax is the maximum permissible parking acceleration; aMin is the maximum permissible parking deceleration; Jerk represents the vehicle's parking jerk; jMax is the maximum permissible parking jerk; and jMin is the minimum permissible parking jerk.
[0085] The extreme values (v) of the reciprocals of different physical quantities in the planning results are obtained by differentiating them with respect to time. extrem ,a extrem ,j extrem ):
[0086]
[0087] In the above formula, v extrem This represents the extreme value where the reciprocal of the derivative of velocity with respect to time is zero; a extrem This represents the extreme value of acceleration obtained by differentiating the acceleration with respect to time, where its reciprocal is zero; j extrem This represents the extreme value where the reciprocal of the jerk is zero when the derivative of the jerk with respect to time is taken.
[0088] The overshoot (δ) is obtained by subtracting the extreme values from the constraint conditions. v ,δ a ,δ j ):
[0089]
[0090] In the above formula, δ v Overshoot representing velocity; δ a The overshoot of acceleration; δ j This represents the overshoot of the jerk. Unifying the three different overshoot values yields the combined overshoot:
[0091]
[0092] Finally, the overall overshoot was obtained under eight time planning conditions with different conditions. The planning results were obtained by sorting and selecting the time planning with the highest overshoot.
[0093] like Figure 3 The image shown is a graph depicting the relationship between distance and time after planning. Figure 3 It can be seen that when the time is 0, the distance is not 0, indicating that the vehicle has an initial distance before entering the parking path, and the distance increases accordingly as time increases; for example Figure 4 The figure shown is a graph depicting the relationship between speed and time after planning. Figure 4 It can be seen that when the time is 0, the speed is not 0, indicating that the vehicle has an initial speed before entering the parking path. As time increases, the vehicle's speed also increases accordingly, until it reaches the target speed at approximately 1.8 seconds and then remains stable. Figure 5 The figure shown is a graph depicting the relationship between acceleration and time after planning. Figure 5 It can be seen that when the time is 0, the acceleration is not 0, indicating that the vehicle has an initial acceleration before entering the parking path. As time increases, the vehicle's acceleration decreases until it reaches 0 at around 1.8 seconds, at which point the speed reaches the target speed and remains stable. Figure 6 The figure shown is a graph depicting the relationship between jerk and time after planning. Figure 6 It can be seen that when the time is 0, the jerk is not 0, indicating that the vehicle has jerk before entering the parking path. As time increases, the vehicle's jerk also increases accordingly, until it reaches its maximum value at about 1.4s and then decreases.
[0094] S203: Construct an evaluation function based on the evaluation conditions to filter the selected planning results through the evaluation function in order to obtain parking control parameters.
[0095] Based on time planning, the timeliness of the planning results is determined, which is characterized by the length of vehicle parking time; based on acceleration, the comfort of the planning results is determined, which is characterized by the dynamic characteristics of the vehicle; the planning results that satisfy both comfort and timeliness are selected as parking control parameters.
[0096] Specifically, the remaining planning results after conditional constraints are used to calculate the Cost using trajectory functions with different time-domain parameters:
[0097]
[0098] In the above formula, K is the calibrable gain, with a calibration value ranging from 0 to 1. K close to 0 indicates a slower vehicle speed and lower timeliness; K close to 1 indicates a faster vehicle speed and higher timeliness; τ is the unit time. K·t is used to measure the timeliness of a single planning operation, i.e., the overall parking time. It is used to measure the dynamic characteristics of a single planning operation, reflecting the comfort level during the parking process. The two are mutually inhibiting; an optimal Cost that meets the requirements is defined by adjusting K. * The function serves as the filtering standard. It is worth noting that the specific value of K can be selected by the driver. For example, if the driver wants to prioritize comfort, the selected K value can be 0.1, 0.2, 0.25, etc. If the driver wants to prioritize timeliness, the selected K value can be 0.8, 0.9, 0.95, etc. Based on the selection result, different planning results are selected through the filtering function.
[0099] S3: Controls the vehicle to travel along the parking path based on parking control parameters.
[0100] Specifically, after obtaining the parking control parameters, namely acceleration, the vehicle is controlled to park along the parking path based on the magnitude of the acceleration.
[0101] like Figure 7 The diagram shows a parking path planning application scenario provided in this application embodiment. In the diagram, the vehicle 100 obtains the current vehicle position and the target parking position 300 through the positioning sensor and then plans a parking path 200. Combined with the vision sensor, it observes the obstacle information around the vehicle 100 and near the parking path 200. When no obstacle is found, the vehicle system obtains the distance and speed parameters when there are no obstacles based on the parking path. The distance and speed parameters are input into the parking planning model, and the corresponding control parameters are obtained through the parking planning model to control the vehicle to park at the target parking position 300.
[0102] like Figure 8 The diagram illustrates a parking path planning application scenario provided by some embodiments of this application. In the diagram, vehicle 100 obtains its current vehicle position and target parking position 300 via a positioning sensor, then plans a parking path 200. Combined with a vision sensor, it observes obstacle information around vehicle 100 and near the parking path 200. Upon detecting an obstacle, the positioning sensor determines its location. The vehicle system obtains distance and speed parameters based on the parking path 200 and the obstacle's location, inputs these parameters into the parking planning model, and obtains corresponding control parameters through the parking planning model to control the vehicle to park at the target parking position 300.
[0103] like Figure 9 The diagram shown is a schematic of a parking path planning application scenario provided by another embodiment of this application. After a pause, the vehicle 100 re-obtains its current vehicle position and target parking position 300 through the positioning sensor and plans a parking path 200. Then, combined with the vision sensor, it observes the obstacle information around the vehicle 100 and near the parking path 200. After no obstacles are found, the vehicle system obtains distance and speed parameters based on the parking path 200. Since the parking path is small, the speed parameter will be set even smaller. The distance and speed parameters are input into the parking planning model, and the corresponding control parameters are obtained through the parking planning model to control the vehicle to stop at the target parking position 300.
[0104] Understandably, the parking control method provided in this application only needs to plan a parking path based on the vehicle's current position, target parking position, and obstacle positions obtained from positioning sensors and fisheye cameras. The distance and speed parameters determined based on the parking path are then input into a pre-built parking planning model. The parking planning model calculates the parking control parameters, and the vehicle's infotainment system can then control the vehicle to park safely and comfortably based on these parameters. The parking control method provided in this application uses fewer types of sensors, does not require specific calibration of various environmental parameters, has a small calibration data volume, short computation time for the vehicle's infotainment system, and results in high parking efficiency and comfort.
[0105] Additionally, in other parking control scenarios, such as parking braking control, the parking braking phase refers to the final braking phase during each maneuver in parking.
[0106] Current parking braking accuracy control is generally affected by three factors: 1. Environmental factors around the vehicle, including road slope, road depressions, undulations, and road traffic facilities (speed bumps, manhole covers, parking locks); 2. Control accuracy of the actuators; 3. Sensing accuracy of vehicle sensors (wheel speed sensors, vehicle inertial sensors, steering wheel angle sensors). These factors combined make it difficult to achieve uniform accuracy control in parking braking, requiring specific logic design and targeted calibration based on the scenario to overcome inconsistencies. Control methods generally include: traditional controller solutions (such as PID controllers and sliding mode controllers) and model-based controller solutions (LQR controllers and MPC controllers). Controllers often require manual calibration or the addition of logic to adapt to the characteristics of different actuators, making it difficult to achieve self-learning and adjust their control strategies or performance according to the characteristics of the actuators.
[0107] Research has revealed the following pain points in current parking and braking systems:
[0108] (1) Parking control algorithms require high precision, but the control precision of the actuators varies under different conditions (vehicle speed, braking force, and different specific actuators involved in the execution). A single control algorithm cannot achieve uniform high-precision control for braking mechanisms from different suppliers under a set of calibration parameters;
[0109] (2) For the control during the braking phase, due to the lack of effective sensor information, or the difficulty in providing high-precision signals from the vehicle body sensors when the vehicle is close to stopping, it is difficult to achieve feedback control through sensor information in this situation.
[0110] (3) Traditional control algorithms only provide feedback on input errors, making it difficult to observe actuator characteristics and thus adjust control strategies accordingly.
[0111] (4) Traditional PID control schemes, while relatively simple to implement, generally employ open-loop control based on experience or designing corresponding models to address the problem of rapid error convergence when the system has multiple control equilibrium points. This approach typically requires a large calibration quantity, and the calibration process is lengthy and complex, demanding a high level of experience from the calibration personnel.
[0112] (5) Traditional control schemes involve a large amount of calibration and calibration work. The internal calibration parameters are generally difficult to completely decouple for each scenario. Calibration personnel need to calibrate each scenario individually, and then verify all scenarios as a whole after all scenarios have been calibrated individually, which is time-consuming. If the scenario parameters cannot be completely decoupled, calibration personnel need to make some compromises on performance.
[0113] (6) While model-based control schemes have high compatibility with general scenarios, they also require higher computing power from the hardware platform compared to traditional control schemes. Furthermore, model-based control algorithms cannot completely simulate reality; therefore, any discrepancies between the two still need to be defined and calibrated manually. Thus, calibration cannot be completely avoided.
[0114] In view of this, the parking control method of this application embodiment can also learn the characteristics of the actuator to make targeted control schemes based on the characteristics of the actuator to the greatest extent. It can not only adapt to actuators provided by different suppliers, but also has a very simple implementation. The requirements for computing power and memory are much smaller than those of model-based control schemes. With the help of the designed machine learning algorithm, high-precision braking control with reduced or even no calibration is achieved, thereby helping to solve at least part of the above problems.
[0115] The application of the parking control method of this application in the above scenario will be specifically described below with reference to specific embodiments.
[0116] like Figure 10 As shown, in some embodiments, the parking control parameters include the predicted drive torque of the vehicle on a target path, which is at least a portion of the parking path. Step S3 can control the vehicle to travel along the parking path based on the parking control parameters by: determining the final acceleration based on the predicted drive torque, the actual drive torque of the vehicle, and the initial acceleration of the vehicle on the target path; and controlling the vehicle to travel along the target path until it stops based on the final acceleration.
[0117] For example, target path 210 is the final path in parking path 200, that is, the final braking phase in the entire parking path 200 (e.g., Figure 10 , Figure 11 and Figure 12 (As shown).
[0118] Specifically, vehicle acceleration is a physical quantity that describes how quickly a vehicle's speed changes, reflecting its ability to increase (or decrease) speed per unit time. When acceleration is positive, the vehicle's speed continuously increases, corresponding to the acceleration process; when acceleration is negative, the vehicle's speed continuously decreases, corresponding to the deceleration process. Vehicle driving torque refers to the torque by which the power output from the engine (or electric motor) is transmitted to the wheels through the transmission system, causing the wheels to rotate and propelling the vehicle forward (or backward). Torque is the product of force and lever arm, and its unit is Newton-meter (N·m).
[0119] It is worth noting that the vehicle's braking phase is independent of the target location, the length and curvature of the parking path 200, and the vehicle's direction of travel. During the braking phase, the vehicle's acceleration before entering the target path 210 is first acquired and used as the initial acceleration for that path. The relationship between vehicle torque and acceleration is influenced by many objective factors, such as vehicle weight, ground resistance, and air resistance. Therefore, it is difficult to establish a simple formula relating torque and acceleration. The initial acceleration is input into the parking planning model to obtain the predicted drive torque. The predicted drive torque is compared with the actual drive torque to determine the vehicle's final acceleration. This final acceleration is then sent to the vehicle's actuators, which control the braking system to bring the vehicle to a stop.
[0120] Understandably, by predicting the vehicle's driving torque during parking by measuring the vehicle's acceleration, and then determining the final acceleration based on the predicted and actual driving torque, effective control of the vehicle's braking can be achieved, thus improving the accuracy of vehicle braking.
[0121] In some embodiments, determining the final acceleration based on the predicted drive torque, the actual drive torque of the vehicle, and the initial acceleration of the vehicle on the target path includes: determining the deviation between the predicted drive torque and the actual drive torque of the vehicle; obtaining an acceleration compensation value based on the deviation; and determining the final acceleration based on the acceleration compensation value and the initial acceleration.
[0122] Specifically, a parking planning model establishes a relationship between initial acceleration and driving torque, thereby predicting the vehicle's current driving torque based on the initial acceleration. To obtain more accurate data, the difference between the predicted and actual driving torque is calculated to obtain the deviation result. When the difference between the predicted and actual driving torque is greater than 0, it indicates that the predicted driving torque is larger than the actual driving torque. In this case, the acceleration compensation value is set to a negative value, and the initial acceleration is added to the acceleration compensation value to obtain a smaller final acceleration. The vehicle is then stopped by controlling the vehicle with the reduced final acceleration. When the difference between the predicted and actual driving torque is less than 0, it indicates that the predicted driving torque is smaller than the actual driving torque. In this case, the acceleration compensation value is set to a positive value, and the initial acceleration is added to the acceleration compensation value to obtain a larger final acceleration. The vehicle is then stopped by controlling the vehicle with the increased final acceleration. When the difference between the predicted and actual driving torque is equal to 0, it indicates that the predicted and actual driving torques are equal. In this case, the acceleration compensation value is set to 0, and the initial acceleration becomes the final acceleration. The vehicle is then stopped by controlling the vehicle with the final acceleration.
[0123] Understandably, by predicting the vehicle's driving torque through a parking planning model, the deviation between the predicted and actual driving torque is used to correct the vehicle's acceleration, thereby achieving effective control over the vehicle's braking and improving the accuracy of braking.
[0124] In some embodiments, the method for constructing a parking planning model includes: constructing an initial planning model based on an LSTM network; obtaining historical motion parameters of the vehicle, including historical distance parameters and historical speed parameters; preprocessing the historical motion parameters and training the initial planning model based on the preprocessed historical motion parameters to obtain a trained parking planning model; the historical speed parameters include historical driving speed, historical acceleration, historical jerk, historical driving torque, and time planning; the historical distance parameter is used to characterize the historical remaining distance of the parking path during the parking process, the historical driving torque is used to characterize the actual torque transmitted from the vehicle's drive system to the drive wheels, and the time planning is used to characterize the time it takes for the vehicle to complete the parking path.
[0125] Specifically, the historical driving speed, historical acceleration, historical jerk, historical driving torque, time planning, and historical remaining distance are first input into the constructed initial planning model for training. Training with multiple sets of data enables the initial planning model to establish a mapping relationship between driving speed, acceleration, jerk, remaining distance, and driving torque. Through continuous training, a parking planning model is obtained. The completed parking planning model can predict the vehicle's driving torque using driving speed, acceleration, jerk, and remaining distance.
[0126] In some embodiments, the initial planning model is built using an LSTM network, which includes a six-layer structure, such as... Figure 13 As shown, the layers are: first layer: Long Short-Term Memory network (LSTM); second layer: normalization layer; third layer: Long Short-Term Memory network; fourth layer: normalization layer; fifth layer: fully connected layer; and sixth layer: cross-entropy loss layer. The first-layer LSTM network acts as a primary sequence feature extractor, capturing the local dependencies of historical motion parameters of the input through gating mechanisms (input gate, forget gate, output gate). The second-layer normalization layer normalizes the feature dimensions of the output of the first-layer LSTM network (mean 0, variance 1) and introduces learnable scaling and translation parameters to stabilize the input distribution of subsequent layers. The third-layer LSTM network acts as a higher-order sequence feature extractor, further mining the global dependencies of the data from the normalized primary features. The fourth-layer Layer Normalization layer normalizes the output of the third-layer LSTM network again, providing stable input for the fully connected layer. At this time, the normalization parameters may learn a different scaling strategy than the previous layer to adapt to the distribution characteristics of higher-order features. The fifth-layer Fully Connected layer achieves a non-linear mapping from features to the target through weight matrices and non-linear activation functions (such as Softmax), thereby integrating the temporal features extracted by the LSTM network and generating a task-oriented output space mapping. The sixth layer Cross Entropy (cross-entropy loss) drives parameter optimization through backpropagation to measure the difference between the model's predicted distribution and the true label distribution.
[0127] It is worth noting that the parking planning model incorporates a nested combination of two LSTM networks and a Layer Normalization layer. The first LSTM layer is primarily used to extract local features from the data, while the third LSTM layer focuses on global features. Both layers are stabilized through a Layer Normalization layer, ensuring that the learning of higher-order features is not affected by lower-order distribution shifts, thereby enhancing gradient stability.
[0128] Understandably, an initial planning model is obtained by building an LSTM network and Layer Normalization layer, and historical motion parameters are input into the initial planning model for training. This results in a parking planning model that can predict the driving torque of the vehicle based on driving speed, acceleration, jerk, and remaining distance, thus realizing the construction of the mapping relationship between driving torque and driving speed, acceleration, jerk, and remaining distance.
[0129] In some embodiments, the method for preprocessing historical motion parameters includes: normalizing the vehicle speed, acceleration, jerk, and drive torque; standardizing the time plan; arranging the normalized vehicle speed, acceleration, jerk, and drive torque in sequence with the standardized time plan; cyclically slicing the matched historical motion parameters based on a preset data extraction window to obtain multiple datasets; dividing the multiple datasets into a training set and a validation set, wherein the training set is used to train the initial planning model and the validation set is used to validate the initial planning model.
[0130] Specifically, the acquired historical vehicle movement data is converted into a data format that is easy to process or standardized using appropriate format conversion tools. For example... Figure 14 As shown, this demonstrates how to convert a file with the ".db" extension to a ".csv" file, or vice versa. Typically, the SQLite command-line tool can be used to convert ".db" files to ".csv" files. The process is as follows: First, open the command line and navigate to the directory containing the ".db" file; second, connect to the database; third, execute the SQL query and export the result as a ".csv" file. Alternatively, a Python script can be written to perform the conversion. The ROS command-line tools `rosbag` and `rostopic` can be used to convert ".bag" files to ".csv" files.
[0131] In some embodiments, the data in the file is filtered after the file is converted to a uniform format. For example, when using a Python script to extract data from a ".csv" file, the validity of the data needs to be judged during the extraction process. The judgment criteria are: (1) whether the final braking accuracy of the vehicle is less than the braking threshold; (2) whether there is a parking interruption during the parking process; and (3) whether the braking process of the vehicle is complete. In point (1), the braking accuracy of the vehicle is judged by the distance between the final braking position of the vehicle and the target position, and the braking threshold is 5cm-15cm. In point (2), the speed and time during the parking process are judged by whether they are continuous. If they are not continuous, it means that the vehicle has experienced a parking interruption. If they are continuous, it means that the vehicle has not experienced a parking interruption. In point (3), the parking planning path and the actual path are judged by whether they are the same. If they are the same, it means that the parking process is complete. If it is determined that the vehicle meets the following conditions during the parking process: braking accuracy is less than the braking threshold, parking has not been interrupted, and the parking process is complete, then the data is considered valid. Figure 14 As shown in the graph, the parking process was interrupted in the "Emergency Braking" section, therefore this data is not accepted. Similarly, the parking process was incomplete in the graph for the "Comfort Braking Abort" section, so this data is also unacceptable. However, the parking process was complete and uninterrupted in the graph for the "Regular Comfort Braking" section, therefore this data is acceptable. The valid data obtained after filtering includes:
[0132] Vehicle speed: {v n |v1,v2,v3,…,v n};
[0133] Vehicle acceleration: {a n |a1,a2,a3,…,a n};
[0134] Vehicle driving torque: {T n |T1,T2,T3,…,T n};
[0135] Vehicle time planning: {t n |t1,t2,t3,…,t n};
[0136] Based on the above data, the vehicle's jerk can be determined using either the first derivative of acceleration with respect to time or the second derivative of velocity with respect to time. Vehicle jerk: {jerk} n |jerk1,jerk2,jerk3,…,jerk n};like Figure 15As shown, the vehicle's speed, acceleration, jerk, and drive torque are processed using the Min_MaxNormalization() function to perform minimum-maximum normalization, scaling these parameters to a uniform range (e.g., [0,1] or a custom interval) to eliminate the impact of dimensional differences on the initial planning model. The vehicle's time planning is processed using the standardization() function, scaling the time features to a distribution with a mean of 0 and a standard deviation of 1, thereby improving the training efficiency and stability of the initial planning model. The speed, acceleration, drive torque, and jerk at the same moment are packaged and arranged in chronological order.
[0137] like Figure 16 As shown, after filtering, a data slicing window can be set up for the valid data. Slicing is done in groups of 50 data points, converting the sliced data into a Tensor type dataset. This dataset is then divided into a training set {X_train} and a validation set {X_test}. The training set {X_train} is used for initial training of the initial planning model, while the validation set {X_test} is used for later validation, thereby improving the model's prediction accuracy. Typically, the ratio of training to validation data is 7:3, but this ratio can be adjusted as needed.
[0138] It is worth noting that, such as Figure 16 As shown, X1 represents the data for the first second, including the vehicle speed v in the first second. 1s The acceleration a in the first second 1s Jerk's accelerometer in the first second 1s The driving torque T in the first second 1s And the time t of the first second 1s Additionally, h1 to h 52 For the hidden layers of the primary planning model, U1 to U 52 This is the output layer of the primary planning model.
[0139] The training set {X_train} obtained after slicing is input into the primary planning model for training. The trained primary planning model is then validated using the validation set {X_test}. The final parking planning model is obtained when the difference between the predicted driving torque output by the primary planning model and the historical driving torque is within the error range.
[0140] like Figure 20 As shown, a parking planning model is applied to a vehicle to predict its driving torque. The initial driving speed v during the parking process is used as the basis for this prediction. in Initial acceleration a in Initial jerkin The parking planning model is input with the initial time, and the predicted driving torque T is obtained from the parking planning model. OUT The output is sent to the feedback controller, which then receives the actual driving torque T of the vehicle. in With predicted drive torque T OUT By comparison, the acceleration compensation value a is obtained. offset And the acceleration compensation value a is added through an adder. offset With initial acceleration a in The final acceleration a is obtained by adding them together. final Through the final acceleration a final Control the vehicle to park until it comes to a complete stop. For example... Figure 17 The image shows a comparison between the predicted drive torque and the actual drive torque. Figure 18 The figure shows the curves illustrating the change in acceleration of a vehicle during the same time period. Figure 19 The graph shows the changes in speed V, acceleration a, jerk, and driving torque T over time during the vehicle's braking process.
[0141] It is understandable that by acquiring historical vehicle motion data and performing a series of preprocessing steps on the historical motion data, the format and type of the data can be standardized to meet the training and validation conditions of the model, thereby improving the training efficiency of the model and further enhancing its stability.
[0142] In summary, this embodiment of the application preprocesses the historical motion data of the vehicle and then slices the data. The standardized discrete data is then reorganized according to the time dimension to obtain a dataset. The dataset is then split into a training set and a validation set for training and validation of the established initial planning model. After training, the real-time parameters of the vehicle are input into the parking planning model for predicting the actuator braking torque. Finally, closed-loop control is performed with the actual driving torque, and the acceleration compensation value is updated according to the error between the predicted driving torque and the actual torque to achieve high-precision control of the vehicle braking to a stop.
[0143] Correspondingly, such as Figure 21 As shown in the figure, this application embodiment also provides a parking control system, the system including: a parameter acquisition module, configured to determine distance parameters and speed parameters based on a planned parking path, the distance parameter being used to characterize the remaining distance of the parking path during the parking process, and the speed parameter being used to characterize the vehicle speed during the parking process; a parameter input module, configured to input the distance parameters and speed parameters into a pre-built parking planning model to obtain parking control parameters; and a vehicle control module, configured to control the vehicle to drive along the parking path based on the parking control parameters.
[0144] like Figure 22 As shown, this application embodiment provides a parking control system module control, including fisheye cameras, a vehicle digital cockpit unit, a microcomputer, an automatic transmission control unit, and a vehicle bus; wherein, there are at least four fisheye cameras, respectively arranged around the vehicle body, for acquiring image data of the vehicle's surroundings; the vehicle digital cockpit unit includes a perception module; the microcomputer includes a positioning module, a slot planning module, and a behavior planning module; the automatic transmission control unit includes a vehicle speed control module; the fisheye cameras are used to acquire image data; the perception module is used to acquire image data and process it, transmitting the processed fusion result (including the target parking position and obstacle information) to the microcomputer via Ethernet. Based on the vehicle's current position provided by the vehicle positioning module, the microcomputer, combining the target parking position and obstacle information provided by the vehicle digital cockpit unit, plans a parking path curve in the slot planning module; finally, the behavior planning module post-processes the path curve and converts it into parking longitudinal control compatible interface signals (i.e., distance parameters and speed parameters, where the distance parameter includes the remaining path and the speed parameter includes the vehicle speed during parking). The microcomputer sends the processed interface commands to the automatic transmission control unit via the CAN (Controller Area Network) interface. The vehicle speed control module running in the automatic transmission control unit plans a smooth acceleration curve based on the information provided by the behavior planning module, and sends this acceleration command to the whole vehicle via the CAN network to achieve vehicle control and realize complete closed-loop control verification of the whole vehicle.
[0145] like Figure 23The diagram shows the motion planning algorithm framework for a parking control system, including a parking speed initialization area, a parking speed planning module, and a parking speed control area. The parking speed initialization area includes an algorithm initialization module, which is responsible for initializing the entire parking motion planning algorithm, handling the constraints of parking motion control, and determining whether to perform replanning based on the planned trajectory and the vehicle's tracking status. The parking speed planning area includes time planning algorithms, speed extremum planning algorithms, acceleration extremum algorithms, jerk extremum algorithms, Cost function algorithms, and evaluation function algorithms. The time planning algorithm is used to plan the time taken for the vehicle to park based on the parking path, and the speed extremum planning algorithm is used to calculate and obtain the speed extrema. The acceleration extremum algorithm is used to calculate the extremum of acceleration, the jerk extremum algorithm is used to calculate the extremum of jerk, the Cost function algorithm is used to constrain various planning results according to the constraints, and the evaluation function algorithm is used to evaluate the remaining planning results to obtain the optimal parking control parameters. The parking speed control zone includes the blind spot area, the PD (Proportional-Derivative) controller, and the acceleration limiting module. The parking control zone determines the control blind spot area, tracks the position and speed errors, and limits the output acceleration through the acceleration limiting module according to the characteristics and physical limits of the PD controller.
[0146] Understandably, the parking control system provided in this application only needs to plan and obtain a parking path based on the vehicle's current position, target parking position, and obstacle positions acquired by positioning sensors and fisheye cameras. The distance and speed parameters determined based on the parking path are then input into a pre-built parking planning model. The parking planning model calculates the parking control parameters, and the vehicle's infotainment system can then control the vehicle to park safely and comfortably based on these parameters. The parking control system provided in this application uses fewer types of sensors, does not require specific calibration of various environmental parameters, has a small calibration data volume, short computation time for the vehicle's infotainment system, and offers high parking efficiency and comfort.
[0147] This application provides a parking control system, including: a memory for storing a program; and a processor for executing the program stored in the memory; when the program stored in the memory is executed, the processor executes the parking control method provided in any of the above embodiments.
[0148] This application provides an autonomous driving device, which includes a parking control system as provided in any of the above embodiments.
[0149] This application provides a computer-readable storage medium that stores instructions for execution by a computing device. When the computing device executes the instructions, it implements the parking control method provided in any of the above embodiments.
[0150] The foregoing has provided a detailed description of a parking control method, system, autonomous driving device, and readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the technical solutions and core ideas of this application. Those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
Claims
1. A parking control method characterized by, The method comprises the following steps: determining a distance parameter and a speed parameter based on a planned parking path, the distance parameter being used to represent a remaining distance of the parking path during parking of the vehicle, and the speed parameter being used to represent a vehicle speed during parking of the vehicle; inputting the distance parameter and the speed parameter into a pre-constructed parking planning model to obtain a parking control parameter; controlling the vehicle to travel along the parking path based on the parking control parameter.
2. The parking control method according to claim 1, characterized by, The method for constructing the parking planning model comprises the following steps: constructing a motion planning function based on vehicle motion parameters, the motion planning function being configured to output a plurality of planning results based on the distance parameter and the speed parameter; constructing a constraint function based on a constraint condition, so as to select the planning result meeting the constraint condition from the plurality of planning results through the constraint function; constructing an evaluation function based on an evaluation condition, so as to screen the selected planning result through the evaluation function to obtain the parking control parameter.
3. The parking control method according to claim 2, characterized by, The method for constructing the motion planning function based on vehicle motion parameters comprises the following steps: constructing a plurality of polynomial functions based on vehicle motion parameters, the vehicle motion parameters comprising a driving distance, a driving speed, an acceleration and a jerk; determining a time planning based on the parking path, the time planning being used to represent a time for the vehicle to complete the parking path; adjusting the plurality of polynomial functions based on the time planning to obtain the motion planning function.
4. The parking control method according to claim 2, characterized by, The constraint condition comprises at least one of a maximum parking speed, a maximum acceleration, a minimum acceleration, a maximum jerk and a minimum jerk; The method for selecting the planning result meeting the constraint condition from the plurality of planning results through the constraint function comprises the following steps: deriving the planning result to obtain extreme values corresponding to different vehicle motion parameters; obtaining an overshoot amount by subtracting the constraint condition from the extreme values; screening the planning result based on the size of the overshoot amount.
5. The parking control method according to claim 3, characterized by, The method for screening the selected planning result through the evaluation function to obtain the parking control parameter comprises the following steps: determining a timeliness of the planning result based on the time planning, the timeliness being represented by a length of the parking time of the vehicle; determining a comfort of the planning result based on the jerk, the comfort being represented by a dynamic characteristic of the vehicle; selecting the planning result meeting the comfort and the timeliness from the planning result as the parking control parameter.
6. The parking control method according to claim 3, characterized by, The method for planning the parking path comprises the following steps: determining a current position of a vehicle, a target parking position and a position of an obstacle; planning a parking path based on the current position of the vehicle, the target parking position and the position of the obstacle.
7. The parking control method according to claim 6, characterized by, The distance parameter comprises a first distance between the vehicle and the target parking position and a second distance between the vehicle and the target parking position during motion, the speed parameter comprises an initial speed of the vehicle and a target speed of the vehicle along the parking path, and the parking control parameter comprises the acceleration of the vehicle at each time point. The method for inputting the distance parameter and the speed parameter into a pre-constructed parking planning model to obtain a parking control parameter comprises: inputting the first distance, the second distance, the initial speed and the target speed into the parking planning model to obtain the acceleration of the vehicle at each time point.
8. The parking control method according to claim 7, characterized by, The method for determining the target speed of the vehicle along the parking path comprises: determining the target speed of the vehicle along the parking path based on the first distance; the target speed is positively correlated with the first distance.
9. The parking control method according to claim 1, characterized by, The parking control parameter comprises a predicted driving torque of the vehicle on a target path, the target path being at least part of the parking path. The method for controlling the vehicle to travel along the parking path based on the parking control parameter comprises: determining a final acceleration based on the predicted driving torque, an actual driving torque of the vehicle and an initial acceleration of the vehicle on the target path; controlling the vehicle to travel along the target path until stopping based on the final acceleration.
10. The parking control method according to claim 9, characterized by, Determining a final acceleration based on the predicted driving torque, an actual driving torque of the vehicle and an initial acceleration of the vehicle on the target path comprises: determining a deviation result of the predicted driving torque and the actual driving torque of the vehicle; obtaining an acceleration compensation value based on the deviation result; determining the final acceleration based on the acceleration compensation value and the initial acceleration.
11. The parking control method according to claim 1, characterized by, The method for constructing the parking planning model comprises: constructing an initial planning model based on an LSTM network; obtaining historical motion parameters of the vehicle, the historical motion parameters comprising historical distance parameters, historical speed parameters and historical control parameters; preprocessing the historical motion parameters, and training the initial planning model based on the preprocessed historical motion parameters to obtain the trained parking planning model; the historical speed parameters comprise historical driving speed, historical acceleration, historical jerk and time planning; the historical distance parameters are used to represent historical remaining distances of the parking path in the parking process of the vehicle; the historical control parameters comprise a historical driving torque, the historical driving torque being used to represent an actual torque transmitted to a driving wheel by a driving system of the vehicle; and the time planning is used to represent a time for completing the parking path by the vehicle.
12. The parking control method according to claim 11, characterized by, The method for preprocessing the historical motion parameters comprises: normalizing the driving speed, the acceleration, the jerk and the driving torque; standardizing the time planning; sequentially arranging the normalized driving speed, the normalized acceleration, the normalized jerk and the normalized driving torque and the standardized time planning; performing cyclic slicing on the matched historical motion parameters based on a preset data extraction window to obtain a plurality of data sets; dividing the plurality of data sets into a training set and a verification set, the training set being used for data training of the initial planning model, and the verification set being used for data verification of the initial planning model.
13. The parking control method according to any one of claims 9 to 12, characterized by, The target path is a final stopping path of the parking path.
14. A parking control system, characterized by The system comprises: The parameter acquisition module is configured to determine a distance parameter and a speed parameter based on the planned parking path, the distance parameter being used to represent a remaining distance of the parking path in the parking process of the vehicle, and the speed parameter being used to represent a vehicle speed in the parking process of the vehicle. The parameter input module is configured to input the distance parameter and the speed parameter into a pre-constructed parking planning model to obtain a parking control parameter. The vehicle control module is configured to control the vehicle to travel along the parking path based on the parking control parameter.
15. A parking control system, characterized by The computer readable medium stores instructions for a computing device to execute, and the computing device, when executing the instructions, implements the parking control method according to any one of claims 1-13. The automatic driving device comprises the parking control system according to claim 14 or 15. The computer readable medium stores instructions for a computing device to execute, and the computing device, when executing the instructions, implements the parking control method according to any one of claims 1-13. 16. An autonomous driving device, comprising: 17. A computer-readable storage medium, characterized in that,