Automatic parking method and system oriented to complex environment

Abstract

The invention provides an automatic parking method and system for a complex environment, and the method comprises the steps: building a grid map, and determining an initial position and a target position; determining a search direction based on a dynamic weighted heuristic function; setting the initial position as a father node, and expanding child nodes according to the search direction; calculating the estimated cost and the actual cost of the extended child nodes, and determining a target node; smoothly planning the trajectory according to the target node to obtain a global optimal trajectory; and updating the global optimal track in real time, and controlling the intelligent vehicle to park based on the real-time global optimal track. According to the automatic parking method and system oriented to the complex environment, the grid map is established, the initial position and the target position are determined, the dynamic weighted heuristic function is introduced to determine the search direction, the nodes are expanded, and the expanded child nodes are searched and obtained; the target node is determined according to the estimated cost and the actual cost of the extended child node, and the trajectory is smoothly planned in combination with the artificial potential field, so that the parking process is safer, and the stability is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic parking, and in particular to an automatic parking method and system for complex environments. Background Art

[0002] As the number of cars increases, the parking environment drivers face becomes more and more complex, making parking a challenge. Automatic parking technology integrates advanced sensors and complex algorithms to enable vehicles to perform precise parking operations. The specific process generally includes: environmental perception, parking space detection and recognition, parking path planning, parking path following control, and simulation display. Automatic parking technology can save drivers' time and reduce the complexity of parking.

[0003] However, current automatic parking algorithms are mostly global static planning, and have problems such as low safety and poor efficiency when faced with complex parking environments (such as busy commercial areas, multi-storey parking lots, etc.).

[0004] In view of this, there is an urgent need for an automatic parking method and system for complex environments to at least solve the above-mentioned deficiencies. Summary of the invention

[0005] One of the purposes of the present invention is to provide an automatic parking method and system for complex environments, establish a grid map of the parking site and determine the starting position and target position, introduce a dynamic weighted heuristic function to determine the search direction, expand the node search to obtain the extended sub-node, determine the target node according to the estimated cost and actual cost of the extended sub-node, and combine the artificial potential field to smoothly plan the trajectory, which greatly improves the safety, real-time and stability of the vehicle.

[0006] An automatic parking method for complex environments provided by an embodiment of the present invention includes:

[0007] Create a grid map of the parking site and determine the starting and target locations;

[0008] Determine the search direction based on the dynamic weighted heuristic function;

[0009] Set the starting position to the parent node and get the extended child nodes according to the search direction;

[0010] Calculate the estimated cost and actual cost of expanding the child node and determine the target node;

[0011] According to the smooth planning trajectory of the target node, the global optimal trajectory is obtained;

[0012] The global optimal trajectory is updated in real time, and the smart car is controlled to park based on the real-time global optimal trajectory.

[0013] Preferably, a grid map of the parking scene is established to determine the starting position and the target position, including:

[0014] Obtain parking environment information;

[0015] A grid map is created based on the parking environment information to determine the starting position and target position.

[0016] Preferably, calculating the estimated cost and the actual cost of expanding the child node and determining the target node includes:

[0017] According to the estimated cost of the expanded child node, the redundant nodes with the same estimated cost in the adjacent trajectories are eliminated, and then the target node with the minimum estimated cost in the adjacent trajectories is determined.

[0018] Preferably, the trajectory is smoothly planned according to the target node to obtain the global optimal trajectory, including:

[0019] Determine the introduction field according to the target node;

[0020] Determine the repulsive field based on the non-search direction and the redundant nodes;

[0021] Use the gravitational field and repulsive field to smoothly plan the trajectory and obtain the global optimal trajectory.

[0022] Preferably, establishing a grid map based on the parking environment information and determining the starting position and the target position includes:

[0023] Try to obtain the parking space selection signal of the smart car driver;

[0024] If the parking space selection signal is successfully acquired, the parking space selection signal is parsed to determine the target location;

[0025] If the parking space selection signal acquisition attempt fails, the target location will be searched intelligently.

[0026] Preferably, attempting to obtain a parking space selection signal from a driver of the smart car includes:

[0027] Displaying a three-dimensional panoramic image of the parking environment to the driver of the smart car;

[0028] Acquire a local three-dimensional panoramic image selected by a driver of the smart car;

[0029] Extracting image features of the local 3D panoramic image, the image features including: traffic parameters and risk parameters of the actual area corresponding to the local 3D panoramic image;

[0030] If the image features meet the standard image features, the local three-dimensional panoramic image is rasterized to determine the grid coordinates, and the parking space selection signal is determined according to the grid coordinates;

[0031] If the image features do not meet the standard image features, it is determined that the attempt to obtain the parking space selection signal has failed.

[0032] Preferably, intelligently searching for a target location includes:

[0033] Based on the grid map, try to find the standard parking space line area;

[0034] If the attempt to find is successful, the standard parking space line area is used as the target location;

[0035] If the search attempt fails, the parking space characteristics are obtained based on the three-dimensional panoramic model of the parking environment;

[0036] Determine the target parking space characteristics according to the parking space characteristics and the body parameters of the smart car;

[0037] Determine the simulated parking space according to the characteristics of the target parking space;

[0038] Obtain the driving control parameters of the smart car;

[0039] According to the driving control parameters and the body parameters of the smart car, the smart car model is simulated to park in the simulated parking space in the 3D panoramic model;

[0040] During the simulated parking process, determine whether there is a simulated parking situation in which no scratches occur with the obstacle model;

[0041] If so, the grid coordinates corresponding to the target parking space features are used as the target position.

[0042] Preferably, according to the driving control parameters and the body parameters of the smart car, simulating the smart car model to park in the simulated parking space in the three-dimensional panoramic model includes:

[0043] Get the simulated parking type;

[0044] Obtain historical parking records based on simulated parking types;

[0045] According to the historical parking records, extract the relative relationship feature sequence between the parking space line and the vehicle body;

[0046] Determine, based on the driving control parameters and the simulated parking space line, a first adjustment cost of adjusting the simulated intelligent vehicle model to a relative relationship feature;

[0047] Determine a second adjustment cost of the relative relationship feature corresponding to the historical vehicle posture;

[0048] Determine a first target relative relationship feature whose first adjustment cost is less than a second adjustment cost, and use the first target relative relationship feature which is relatively late in the relative relationship feature sequence as the second target relative relationship feature;

[0049] The second target relative relationship feature and the relative relationship feature after the second target relative relationship feature in the relative relationship feature sequence are used as a simulated intelligent vehicle model relative feature sequence;

[0050] According to the relationship feature sequence of the simulated smart car model, the smart car model is simulated to park in the simulated parking space in the three-dimensional panoramic model.

[0051] An automatic parking method for complex environments provided by an embodiment of the present invention further includes:

[0052] Obtaining a target adjustment cost of the target vehicle corresponding to the relative relationship feature;

[0053] Wherein, obtaining the target adjustment cost of the target vehicle corresponding to the relative relationship feature includes:

[0054] Obtaining a manual driving model of a target vehicle;

[0055] Determine vehicle control parameters based on relative relationship characteristics and artificial driving model;

[0056] Determine a steering wheel steering increment and a travel increment according to vehicle control parameters;

[0057] The steering wheel steering increment and travel increment are normalized to obtain the target adjustment cost.

[0058] An embodiment of the present invention provides an automatic parking system for complex environments, comprising:

[0059] A grid map building subsystem is used to build a grid map of the parking site and determine the starting position and the target position;

[0060] A search direction determination subsystem, used to determine the search direction based on a dynamic weighted heuristic function;

[0061] The child node expansion subsystem is used to set the starting position to the parent node and obtain the expanded child nodes according to the search direction;

[0062] The target node determination subsystem is used to calculate the estimated cost and actual cost of the extended subnode and determine the target node;

[0063] The global optimal trajectory planning subsystem is used to smoothly plan the trajectory according to the target node to obtain the global optimal trajectory;

[0064] The parking subsystem is used to update the global optimal trajectory in real time and control the smart car to park based on the real-time global optimal trajectory.

[0065] The beneficial effects of the present invention are:

[0066] The present invention establishes a grid map of the parking scene and determines the starting position and the target position, introduces a dynamic weighted heuristic function to determine the search direction, expands the node search to obtain the extended sub-node, determines the target node according to the estimated cost and actual cost of the extended sub-node, and combines the artificial potential field to smoothly plan the trajectory, which greatly improves the safety, real-time and stability of the vehicle.

[0067] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the present application documents.

[0068] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0070] Figure 1 is a schematic diagram of an automatic parking method for complex environments according to an embodiment of the present invention;

[0071] Figure 2 Schematic diagram of an automatic parking system for complex environments in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0073] The embodiment of the present invention provides an automatic parking method for complex environments, such as Figure 1 As shown, including:

[0074] Step 1: Create a grid map of the parking scene and determine the starting position and target position; the grid map is a data structure that divides the parking environment into grids, each grid represents a portion of the three-dimensional point cloud data of the parking environment, and the gridding rules are manually preset; the starting position is the relative position of the current position of the smart car that needs to be parked in the grid map; the target position is the relative position of the parking position in the grid map;

[0075] Step 2: Determine the search direction based on the dynamic weighted heuristic function; the heuristic function is: a function that estimates the cost of the starting position and the target position, including the actual cost and the heuristic estimated cost. In order to improve the search efficiency, a dynamically changing weight coefficient is introduced before the heuristic estimated cost. At the beginning of the search, it is more important to quickly reach the destination area, so a larger weight coefficient can be set to speed up the search; at the end of the search, it is more important to obtain the best path to the target, so the weight coefficient can be reduced to improve the optimality of the path. The specific weight coefficient value is set manually after meeting the above setting rules; the search direction is: the path node that is preferentially expanded according to the dynamic weighted heuristic function;

[0076] Step 3: Set the starting position as the parent node, and obtain the extended child nodes according to the search direction; Set the starting position as the parent node, and store the end position in the memory list, and expand the child nodes based on the search direction;

[0077] Step 4: Calculate the estimated cost and actual cost of the expanded child node and determine the target node; the estimated cost is the estimated cost of the best path from the current node to the target node; the actual cost is the cumulative cost of the path from the starting point to the current node;

[0078] Step 5: Smoothly plan the trajectory according to the target node to obtain the global optimal trajectory;

[0079] Step 6: Update the global optimal trajectory in real time, and control the smart car to park based on the real-time global optimal trajectory;

[0080] The step 1: establishing a grid map of the parking site and determining the starting position and the target position includes:

[0081] Acquire parking environment information; wherein the parking environment information is: shooting data and radar scanning data acquired by the on-board camera of the smart car;

[0082] A grid map is established based on the parking environment information to determine the starting position and the target position; the target position includes the parking spaces input manually and the parking spaces intelligently searched by the intelligent driving system;

[0083] The step 4: calculating the estimated cost and actual cost of expanding the child node and determining the target node includes:

[0084] According to the estimated cost of the expanded child node, the redundant nodes with the same estimated cost in the adjacent trajectories are eliminated, and then the target node with the minimum estimated cost in the adjacent trajectories is determined;

[0085] The step 5: smoothly planning the trajectory according to the target node to obtain the global optimal trajectory includes:

[0086] Determine the introduction field according to the target node;

[0087] Determine the repulsive field based on the non-search direction and the redundant nodes;

[0088] The trajectory is planned by smoothing the gravitational field and the repulsive field to obtain the global optimal trajectory. When the trajectory is smoothed by the gravitational field and the repulsive field, the path is adjusted by simulating the introduction and repulsion in the physical field. The gravitational force will attract the vehicle to move to the target position, while the repulsive force will prevent the vehicle from colliding with obstacles.

[0089] The working principle and beneficial effects of the above technical solution are:

[0090] The present invention establishes a grid map of the parking scene and determines the starting position and the target position, introduces a dynamic weighted heuristic function to determine the search direction, expands the node search to obtain the extended sub-node, determines the target node according to the estimated cost and actual cost of the extended sub-node, and combines the artificial potential field to smoothly plan the trajectory, which greatly improves the safety, real-time and stability of the vehicle.

[0091] In one embodiment, a grid map is established based on parking environment information to determine a starting position and a target position, including:

[0092] Try to obtain the parking space selection signal of the smart car driver; wherein the parking space selection signal is: an electrical signal of the selection information of the manual parking space selection during assisted parking;

[0093] If the parking space selection signal is successfully acquired, the parking space selection signal is parsed to determine the target location;

[0094] If the parking space selection signal acquisition attempt fails, the target location will be searched intelligently.

[0095] The working principle and beneficial effects of the above technical solution are:

[0096] The present invention introduces two methods to obtain the target position. The first method is to determine the target position according to the parking space selected by the driver. The second method is to automatically search for the parking space. The target position acquisition process is more intelligent and more reasonable.

[0097] In one embodiment, attempting to obtain a parking space selection signal from a driver of a smart car includes:

[0098] The three-dimensional panoramic image of the parking environment is displayed to the driver of the smart car; wherein the three-dimensional panoramic image is: a top view image of the three-dimensional environment model of the parking environment;

[0099] Acquire a local three-dimensional panoramic image selected by a driver of the smart car;

[0100] Extract image features of the local 3D panoramic image, the image features include: traffic parameters and risk parameters of the actual area corresponding to the local 3D panoramic image; the traffic parameters of the actual area are: the area length and area width of the actual area and the area parameters of the 3D panoramic image connected to the area opening corresponding to the global area; the risk parameters are: obstacle features of obstacles adjacent to the local 3D panoramic image, the obstacle features can read the risk value by comparing with the preset risk value library, the risk value library includes one-to-one corresponding pre-selected obstacle features and risk values, for example: the pre-selected obstacle feature is that the distance between the virtual front of the actual area and the front vehicle is 10cm, and the risk value is: 50;

[0101] If the image features meet the standard image features, the local three-dimensional panoramic image is rasterized to determine the grid coordinates, and the parking space selection signal is determined according to the grid coordinates; wherein the standard image features are extracted according to the parking process image that can successfully park, including: the distance between the successfully parked vehicle and the obstacle, the area passed by the successfully parked vehicle, and the area shape composed of the successful parking position;

[0102] If the image features do not meet the standard image features, it is determined that the attempt to obtain the parking space selection signal has failed.

[0103] The working principle and beneficial effects of the above technical solution are:

[0104] When automatic parking intervenes, humans can also select the parking space they want to park by themselves, and display the three-dimensional panoramic image of the parking environment to the smart car driver through the car screen. The smart car driver selects a local three-dimensional panoramic image through the car screen and extracts image features of the local three-dimensional panoramic image. If the image features meet the standard image features, that is, the actual area corresponding to the local three-dimensional panoramic image can accommodate the body of the smart car, the three-dimensional panoramic image connected by the regional openings corresponds to the global area that can accommodate the body posture space corresponding to the parking trajectory of the smart car, and the risk parameter is less than a preset risk parameter threshold (the preset risk parameter threshold is pre-set by humans, for example: 60), the local three-dimensional panoramic image at this time is rasterized to obtain grid coordinates, and the grid coordinates are converted into a parking space selection signal based on a preset parking space selection signal conversion template to obtain a parking space selection signal. If the image features do not meet the standard image features, it means that the area selected by the driver is not suitable for parking, and a reminder is issued to the driver.

[0105] In one embodiment, intelligently searching for a target location includes:

[0106] According to the grid map, try to find the standard parking line area; the standard parking line area is: the grid map area where the parking lines are drawn;

[0107] If the attempt to find is successful, the standard parking space line area is used as the target location;

[0108] If the attempt to find the parking space fails, the parking space characteristics are obtained according to the three-dimensional panoramic model of the parking environment; wherein the three-dimensional panoramic model of the parking environment is: a three-dimensional reconstruction model of the parking environment, which is reconstructed based on the shooting parameters of the vehicle body camera and the radar scanning data; the parking space characteristics are: three-dimensional data of the parking environment;

[0109] Determine the target parking space characteristics according to the parking space characteristics and the body parameters of the smart car; wherein the target parking space characteristics are: the coordinates of the space area on the ground in the parking environment that can accommodate the body of the smart car;

[0110] Determine the simulated parking space according to the characteristics of the target parking space;

[0111] Obtaining driving control parameters of the smart car; wherein the driving control parameters include: vehicle turning radius, wheel driving mode, etc.;

[0112] According to the driving control parameters and the body parameters of the smart car, the smart car model is simulated to park in the simulated parking space in the 3D panoramic model;

[0113] During the simulated parking process, it is determined whether there is a simulated parking situation in which no scratches occur with the obstacle model; wherein the obstacle model is: other person and object models except the smart car model in the three-dimensional panoramic model;

[0114] If so, the grid coordinates corresponding to the target parking space features are taken as the target position;

[0115] If not, the output is that there is no suitable parking space in the parking environment.

[0116] The working principle and beneficial effects of the above technical solution are:

[0117] When no parking space preference is manually input, the assisted driving system can also automatically find a parking space after starting assisted parking, establish a three-dimensional panoramic model of the parking environment, extract the parking space characteristics of the parking environment, and determine the spatial area coordinates on the ground in the parking environment that can accommodate the smart car body (target parking space characteristics) based on the parking space characteristics. In addition, different driving control parameters have different parking requirements. For example, a vehicle with a smaller turning radius is more flexible than a vehicle with a larger turning radius, so the requirement for adjustment space is smaller. The smart car model is simulated in the three-dimensional panoramic model to park in a simulated parking space based on the driving control parameters and the smart car body parameters, and scratch information during the simulation process is obtained. When there is a simulated parking situation without scratching the obstacle model, it means that it can be parked safely, and the grid coordinates corresponding to the target parking space characteristics are used as the target position, otherwise the output is that there is no suitable parking space, which improves the suitability and rationality of intelligent parking space search.

[0118] In one embodiment, according to the driving control parameters and the body parameters of the smart car, simulating the smart car model to park in a simulated parking space in the three-dimensional panoramic model includes:

[0119] Get the simulated parking type; the simulated parking types include: long side parking and short side parking;

[0120] According to the simulated parking type, historical parking records are obtained; wherein the historical parking records are parking video records of the simulated parking type collected by big data;

[0121] Extract the relative relationship feature sequence between the parking space line and the vehicle body according to the historical parking records; the relative relationship feature sequence is: different parking process and the corresponding parking space line-vehicle relationship, for example: {the vehicle body is 45 degrees with the short side of the parking space, the left rear wheel presses the short side of the parking space and is 30cm away from the left corner of the parking space, the right rear wheel presses the short side of the parking space, and the left and right sides of the vehicle body are 30cm away from the long side of the parking space...};

[0122] Determine, according to the driving control parameters and the simulated parking space line, a first adjustment cost of adjusting the simulated intelligent vehicle model to the relative relationship feature; wherein the first adjustment cost represents the difficulty of adjusting the relative position relationship between the simulated intelligent vehicle model and the simulated parking space line to the relative position relationship corresponding to the relative relationship feature;

[0123] Determine a second adjustment cost of the relative relationship feature corresponding to the historical vehicle posture; wherein the second adjustment cost is: the difficulty of adjustment of the historical vehicle corresponding to the extracted relative relationship feature;

[0124] Determine the first target relative relationship feature whose first adjustment cost is less than the second adjustment cost, and use the first target relative relationship feature that is relatively late in the relative relationship feature sequence as the second target relative relationship feature; wherein, the later the first target relative relationship feature is, the smaller the corresponding subsequent adjustment computing power required;

[0125] The second target relative relationship feature and the relative relationship feature after the second target relative relationship feature in the relative relationship feature sequence are used as a simulated intelligent vehicle model relative feature sequence;

[0126] According to the relationship feature sequence of the simulated smart car model, the smart car model is simulated to park in the simulated parking space in the three-dimensional panoramic model.

[0127] The working principle and beneficial effects of the above technical solution are:

[0128] The present invention introduces a simulated parking type, retrieves historical parking records corresponding to the simulated parking type (for example, parking records of reversing into a garage), and extracts a relative relationship feature sequence of the relationship between the parking space line and the vehicle body according to the historical parking records. The sequence includes different garage entry processes and their corresponding parking space line-vehicle relationships. When the corresponding garage entry process is reached, the corresponding relative relationship feature deviation is less than the deviation threshold, which can ensure that the vehicle can be garaged in a standardized manner when the final garage entry process is reached; however, each vehicle driving control parameter is different, such as the wheel drive form and the turning radius of the car. Therefore, it is not necessary to strictly follow the relative relationship feature sequence. If the vehicle simulation model can meet a certain relative relationship feature in the relative relationship feature sequence, then the relative relationship feature of the predecessor of the relative relationship feature that meets the requirement in the relative relationship feature sequence does not need to be fully followed, for example, when a car reverses into a garage When parking, first make the car body parallel to the storage position, then turn right to form a 45° angle with the storage line, and then reverse so that the left rear wheel presses the storage line, and the pressure point is 30cm-50cm away from the upper left corner of the storage. However, this is the storage logic of historical vehicles. Four-wheel independent drive vehicles can control the wheel steering individually to make the body and the storage line form a 45° angle, without the need for tedious pre-operation. Determine the first adjustment cost of the simulated intelligent car model to adjust to the relative relationship feature, and at the same time, determine the second adjustment cost of the relative relationship feature corresponding to the historical vehicle posture, select the second target relative relationship feature whose first adjustment cost is less than the second adjustment cost and whose sequence position is relatively late in the relative relationship feature sequence, select the second target relative relationship feature and the relative relationship feature thereafter as the simulated intelligent car model relationship feature sequence and perform subsequent simulated parking, thereby improving the efficiency of simulated parking.

[0129] In one embodiment, according to the driving control parameters and the body parameters of the smart car, simulating the smart car model to park in the simulated parking space in the three-dimensional panoramic model also includes:

[0130] Obtaining a target adjustment cost of the target vehicle corresponding to the relative relationship feature;

[0131] Wherein, obtaining the target adjustment cost of the target vehicle corresponding to the relative relationship feature includes:

[0132] Obtaining a manual driving model of a target vehicle; wherein the target vehicle is: a vehicle for which adjustment cost calculation is performed; and the manual driving model is: an automatic driving model trained according to driving control parameters of the target vehicle and manual driving records;

[0133] Determine the vehicle control parameters according to the relative relationship characteristics and the manual driving model; wherein the vehicle control parameters are: control parameters for driving the target vehicle so that its relationship with the parking space line conforms to the relative relationship characteristics;

[0134] The steering wheel steering increment and the travel increment are determined according to the vehicle control parameters; wherein the steering wheel steering increment is: the change in steering wheel rotation, for example: first turn right 90°, then turn left 45°, the corresponding steering wheel steering increment is 135°; the travel increment is: the change in mileage, for example: first shift to D gear and drive for 10m, then shift to R gear and drive for 5m, the corresponding travel increment is: 15m;

[0135] The steering wheel increment and the travel increment are normalized to obtain the target adjustment cost. The normalization rule is preset manually to satisfy the relationship that the larger the steering wheel increment and the travel increment, the higher the corresponding target adjustment cost.

[0136] The working principle and beneficial effects of the above technical solution are:

[0137] The present invention introduces an artificial driving model of a target vehicle (trained according to the driving control parameters of the target vehicle and artificial driving records), determines the vehicle control parameters according to the relative relationship characteristics and the artificial driving model, analyzes the vehicle control parameters to obtain the steering wheel increment and the travel increment and calculates the target adjustment cost, thereby improving the accuracy of the target adjustment cost calculation and further improving the accuracy of constructing the subsequent simulated intelligent vehicle model relationship feature sequence.

[0138] The embodiment of the present invention provides an automatic parking system for complex environments, such as Figure 2 As shown, including:

[0139] A grid map building subsystem 1 is used to build a grid map of the parking scene and determine the starting position and the target position;

[0140] A search direction determination subsystem 2, used for determining a search direction based on a dynamic weighted heuristic function;

[0141] The child node expansion subsystem 3 is used to set the starting position to the parent node and obtain the expanded child node according to the search direction;

[0142] The target node determination subsystem 4 is used to calculate the estimated cost and actual cost of the extended subnode and determine the target node;

[0143] The global optimal trajectory planning subsystem 5 is used to smoothly plan the trajectory according to the target node to obtain the global optimal trajectory;

[0144] The parking subsystem 6 is used to update the global optimal trajectory in real time and control the smart car to park based on the real-time global optimal trajectory.

[0145] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An automatic parking method for complex environments, characterized in that: include: Create a grid map of the parking site and determine the starting and target locations; Determine the search direction based on the dynamic weighted heuristic function; Set the starting position to the parent node and get the extended child nodes according to the search direction; Calculate the estimated cost and actual cost of expanding the child node and determine the target node; According to the smooth planning trajectory of the target node, the global optimal trajectory is obtained; The global optimal trajectory is updated in real time, and the smart car is controlled to park based on the real-time global optimal trajectory.

2. The automatic parking method for complex environments as claimed in claim 1, characterized in that: Create a grid map of the parking site and determine the starting and target locations, including: Obtain parking environment information; A grid map is created based on the parking environment information to determine the starting position and target position.

3. The automatic parking method for complex environments as claimed in claim 1, characterized in that: Calculate the estimated cost and actual cost of expanding the child node and determine the target node, including: According to the estimated cost of the expanded child node, the redundant nodes with the same estimated cost in the adjacent trajectories are eliminated, and then the target node with the minimum estimated cost in the adjacent trajectories is determined.

4. The automatic parking method for complex environments as claimed in claim 1, characterized in that: According to the target node smooth planning trajectory, the global optimal trajectory is obtained, including: Determine the introduction field according to the target node; Determine the repulsive field based on the non-search direction and the redundant nodes; Use the gravitational field and repulsive field to smoothly plan the trajectory and obtain the global optimal trajectory.

5. The automatic parking method for complex environments as claimed in claim 2, characterized in that: A grid map is created based on the parking environment information to determine the starting position and target position, including: Try to obtain the parking space selection signal of the smart car driver; If the parking space selection signal is successfully acquired, the parking space selection signal is parsed to determine the target location; If the parking space selection signal acquisition attempt fails, the target location will be searched intelligently.

6. The automatic parking method for complex environments as claimed in claim 5, characterized in that: Try to obtain the parking space selection signal of the smart car driver, including: Displaying a three-dimensional panoramic image of the parking environment to the driver of the smart car; Acquire a local three-dimensional panoramic image selected by a driver of the smart car; Extracting image features of the local 3D panoramic image, the image features including: traffic parameters and risk parameters of the actual area corresponding to the local 3D panoramic image; If the image features meet the standard image features, the local three-dimensional panoramic image is rasterized to determine the grid coordinates, and the parking space selection signal is determined according to the grid coordinates; If the image features do not meet the standard image features, it is determined that the attempt to obtain the parking space selection signal has failed.

7. The automatic parking method for complex environments as claimed in claim 5, characterized in that: Intelligently search for target locations, including: Based on the grid map, try to find the standard parking space line area; If the attempt to find is successful, the standard parking space line area is used as the target location; If the search attempt fails, the parking space characteristics are obtained based on the three-dimensional panoramic model of the parking environment; Determine the target parking space characteristics according to the parking space characteristics and the body parameters of the smart car; Determine the simulated parking space according to the characteristics of the target parking space; Obtain the driving control parameters of the smart car; According to the driving control parameters and the body parameters of the smart car, the smart car model is simulated to park in the simulated parking space in the 3D panoramic model; During the simulated parking process, determine whether there is a simulated parking situation in which no scratches occur with the obstacle model; If so, the grid coordinates corresponding to the target parking space features are used as the target position.

8. The automatic parking method for complex environments as claimed in claim 7, characterized in that: According to the driving control parameters and the body parameters of the smart car, the smart car model is simulated to park in the simulated parking space in the 3D panoramic model, including: Get the simulated parking type; Obtain historical parking records based on simulated parking types; According to the historical parking records, extract the relative relationship feature sequence between the parking space line and the vehicle body; Determine, based on the driving control parameters and the simulated parking space line, a first adjustment cost of adjusting the simulated intelligent vehicle model to a relative relationship feature; Determine a second adjustment cost of the relative relationship feature corresponding to the historical vehicle posture; Determine a first target relative relationship feature whose first adjustment cost is less than a second adjustment cost, and use the first target relative relationship feature which is relatively late in the relative relationship feature sequence as the second target relative relationship feature; The second target relative relationship feature and the relative relationship feature after the second target relative relationship feature in the relative relationship feature sequence are used as a simulated intelligent vehicle model relative feature sequence; According to the relationship feature sequence of the simulated smart car model, the smart car model is simulated to park in the simulated parking space in the three-dimensional panoramic model.

9. The automatic parking method for complex environments as claimed in claim 8, further comprising: Obtaining a target adjustment cost of the target vehicle corresponding to the relative relationship feature; Wherein, obtaining the target adjustment cost of the target vehicle corresponding to the relative relationship feature includes: Obtaining a manual driving model of a target vehicle; Determine vehicle control parameters based on relative relationship characteristics and artificial driving model; Determine a steering wheel steering increment and a travel increment according to vehicle control parameters; The steering wheel steering increment and travel increment are normalized to obtain the target adjustment cost.

10. An automatic parking system for complex environments, characterized in that: include: A grid map building subsystem is used to build a grid map of the parking site and determine the starting position and the target position; A search direction determination subsystem, used to determine the search direction based on a dynamic weighted heuristic function; The child node expansion subsystem is used to set the starting position to the parent node and obtain the expanded child nodes according to the search direction; The target node determination subsystem is used to calculate the estimated cost and actual cost of the extended subnode and determine the target node; The global optimal trajectory planning subsystem is used to smoothly plan the trajectory according to the target node to obtain the global optimal trajectory; The parking subsystem is used to update the global optimal trajectory in real time and control the smart car to park based on the real-time global optimal trajectory.

Images