Vehicle control method and system based on artificial intelligence

By considering the movement characteristics of the obstacle vehicle and the distance of the non-obstruction grid in the vehicle avoidance route planning, a scientific and reasonable escape route is generated, which solves the problem of inaccurate escape routes in the existing technology, and realizes the scientific, effective and safe vehicle avoid.

CN118494468BActive Publication Date: 2024-11-08ROPEOK TECHNOLOGY GROUP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410954226.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-08
Estimated Expiration
2044-07-17

AI Technical Summary

Technical Problem

The actual operating status of the obstacle vehicle is not considered in the planning of existing vehicle evacuation routes, resulting in inaccurate avoidance routes generated and causing safety hazards such as scratches and collisions.

Method used

Using an artificial intelligence-based vehicle control method, by obtaining the relative information between the barrier vehicle grid and the initial grid in the two-dimensional grid map, the vehicle avoidance coefficient and obstacle impact coefficient are calculated, and the initial vehicle recommended route is generated, and the initial vehicle is smoothed. Processing to obtain the optimal route for scientific, effective and safe avoidance.

Benefits of technology

By considering the movement characteristics of the obstacle vehicle and the distance of the non-obstruction grid, a scientific and reasonable avoidance path is generated, the risk of avoidance control is reduced, and the vehicle's scientific, effective and safe avoidance is achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118494468B_ABST
    Figure CN118494468B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of vehicle avoidance control technology, and in particular to a vehicle control method and system based on artificial intelligence. The method is based on the relative information between the obstacle vehicle grid and the initial grid in a two-dimensional grid map, and determines the vehicle avoidance coefficient of each obstacle vehicle grid according to the relative speed and distance information. The obstacle influence coefficient of the obstacle vehicle grid relative to the non-obstacle grid is obtained based on the motion characteristics. The vehicle avoidance estimation cost obtained according to the obstacle influence coefficient can simultaneously take into account the motion state of the obstacle vehicle and the distance of the non-obstacle grid relative to the target grid, thereby obtaining a scientific and reasonable initial vehicle recommended route. The initial vehicle recommended route is smoothed, and the target vehicle is controlled to avoid using the optimal vehicle recommended route. The present invention obtains a scientific and effective avoidance path by analyzing the motion characteristics of the obstacle vehicle, thereby achieving scientific, effective and safe avoidance of the target vehicle.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle avoidance control, and in particular to a vehicle control method and system based on artificial intelligence. Background Art

[0002] Automatic assisted navigation driving technology is widely used in the automotive industry, especially in new energy vehicles. On highways, because the types of objects on the road are relatively fixed, most of them are moving cars with obvious vehicle characteristics, driving simulation technology can be used on the road to plan excellent avoidance paths in the simulation map for avoidance.

[0003] In the prior art, the A* algorithm can be used to plan the path of the simulated grid map. When it is detected that avoidance is required, the vehicle is controlled to drive through the planned route to achieve avoidance. However, in the existing vehicle avoidance route planning, the length of the generated route is mainly used as a reference. On the actual road, the obstacle vehicle may accelerate and decelerate at any time, that is, the obstacle vehicle has an obvious unstable speed characteristic. If the actual running state of the obstacle vehicle is not considered during the path planning process, the generated avoidance route will be inaccurate, causing safety hazards such as scratches and collisions. Summary of the invention

[0004] In order to solve the technical problem that the existing technology does not consider the actual running state of the obstacle vehicle during the path planning process, resulting in inaccurate avoidance routes and causing safety hazards such as scratches and collisions, the purpose of the present invention is to provide a vehicle control method and system based on artificial intelligence. The technical solutions adopted are as follows:

[0005] The present invention proposes a vehicle control method based on artificial intelligence, the method comprising:

[0006] Acquire a two-dimensional grid map of the target vehicle on the road in real time; the two-dimensional grid map includes an obstacle vehicle grid representing an obstacle vehicle, an initial grid representing an initial position of the target vehicle, a target grid representing an obstacle avoidance end point of the target vehicle, and a non-obstacle grid;

[0007] According to the distance between the obstacle vehicle grid and the initial grid, and the relative speed between the obstacle vehicle and the target vehicle, the vehicle avoidance coefficient of each obstacle vehicle grid is obtained;

[0008] For any non-obstruction grid, the obstacle influence coefficient of the non-obstruction grid is obtained according to the distance between each obstacle vehicle grid and the non-obstruction grid and the corresponding vehicle avoidance coefficient; the obstacle influence coefficient is adjusted according to the distance between the non-obstruction grid and the target grid to obtain the estimated vehicle avoidance cost of each non-obstruction grid, and the initial vehicle recommended route of the target vehicle is obtained according to the estimated vehicle avoidance cost;

[0009] Smoothing the initial vehicle recommended route according to the distribution characteristics of the estimated vehicle avoidance cost within the neighborhood of each route grid in the initial vehicle recommended route to obtain the optimal vehicle recommended route;

[0010] The target vehicle is controlled to avoid the obstacle according to the optimal vehicle recommended route.

[0011] Furthermore, the method for obtaining the relative speed includes:

[0012] The two-dimensional grid map at the real time is compared with the two-dimensional grid map at the previous moment, and the relative speed is obtained according to the position change of the obstacle vehicle grid.

[0013] Further, obtaining the relative speed according to the change of the distance between the obstacle vehicle grid and the target grid includes:

[0014] If the same obstacle vehicle grid exists in two adjacent two-dimensional grid maps, the relative speed of the corresponding obstacle vehicle grid is obtained according to the position distance between the same obstacle vehicle grids; if the obstacle vehicle grid at the real time moment does not exist in the two-dimensional grid map at the previous moment, the relative speed of the corresponding obstacle vehicle grid at the real time moment is set to a preset speed value.

[0015] Furthermore, the method for obtaining the vehicle avoidance coefficient includes:

[0016] The relative speed is used as a numerator and the square of the distance between the obstacle vehicle grid and the initial grid is used as a denominator to obtain the vehicle avoidance coefficient.

[0017] Furthermore, the method for obtaining the obstacle influence coefficient includes:

[0018] For any non-obstruction grid, the distance influence coefficient between each obstacle vehicle grid and the non-obstruction grid is obtained according to the distance between the obstacle vehicle grid and the non-obstruction grid; the initial influence coefficient between each obstacle vehicle grid and the non-obstruction grid is obtained according to the distance influence coefficient and the vehicle avoidance coefficient of the obstacle vehicle grid; the initial influence coefficients between all obstacle vehicle grids and the non-obstruction grid are accumulated as the obstacle influence coefficient of the non-obstruction grid.

[0019] Furthermore, the method for obtaining the estimated cost of vehicle avoidance includes:

[0020] The distance between the non-obstruction grid and the target grid is multiplied by the obstacle influence coefficient to obtain the vehicle avoidance estimated cost.

[0021] Furthermore, the method for obtaining the optimal vehicle recommended route includes:

[0022] A smoothing window is constructed with the starting point of the initial vehicle recommended route as the center. In the smoothing window, a penalty factor of each route grid is obtained according to the fluctuation degree of the estimated cost of vehicle avoidance in the neighborhood of each route grid; an optimal route grid corresponding to the smoothing window is screened out in the route grid according to the variation characteristics of the penalty factor between adjacent route grids; the smoothing window is continuously used for processing with the optimal route grid as the center until the entire initial vehicle recommended route is traversed; and a route composed of the initial grid, the optimal route grid and the target grid is used as the optimal vehicle recommended route.

[0023] Furthermore, the method for obtaining the penalty factor includes:

[0024] The standard deviation of the estimated cost of vehicle avoidance within the route grid neighborhood is used as the penalty factor.

[0025] Further, the step of selecting the optimal route grid corresponding to the smoothing window from the route grid includes:

[0026] In the smoothing window, along the direction of the initial vehicle recommended route, the penalty factor difference between each route grid and the next route grid is used as a judgment value; the route grid with the judgment value less than 0 is used as a candidate grid;

[0027] If the grid to be selected exists in the smooth window, when the center point of the smooth window is the grid to be selected, the next route grid of the center point is used as the optimal route grid, otherwise the first grid to be selected in the smooth window is used as the optimal route grid; if the grid to be selected does not exist in the smooth window, the last route grid in the smooth window is used as the optimal route grid.

[0028] The present invention also proposes an artificial intelligence-based vehicle control system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the steps of an artificial intelligence-based vehicle control method when executing the computer program.

[0029] The present invention has the following beneficial effects:

[0030] The embodiment of the present invention first determines the vehicle avoidance coefficient of each obstacle vehicle grid according to the relative speed and distance information based on the relative information between the obstacle vehicle grid and the initial grid in the two-dimensional grid map. The vehicle avoidance coefficient can characterize the motion characteristics of the corresponding obstacle vehicle, so the obstacle influence coefficient of the obstacle vehicle grid relative to the non-obstacle grid can be obtained based on the motion characteristics. Because an avoidance path needs to be found in the non-obstacle grid, the vehicle avoidance estimated cost obtained according to the obstacle influence coefficient can simultaneously take into account the motion state of the obstacle vehicle and the distance of the non-obstacle grid relative to the target grid, thereby obtaining a scientific and reasonable initial vehicle recommended route. In order to further reduce the risk of the vehicle during avoidance control, the initial vehicle recommended route is further smoothed, and the optimal vehicle recommended route is used to control the target vehicle to avoid, so as to achieve scientific, effective and safe avoidance. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 A flow chart of a vehicle control method based on artificial intelligence provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the vehicle control method and system based on artificial intelligence proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0034] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0035] The specific scheme of a vehicle control method and system based on artificial intelligence provided by the present invention is described in detail below with reference to the accompanying drawings.

[0036] See also Figure 1 , which shows a flow chart of a vehicle control method based on artificial intelligence provided by an embodiment of the present invention, the method comprising:

[0037] Step S1: Acquire a two-dimensional grid map of the target vehicle on the road in real time.

[0038] The embodiment of the present invention aims at path planning of a car in the process of avoiding a maneuver on the road. The car should determine in real time whether it needs to avoid a maneuver during driving, and then plan a route to achieve avoidance. Therefore, the real-time moment proposed in the embodiment of the present invention is the moment when the avoidance operation is to be performed. Because it is necessary to determine in real time whether it is necessary to avoid a maneuver during driving, a two-dimensional grid map of the target vehicle on the road also exists at the historical moment before the real-time moment.

[0039] In a specific implementation of an embodiment of the present invention, obstacle point cloud data of the surrounding environment is collected by a 32-line laser radar installed on the target vehicle. The detection range of the radar is 45 meters wide and 395 meters long. The obstacle point cloud data collected by the target vehicle is constructed into a two-dimensional grid map with a width of 45 meters and a length of 395 meters using a SLAM algorithm, that is, each grid corresponds to a square area with an actual length of 1 meter. In the two-dimensional grid map, the target vehicle position is located at the center, that is, the central grid is the initial grid representing the initial position of the target vehicle, and the grid corresponding to the obstacle is marked as 1 in the two-dimensional grid map, and the non-obstacle grid is marked as 0. Because the road contains not only obstacle vehicles, but also other obstacles such as road fences, it is necessary to further identify the obstacle vehicle grid representing the obstacle vehicle in order to facilitate subsequent calculations. The embodiment of the present invention uses the obstacle point cloud data as input, uses the VoxelNet neural network to classify and identify the obstacle point cloud data, outputs the obstacle identification label of each obstacle point cloud data, and then screens out the obstacle vehicle grid corresponding to the obstacle vehicle. Each obstacle vehicle grid is marked according to the corresponding obstacle vehicle, and a semantic label is added to each obstacle vehicle grid, for example, it can be marked as obstacle vehicle a, obstacle vehicle b, etc., to facilitate subsequent processing of the obstacle vehicle grid.

[0040] In one embodiment of the present invention, considering that the target vehicle is about to perform an avoidance operation, it is best to continue to return to the original lane for driving after the avoidance. Therefore, the target grid is recorded as the last grid along the vehicle driving direction in the grid column where the initial grid is located in the two-dimensional grid map.

[0041] Step S2: Obtain the vehicle avoidance coefficient of each obstacle vehicle grid according to the distance between the obstacle vehicle grid and the initial grid, and the relative speed between the obstacle vehicle and the target vehicle.

[0042] During road driving, if the speed of the obstacle vehicle is significantly higher and the distance is closer than that of the target vehicle, it means that the relative position of the obstacle vehicle will change significantly in a short period of time, and it will threaten the safety of the target vehicle's avoidance driving. If the motion state of the obstacle vehicle is not considered during the avoidance path planning process, the risk of scratches and collisions will increase. Therefore, the embodiment of the present invention obtains the vehicle avoidance coefficient of each obstacle vehicle grid according to the distance between the obstacle vehicle grid and the initial grid, and the relative speed between the obstacle vehicle and the target vehicle. That is, the larger the vehicle avoidance coefficient, the more attention should be paid to the impact of the corresponding obstacle vehicle grid on the path during path planning, and the higher the degree of avoidance of the target vehicle to the obstacle vehicle should be.

[0043] Preferably, in one embodiment of the present invention, the method for acquiring the relative speed includes:

[0044] Because the two-dimensional grid map contains the position information of the obstacle vehicle, the two-dimensional grid map at the real time is compared with the two-dimensional grid map at the previous moment, and the relative speed can be obtained according to the position change of the obstacle vehicle grid.

[0045] Further, in one embodiment of the present invention, if the same obstacle vehicle grid exists in two adjacent two-dimensional grid maps, the relative speed of the corresponding obstacle vehicle grid is obtained according to the position distance between the same obstacle vehicle grids; if the obstacle vehicle grid at the real time moment does not exist in the two-dimensional grid map at the previous moment, it means that the obstacle vehicle has a fast speed, or has a more aggressive driving habit, and directly appears in the monitoring range of the target vehicle for a period of time, so the obstacle vehicle grid corresponding to the obstacle vehicle can be set to a larger value, that is, the relative speed of the corresponding obstacle vehicle grid at the real time moment is set to the preset speed value. In the embodiment of the present invention, the preset speed value is set to 30, that is, the speed of 30 is considered to be the largest value among all obstacle vehicle grids. It should be noted that the specific setting can be made according to the division of the image field of view. In the embodiment of the present invention, after the image field of view is divided, the horizontal and vertical coordinates are respectively divided into 40 units, that is, the maximum relative speed within one second is 40, so the preset speed value is set to 30, which can be a larger value. Similarly, the preset speed value can be specifically set according to the specific division result.

[0046] It should be noted that, because a semantic label is assigned to each obstacle vehicle grid in one embodiment of the present invention, it is convenient to obtain the relative speed. In one embodiment of the present invention, the previous moment of the real time is the moment one second before the real time, that is, a two-dimensional grid map can be obtained every second, so the position distance between obstacle vehicle grids with the same semantic label in the same coordinate system can be directly used as the relative speed of the corresponding obstacle vehicle grid.

[0047] It should be noted that, in the embodiment of the present invention, the distances between grids are obtained by using a Euclidean distance algorithm based on a two-dimensional grid map, which will not be described in detail later.

[0048] Preferably, in one embodiment of the present invention, the method for obtaining the vehicle avoidance coefficient includes:

[0049] The relative speed is used as the numerator and the square of the distance between the obstacle vehicle grid and the initial grid is used as the denominator to obtain the vehicle avoidance coefficient. The embodiment of the present invention uses the ratio to construct a correlation relationship, and further uses the square form to amplify the influence of the distance on the result, that is, the greater the relative speed and the smaller the distance, the greater the avoidance risk of the obstacle vehicle grid, and the greater the vehicle avoidance coefficient.

[0050] It should be noted that in other embodiments of the present invention, the ratio of the relative speed to the distance may be directly used as the vehicle avoidance coefficient, or other basic mathematical operations may be used to construct a correlation relationship, which is not limited or elaborated herein.

[0051] Step S3: For any non-obstacle grid, the obstacle influence coefficient of the non-obstacle grid is obtained according to the distance between each obstacle vehicle grid and the non-obstacle grid and the corresponding vehicle avoidance coefficient; the obstacle influence coefficient is adjusted according to the distance between the non-obstacle grid and the target grid to obtain the estimated vehicle avoidance cost of each non-obstacle grid, and the initial vehicle recommended route of the target vehicle is obtained according to the estimated vehicle avoidance cost.

[0052] In a two-dimensional grid map, path planning needs to find a reasonable avoidance path in the non-obstacle grid, but for an obstacle vehicle, its special driving state may threaten the safety of the non-obstacle grid. That is, for a non-obstacle grid, the closer an obstacle vehicle grid is to it and the larger the vehicle avoidance coefficient of the obstacle vehicle grid, the more likely it is that the obstacle vehicle will drive to the position of the non-obstacle grid, thereby threatening the avoidance driving safety of the target vehicle. Therefore, for any non-obstacle grid, the embodiment of the present invention obtains the obstacle influence coefficient of the non-obstacle grid according to the distance between each obstacle vehicle grid and the non-obstacle grid and the corresponding vehicle avoidance coefficient. The larger the obstacle influence coefficient, the more susceptible the non-obstacle grid is to the influence of the obstacle vehicle grid, and the less likely it should be planned into the path during the path planning process.

[0053] Preferably, in one embodiment of the present invention, the method for obtaining the obstacle influence coefficient includes:

[0054] For any non-obstruction grid, the distance influence coefficient between each obstacle vehicle grid and the non-obstruction grid is obtained according to the distance between the obstacle vehicle grid and the non-obstruction grid. That is, the smaller the distance, the greater the distance influence coefficient.

[0055] According to the distance influence coefficient and the vehicle avoidance coefficient of the obstacle vehicle grid, the initial influence coefficient between each obstacle vehicle grid and the non-obstacle grid is obtained. That is, the initial influence coefficient represents the influence of an obstacle vehicle grid on a non-obstacle grid. The larger the distance influence coefficient and the larger the vehicle avoidance coefficient, the larger the initial influence coefficient. Further, the initial influence coefficients between all obstacle vehicle grids and non-obstacle grids are accumulated as the obstacle influence coefficient of the non-obstacle grid. That is, the obstacle influence coefficient represents the influence of all obstacle vehicle grids on a non-obstacle grid.

[0056] In one embodiment of the present invention, the inverse of the distance between the obstacle vehicle grid and the non-obstruction grid is used as the power of an exponential function with a natural constant as the base, thereby realizing negative correlation mapping of the distance and normalizing it to obtain a distance influence coefficient. The product of the distance influence coefficient and the vehicle avoidance coefficient is used as the initial influence coefficient.

[0057] In other embodiments of the present invention, other basic mathematical operations may also be used to implement functions such as negative correlation mapping and positive correlation, which are not limited or elaborated herein.

[0058] Further considering the distance between the non-obstruction grid and the target grid, the obstacle influence coefficient is adjusted according to the distance between the non-obstruction grid and the target grid to obtain the estimated cost of vehicle avoidance for each non-obstruction grid. When the distance between the non-obstruction grid and the target grid is large, if the non-obstruction grid is included in the avoidance path, the avoidance path will be longer. Therefore, the estimated cost of vehicle avoidance should be large, and the non-obstruction grid should be avoided as a route grid in the avoidance path. At the same time, the larger the obstacle influence coefficient, the greater the impact of the non-obstruction grid on the obstacle vehicle grid, the more likely the non-obstruction grid will cause safety hazards, and the estimated cost of vehicle avoidance will be large, so the non-obstruction grid should be avoided as a route grid in the avoidance path.

[0059] Preferably, in one embodiment of the present invention, it is considered that the distance between the non-obstruction grid and the target grid and the obstacle influence coefficient should be positively correlated with the vehicle avoidance estimated cost, so the distance between the non-obstruction grid and the target grid is multiplied by the obstacle influence coefficient to obtain the vehicle avoidance estimated cost.

[0060] The estimated cost of vehicle avoidance can then be used as the estimated cost from each non-obstacle grid to the target grid in the A* algorithm, and the A* algorithm can be used to output the initial vehicle recommended route in the two-dimensional grid map. Driving on the initial vehicle recommended route can actively avoid obstacle vehicles that are close to the target vehicle, and can also avoid obstacle vehicles with relatively high relative speeds and more aggressive driving, ensuring driving safety.

[0061] It should be noted that the A* algorithm is a path planning algorithm well known to those skilled in the art and will not be described in detail here.

[0062] Step S4: Smoothing the initial vehicle recommended route according to the vehicle avoidance estimated cost distribution characteristics within the neighborhood of each route grid in the initial vehicle recommended route to obtain the optimal vehicle recommended route.

[0063] The initial vehicle recommended route obtained is a route planned taking into account the motion state of the obstacle vehicle. In the actual driving process of the vehicle, frequent turning or excessive turning angle of the vehicle will not only increase the difficulty of vehicle control, but also further increase the safety risk. Therefore, the initial vehicle recommended route needs to be smoothed to obtain the optimal vehicle recommended route with a simple path and uncomplicated vehicle control. Considering that the vehicle avoidance estimated cost can characterize the path availability of the non-obstruction grid, for the route grid in the initial vehicle recommended route, the route grid itself has a small vehicle avoidance estimated cost. If the vehicle avoidance estimated cost distribution within its neighborhood is relatively uniform, it means that the non-obstruction grid is relatively safe. In path planning, it is replaced by the surrounding grids, which has little effect on the safety of the path. Therefore, during the smoothing process, the vehicle avoidance estimated cost distribution characteristics within the neighborhood of each route grid can be smoothed to reduce the route complexity in the optimal vehicle recommended route.

[0064] Preferably, in one embodiment of the present invention, the optimal vehicle recommended route is obtained by adopting the idea of ​​step-by-step traversal, which specifically includes:

[0065] A smoothing window is constructed with the starting point of the initial vehicle recommended route as the center. In the smoothing window, the penalty factor of each route grid is obtained according to the fluctuation degree of the estimated cost of vehicle avoidance within the neighborhood of each route grid.

[0066] If the variation characteristics of the penalty factors between adjacent route grids show an obvious increasing trend, it means that the route grid at the back has a certain instability, and the route grid at the front is relatively more stable. Therefore, the optimal route grid corresponding to the smooth window can be screened out in the route grid according to the variation characteristics of the penalty factors between adjacent route grids. Continue to use the smooth window to process with the optimal route grid as the center until the entire initial vehicle recommended route is traversed; the route composed of the initial grid, the optimal route grid and the target grid is used as the optimal vehicle recommended route. It should be noted that the initial grid, the optimal route grid and the target grid may not be continuous. The path can be filled by an optimization algorithm such as an interpolation algorithm, or the grids can be directly connected to obtain a continuous optimal vehicle recommended route.

[0067] In one embodiment of the present invention, the neighborhood range is set to 8 neighborhoods, and the size of the smoothing window is set to 25×25.

[0068] Preferably, in one embodiment of the present invention, the standard deviation of the estimated cost of vehicle avoidance within the neighborhood of the route grid is used as a penalty factor. That is, the larger the standard deviation, the greater the fluctuation of the estimated cost of vehicle avoidance, the more unstable the route grid, and the larger the penalty factor.

[0069] In other embodiments of the present invention, statistical values ​​such as variance that represent fluctuation characteristics may also be selected as penalty factors, which will not be elaborated or limited here.

[0070] Preferably, in one embodiment of the present invention, selecting the optimal route grid corresponding to the smoothing window from the route grid includes:

[0071] In the smoothing window, along the direction of the initial vehicle recommended route, the difference in penalty factors between each route grid and the next route grid is used as the judgment value. If the judgment value is greater than 0, it means that the penalty factor is decreasing, and if the judgment value is less than 0, it means that the penalty is increasing. The increasing trend indicates that the route grid at the back has a certain instability, and the route grid at the front is relatively more stable. Therefore, the judgment value less than 0 can reflect the stability of the current route grid, so the route grid with a judgment value less than 0 is used as the candidate grid.

[0072] If there is a grid to be selected in the smooth window, the first grid to be selected in the smooth window is used as the optimal route grid. If there is no grid to be selected in the smooth window, the last route grid in the smooth window is used as the optimal route grid. It should be noted that the embodiment of the present invention searches for grids to be selected step by step from the center point of the sliding window along the direction of vehicle travel. If the center point of the sliding window is a grid to be selected, the next route grid of the center point of the sliding window is used as the optimal route grid.

[0073] Step S5: Control the target vehicle to avoid the vehicle according to the optimal vehicle recommended route.

[0074] The optimal vehicle recommended route is planned on the basis of the initial vehicle recommended route. It can actively avoid obstacle vehicles that are close to the target vehicle, and can also avoid obstacle vehicles with a relatively high relative speed and more aggressive driving. Furthermore, the optimal vehicle recommended route does not include routes with obvious turning angles that make vehicle control complicated. Therefore, controlling the target vehicle to avoid according to the optimal vehicle recommended route can achieve an effective and safe avoidance effect.

[0075] In summary, the embodiment of the present invention determines the vehicle avoidance coefficient of each obstacle vehicle grid according to the relative speed and distance information based on the relative information between the obstacle vehicle grid and the initial grid in the two-dimensional grid map. The obstacle influence coefficient of the obstacle vehicle grid relative to the non-obstacle grid is obtained based on the motion characteristics. The estimated vehicle avoidance cost obtained according to the obstacle influence coefficient can simultaneously take into account the motion state of the obstacle vehicle and the distance of the non-obstacle grid relative to the target grid, thereby obtaining a scientific and reasonable initial vehicle recommended route. The initial vehicle recommended route is smoothed, and the optimal vehicle recommended route is used to control the target vehicle to avoid. The present invention obtains a scientific and effective avoidance path by analyzing the motion characteristics of the obstacle vehicle, thereby achieving scientific, effective and safe avoidance of the target vehicle.

[0076] The present invention also proposes an artificial intelligence-based vehicle control system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any step of an artificial intelligence-based vehicle control method is implemented.

[0077] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A vehicle control method based on artificial intelligence, characterized in that: The method comprises: Acquire a two-dimensional grid map of the target vehicle on the road in real time; the two-dimensional grid map includes an obstacle vehicle grid representing an obstacle vehicle, an initial grid representing an initial position of the target vehicle, a target grid representing an obstacle avoidance end point of the target vehicle, and a non-obstacle grid; According to the distance between the obstacle vehicle grid and the initial grid, and the relative speed between the obstacle vehicle and the target vehicle, the vehicle avoidance coefficient of each obstacle vehicle grid is obtained; For any non-obstruction grid, the obstacle influence coefficient of the non-obstruction grid is obtained according to the distance between each obstacle vehicle grid and the non-obstruction grid and the corresponding vehicle avoidance coefficient; the obstacle influence coefficient is adjusted according to the distance between the non-obstruction grid and the target grid to obtain the estimated vehicle avoidance cost of each non-obstruction grid, and the initial vehicle recommended route of the target vehicle is obtained according to the estimated vehicle avoidance cost; Smoothing the initial vehicle recommended route according to the distribution characteristics of the estimated vehicle avoidance cost within the neighborhood of each route grid in the initial vehicle recommended route to obtain the optimal vehicle recommended route; Controlling the target vehicle to avoid the vehicle according to the optimal vehicle recommended route; The method for obtaining the optimal vehicle recommended route includes: A smoothing window is constructed with the starting point of the initial vehicle recommended route as the center. In the smoothing window, a penalty factor of each route grid is obtained according to the fluctuation degree of the estimated cost of vehicle avoidance in the neighborhood of each route grid; an optimal route grid corresponding to the smoothing window is screened out in the route grid according to the variation characteristics of the penalty factor between adjacent route grids; the smoothing window is continuously used for processing with the optimal route grid as the center until the entire initial vehicle recommended route is traversed; the route composed of the initial grid, the optimal route grid and the target grid is used as the optimal vehicle recommended route; The step of selecting the optimal route grid corresponding to the smoothing window from the route grid includes: In the smoothing window, along the direction of the initial vehicle recommended route, the penalty factor difference between each route grid and the next route grid is used as a judgment value; the route grid with the judgment value less than 0 is used as a candidate grid; If the grid to be selected exists in the smooth window, when the center point of the smooth window is the grid to be selected, the next route grid of the center point is used as the optimal route grid, otherwise the first grid to be selected in the smooth window is used as the optimal route grid; if the grid to be selected does not exist in the smooth window, the last route grid in the smooth window is used as the optimal route grid.

2. The vehicle control method based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the relative speed includes: The two-dimensional grid map at the real time is compared with the two-dimensional grid map at the previous moment, and the relative speed is obtained according to the position change of the obstacle vehicle grid.

3. The vehicle control method based on artificial intelligence according to claim 2, characterized in that: The obtaining the relative speed according to the change of the distance between the obstacle vehicle grid and the target grid includes: If the same obstacle vehicle grid exists in two adjacent two-dimensional grid maps, the relative speed of the corresponding obstacle vehicle grid is obtained according to the position distance between the same obstacle vehicle grids; if the obstacle vehicle grid at the real time moment does not exist in the two-dimensional grid map at the previous moment, the relative speed of the corresponding obstacle vehicle grid at the real time moment is set to a preset speed value.

4. The vehicle control method based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the vehicle avoidance coefficient includes: The relative speed is used as a numerator and the square of the distance between the obstacle vehicle grid and the initial grid is used as a denominator to obtain the vehicle avoidance coefficient.

5. The vehicle control method based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the obstacle influence coefficient includes: For any non-obstruction grid, the distance influence coefficient between each obstacle vehicle grid and the non-obstruction grid is obtained according to the distance between the obstacle vehicle grid and the non-obstruction grid; the initial influence coefficient between each obstacle vehicle grid and the non-obstruction grid is obtained according to the distance influence coefficient and the vehicle avoidance coefficient of the obstacle vehicle grid; the initial influence coefficients between all obstacle vehicle grids and the non-obstruction grid are accumulated as the obstacle influence coefficient of the non-obstruction grid.

6. The vehicle control method based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the estimated cost of vehicle avoidance includes: The distance between the non-obstruction grid and the target grid is multiplied by the obstacle influence coefficient to obtain the vehicle avoidance estimated cost.

7. The vehicle control method based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the penalty factor includes: The standard deviation of the estimated cost of vehicle avoidance within the route grid neighborhood is used as the penalty factor.

8. A vehicle control system based on artificial intelligence, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the vehicle control method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.