Intelligent vehicle path prediction system for automatic driving

Abstract

The invention provides an intelligent vehicle path prediction system for automatic driving, and the system comprises a blind spot detection module which is used for obtaining blind spot region information based on environment data collected by a sensor of an automatic driving intelligent vehicle; the judgment module is used for judging whether the current driving path of the vehicle passes through the blind spot area or not, and if yes, an avoidance path is planned; the avoidance path planning module is used for searching an avoidance path scheme for avoiding the blind spot area in the vehicle driving area based on the blind spot area information; the path optimization module is used for optimizing the avoiding path scheme to obtain an optimal path; and the path prediction module is used for obtaining a local motion track and speed planning of the vehicle based on the optimal path, and completing intelligent vehicle path prediction. The problems of blind spot avoidance and path prediction planning of the automatic driving vehicle in a complex environment are effectively solved, the driving safety and the energy efficiency are improved, and reliable technical support is provided for application of an automatic driving technology in an actual road environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and in particular relates to an intelligent vehicle path prediction system for autonomous driving. Background Art

[0002] The core technical challenge faced by autonomous vehicles in complex road environments is how to effectively identify and respond to blind spots and predict paths to ensure driving safety. The existence of blind spots makes it difficult for vehicles to fully perceive the surrounding environment, which can easily lead to collision risks. However, simply avoiding all potential blind spots may cause the vehicle's driving path to be too conservative, affecting driving efficiency. Therefore, the key issue is how to weigh the costs and benefits of path adjustment while ensuring safety. This involves a series of complex processes such as accurate identification of blind spots, real-time matching with driving paths, generation and evaluation of alternative paths. At the same time, energy efficiency needs to be considered to avoid unnecessary energy consumption caused by excessive adjustments. In a dynamic and changing road environment, how to achieve rapid response and smooth avoidance of blind spots is also a major challenge. In addition, different types of blind spots (such as static obstacles, dynamic vehicles, pedestrians, etc.) may require different avoidance strategies. How to design a general and robust avoidance and path prediction method is also a problem that needs to be solved. In general, the blind spot avoidance problem of autonomous vehicles involves multiple levels such as perception, decision-making, and control, and it is necessary to seek the best balance between safety, efficiency, and comfort. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides an intelligent vehicle path prediction system for autonomous driving, which can realize the detection of blind spot areas and the efficient prediction of driving paths.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] An intelligent vehicle path prediction system for autonomous driving, comprising: a blind spot detection module, a judgment module, an avoidance path planning module, a path optimization module and a path prediction module;

[0006] The blind spot detection module is used to obtain blind spot area information based on environmental data collected by sensors of the autonomous driving intelligent vehicle;

[0007] The judgment module is used to match the blind spot area information with the current position and driving path information of the vehicle, and judge whether the current driving path of the vehicle will pass through the blind spot area. If so, plan an avoidance path;

[0008] The avoidance path planning module is used to search for an avoidance path solution that avoids the blind spot area within the vehicle drivable area based on the blind spot area information, combined with the road topology and traffic rules, and using a path search algorithm;

[0009] The path optimization module is used to optimize the avoidance path solution to obtain the optimal path;

[0010] The path prediction module is used to obtain the vehicle's local motion trajectory and speed planning based on the optimal path, combined with the vehicle dynamics model and real-time traffic status information, to complete intelligent vehicle path prediction.

[0011] Preferably, the blind spot detection module includes:

[0012] A data acquisition unit, used to collect surrounding environment data based on vehicle-mounted sensors to obtain obstacle information and road information;

[0013] A data format conversion unit, used to convert the obstacle information and road information into a digital image format, and pre-process the converted digital image;

[0014] A blind spot feature extraction unit, used to segment the pre-processed digital image to obtain a blind spot area feature vector;

[0015] An obstacle detection unit is used to classify the feature vector of the blind spot area based on a support vector machine, determine whether there is an obstacle in the blind spot area, and obtain the position information of the obstacle in the blind spot area;

[0016] The blind spot area information acquisition unit is used to calculate the relative position and distance of the obstacle in the blind spot area based on the position information of the obstacle and the current driving state information of the vehicle to obtain the blind spot area information.

[0017] Preferably, the judgment module includes:

[0018] A first coordinate acquisition unit, used to convert the current position information and driving path information of the vehicle into first coordinate data in a unified coordinate system;

[0019] A second coordinate acquisition unit, used to extract the position coordinate data of the blind spot area information to obtain second coordinate data;

[0020] A coordinate matching unit, used to match the first coordinate data with the second coordinate data to determine whether each position point on the vehicle's driving path falls within a blind spot area;

[0021] The marking unit marks the position coordinate data of the overlapping area as a target area to be avoided if the judgment result indicates that the vehicle's driving path overlaps with the blind spot area.

[0022] Preferably, the avoidance path planning module includes:

[0023] A path reference line acquisition unit, used to search for a global optimal path from a starting point to a destination based on a high-precision map, and use the global optimal path as a path reference line;

[0024] A reference line smoothing unit, used for smoothing the path reference line to obtain a smoothed path reference line;

[0025] An initial path acquisition unit is used to establish a Frenet coordinate system based on the smooth path reference line, perform raster sampling on the current road environment, and search for an initial path using a search algorithm;

[0026] The path space acquisition unit is used to generate a drivable path space avoiding the blind spot area based on the initial path and the road environment; and obtain an avoidance path plan based on the drivable path space.

[0027] Preferably, the path optimization module:

[0028] A collision detection unit, used to determine whether there is a collision risk in the avoidance path scheme based on a collision detection algorithm, and select a safe avoidance path;

[0029] The adjustment cost calculation unit is used to compare the safe avoidance path with the original driving path of the vehicle to obtain a path adjustment cost. When the path adjustment cost meets a preset threshold, the safe avoidance path is used as the optimal path.

[0030] Preferably, the path optimization module includes:

[0031] A traffic prediction unit is used to collect dynamic traffic information around the vehicle in real time based on vehicle-mounted sensors, and to fuse the dynamic traffic information with historical traffic data to obtain a trend of traffic condition changes;

[0032] A waypoint optimization unit, used to discretize the optimal path into waypoints, and perform constraint optimization on each waypoint using a vehicle dynamics model and road curvature and slope;

[0033] A motion state prediction unit, used to predict the motion state of the vehicle based on the optimized waypoints as reference tracks;

[0034] The path and speed prediction unit is used to plan the vehicle's motion trajectory and speed in a short period of time based on the vehicle's motion state prediction results.

[0035] Preferably, the waypoints include vehicle position, speed and acceleration.

[0036] Preferably, the process of constructing the vehicle dynamics model is:

[0037] Based on the actual position vector and the projection vector of the vehicle, a vehicle velocity vector and a projection velocity vector are obtained;

[0038] Calculating the lateral velocity of the vehicle in the Frenet coordinate system based on the vehicle velocity vector and the projected velocity vector;

[0039] Obtaining an error set based on the lateral velocity, the vehicle yaw angle, and the center of mass sideslip angle;

[0040] Based on the error set, a state quantity and a control quantity are obtained;

[0041] The vehicle dynamics model is obtained based on the state quantity and the control quantity.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows: environmental data is collected by sensors, blind spot areas are identified using image processing algorithms, and matched with the vehicle's driving path. If there is an overlap, a path search algorithm is used to generate an avoidance path, and safety assessment and energy consumption optimization are performed to select the optimal path. The present invention also takes into account the path adjustment cost and executes a new path when the threshold conditions are met. Finally, motion trajectory and speed planning are performed to achieve vehicle avoidance control. It effectively solves the blind spot avoidance and path prediction planning problems of autonomous driving vehicles in complex environments, improves driving safety and energy efficiency, and provides reliable technical support for the application of autonomous driving technology in actual road environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. 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 labor.

[0044] Figure 1 Schematic diagram of the structure of an intelligent vehicle path prediction system for autonomous driving according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Embodiment 1

[0048] like Figure 1 As shown, an intelligent vehicle path prediction system for autonomous driving includes: a blind spot detection module, a judgment module, an avoidance path planning module, a path optimization module and a path prediction module;

[0049] A blind spot detection module is used to obtain blind spot area information based on environmental data collected by sensors of the autonomous driving intelligent vehicle; a further implementation method is that the blind spot detection module includes:

[0050] A data acquisition unit, used to collect surrounding environment data based on vehicle-mounted sensors to obtain obstacle information and road information;

[0051] A data format conversion unit, used for converting obstacle information and road information into a digital image format and preprocessing the converted digital image;

[0052] A blind spot feature extraction unit is used to segment the pre-processed digital image to obtain a blind spot area feature vector;

[0053] The obstacle detection unit is used to classify the feature vector of the blind spot area based on the support vector machine, determine whether there is an obstacle in the blind spot area, and obtain the position information of the obstacle in the blind spot area;

[0054] The blind spot area information acquisition unit is used to calculate the relative position and distance of the obstacle in the blind spot area based on the position information of the obstacle and the current driving state information of the vehicle to obtain the blind spot area information.

[0055] In this embodiment, the autonomous driving intelligent vehicle collects environmental data around the vehicle through various sensors such as laser radar, millimeter wave radar, and camera, and converts the collected point cloud data, radar data, etc. into an RGB image format with a resolution of 1920×1080. The converted image is subjected to median filtering to remove salt and pepper noise, and the image contrast is enhanced by the histogram equalization algorithm to improve the image quality. The image is divided into 12 semantic categories such as roads, vehicles, pedestrians, and buildings using the SegNet-based semantic segmentation algorithm, and the semantic mask of the blind spot area is extracted. The color, texture, shape and other features of the blind spot area are extracted using algorithms such as color histogram, GLCM texture features, and Hu invariant moments to construct a 36-dimensional blind spot area feature vector. The support vector machine algorithm is used to perform binary classification on the blind spot area feature vector to determine whether there is an obstacle in the blind spot area, and the classification accuracy reaches more than 95%. According to the pixel coordinates of the obstacles in the blind spot area, combined with the vehicle's speed, direction, IMU and other driving status information, the relative position and distance of the obstacles are calculated using the triangulation principle to determine the range of the blind spot area.

[0056] A judgment module is used to match the blind spot area information with the current position and driving path information of the vehicle, and judge whether the current driving path of the vehicle will pass through the blind spot area. If so, an avoidance path is planned; a further implementation method is that the judgment module includes:

[0057] A first coordinate acquisition unit, used to convert the current position information and driving path information of the vehicle into first coordinate data in a unified coordinate system;

[0058] A second coordinate acquisition unit, used to extract the position coordinate data of the blind spot area information to obtain second coordinate data;

[0059] A coordinate matching unit, used to match the first coordinate data with the second coordinate data to determine whether each position point on the vehicle's driving path falls within a blind spot area;

[0060] The marking unit marks the position coordinate data of the overlapping area as a target area to be avoided if the judgment result indicates that the vehicle's driving path overlaps with the blind spot area.

[0061] In this embodiment, when obtaining the current position information of the vehicle and the preset driving path information, the vehicle-mounted GPS positioning system can be used to obtain the latitude and longitude coordinates of the vehicle (such as 122.34 degrees east longitude and 35.67 degrees north latitude), and convert them into coordinate points in a unified plane rectangular coordinate system (such as X=123.5 meters, Y=56.79 meters) to form a first coordinate data set of the vehicle's driving path. At the same time, using the position information of the road blind spot area identified in the previous step, the coordinates of the center point of the blind spot area are extracted (such as X=23.46 meters, Y=67.80 meters) to obtain a second coordinate data set. The first coordinate data set is spatially matched with the second coordinate data set, and the nearest neighbor algorithm is used to calculate the Euclidean distance from each coordinate point on the vehicle's driving path to the center point of the blind spot area. If the distance is less than the radius of the blind spot area (such as 50 meters), it is determined that the coordinate point falls within the blind spot area and needs to be avoided. Cluster analysis is performed on the coordinate points that are determined to need to be avoided to obtain several target areas that need to be avoided.

[0062] The avoidance path planning module is used to search for an avoidance path solution that avoids the blind spot area in the vehicle drivable area based on the blind spot area information, combined with the road topology and traffic rules, using a path search algorithm; a further implementation method is that the avoidance path planning module includes:

[0063] A path reference line acquisition unit is used to search for a global optimal path from a starting point to a destination based on a high-precision map, and use the global optimal path as a path reference line;

[0064] The reference line smoothing unit is used to smooth the path reference line to obtain a smooth path reference line; in this embodiment, the smoothing optimization problem is solved from three aspects, including optimization variables, optimization objectives, and constraints. Among them, the optimization variables are a series of discrete points to be smoothed, and the optimization objective is a smoothness cost function constructed based on the geometric relationship between the discrete points, including smoothness cost, geometric similarity cost, and length cost. The constraint condition is the numerical range between the difference and sum of the original path coordinates and the reference point relative to the original path point offset threshold.

[0065] The initial path acquisition unit is used to establish a Frenet coordinate system based on the smooth path reference line, perform raster sampling on the current road environment, and search for an initial path using a search algorithm;

[0066] The path space acquisition unit is used to generate a drivable path space that avoids the blind spot area based on the initial path and the road environment; and obtain an avoidance path plan based on the drivable path space.

[0067] A path optimization module is used to optimize the avoidance path solution to obtain an optimal path; a further implementation method is that the path optimization module:

[0068] The collision detection unit is used to determine whether there is a collision risk in the avoidance path scheme based on the collision detection algorithm, and to screen out a safe avoidance path; in this embodiment, a time-based dynamic collision detection (TDCD) algorithm is used, which can predict the relative position and speed of the vehicle and other obstacles in the next few seconds, so as to evaluate the probability of collision. Assume that on a certain path, the TDCD algorithm predicts that the minimum distance between the vehicle and the pedestrian is 5 meters, which is far below the safety threshold of 2 meters, so the path is marked as high risk. Then, a safety margin analysis is performed on all paths marked as low risk, such as calculating the average safety distance on each path and the shortest stopping distance of the vehicle in an emergency. With these data, safer paths can be further screened out. After determining the safe and feasible paths, these paths are further sorted, considering factors including path length and avoidance efficiency.

[0069] The adjustment cost calculation unit is used to compare the safe avoidance path with the original driving path of the vehicle to obtain the path adjustment cost. When the path adjustment cost meets the preset threshold, the safe avoidance path is used as the optimal path.

[0070] In this embodiment, the path data of the recommended avoidance path and the original planned path of the vehicle are obtained, including the starting point, end point, waypoint, section distance, section speed limit and other information of the path. The dynamic programming algorithm is used to calculate the driving time of the recommended path and the original planned path. By considering factors such as section distance and speed limit, the estimated driving time of the two paths is obtained. According to the power system parameters and section information of the vehicle, the energy consumption model is used to calculate the estimated energy consumption of the recommended path and the original planned path. The energy consumption values ​​of the two paths are determined by comprehensively considering factors such as the slope, speed, and acceleration of the section. The driving time of the recommended path is compared with the driving time of the original planned path to calculate the incremental driving time. At the same time, the energy consumption of the recommended path is compared with the energy consumption of the original planned path to obtain the incremental energy consumption. The incremental driving time and the incremental energy consumption are weighted and summed to obtain the comprehensive adjustment cost. By setting the weight coefficients of time and energy consumption, the two incremental values ​​are normalized, and the weighted sum is calculated as the comprehensive adjustment cost. It is judged whether the comprehensive adjustment cost exceeds the preset cost threshold. If the threshold is exceeded, the original planned path is maintained unchanged; if the threshold is not exceeded, the recommended path is used as the new driving path.

[0071] The path prediction module is used to obtain the vehicle's local motion trajectory and speed planning based on the optimal path, combined with the vehicle dynamics model and real-time traffic status information, to complete intelligent vehicle path prediction.

[0072] A further implementation is that the path optimization module includes:

[0073] Traffic prediction unit, which is used to collect dynamic traffic information around the vehicle in real time based on vehicle-mounted sensors, and integrate dynamic traffic information with historical traffic data to obtain traffic condition change trends;

[0074] The waypoint optimization unit is used to discretize the optimal path into waypoints and perform constraint optimization on each waypoint using the vehicle dynamics model and the road curvature and slope.

[0075] A motion state prediction unit, used to predict the motion state of the vehicle based on the optimized waypoints as reference tracks;

[0076] The path and speed prediction unit is used to plan the vehicle's motion trajectory and speed in a short period of time based on the vehicle's motion state prediction results.

[0077] A further embodiment is that the waypoint includes vehicle position, speed and acceleration.

[0078] A further implementation method is that the process of constructing the vehicle dynamics model is:

[0079] Based on the actual position vector and the projection vector of the vehicle, a vehicle velocity vector and a projection velocity vector are obtained;

[0080] Based on the vehicle velocity vector and the projected velocity vector, calculate the lateral velocity of the vehicle in the Frenet coordinate system;

[0081] Based on the lateral velocity, the vehicle yaw angle and the sideslip angle of the center of mass, an error set is obtained; in this embodiment, the error set includes a lateral error, a heading error, a position error and a velocity error.

[0082] Based on the error set, the state quantity and the control quantity are obtained; the state quantity is the transpose of the matrix composed of the error set, and the control quantity is the transpose of the matrix composed of the vehicle acceleration change and the front wheel steering angle.

[0083] Based on the state quantity and control quantity, the vehicle dynamics model is obtained.

[0084] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An intelligent vehicle path prediction system for autonomous driving, characterized in that: include: Blind spot detection module, judgment module, avoidance path planning module, path optimization module and path prediction module; The blind spot detection module is used to obtain blind spot area information based on environmental data collected by sensors of the autonomous driving intelligent vehicle; The judgment module is used to match the blind spot area information with the current position and driving path information of the vehicle, and judge whether the current driving path of the vehicle will pass through the blind spot area. If so, plan an avoidance path; The avoidance path planning module is used to search for an avoidance path solution that avoids the blind spot area within the vehicle drivable area based on the blind spot area information, combined with the road topology and traffic rules, and using a path search algorithm; The path optimization module is used to optimize the avoidance path solution to obtain the optimal path; The path prediction module is used to obtain the vehicle's local motion trajectory and speed planning based on the optimal path, combined with the vehicle dynamics model and real-time traffic status information, to complete intelligent vehicle path prediction.

2. The intelligent vehicle path prediction system for autonomous driving according to claim 1, characterized in that: The blind spot detection module comprises: A data acquisition unit, used to collect surrounding environment data based on vehicle-mounted sensors to obtain obstacle information and road information; A data format conversion unit, used to convert the obstacle information and road information into a digital image format, and pre-process the converted digital image; A blind spot feature extraction unit, used to segment the pre-processed digital image to obtain a blind spot area feature vector; An obstacle detection unit is used to classify the feature vector of the blind spot area based on a support vector machine, determine whether there is an obstacle in the blind spot area, and obtain the position information of the obstacle in the blind spot area; The blind spot area information acquisition unit is used to calculate the relative position and distance of the obstacle in the blind spot area based on the position information of the obstacle and the current driving state information of the vehicle to obtain the blind spot area information.

3. The intelligent vehicle path prediction system for autonomous driving according to claim 1, characterized in that: The judging module comprises: A first coordinate acquisition unit, used to convert the current position information and driving path information of the vehicle into first coordinate data in a unified coordinate system; A second coordinate acquisition unit, used to extract the position coordinate data of the blind spot area information to obtain second coordinate data; A coordinate matching unit, used to match the first coordinate data with the second coordinate data to determine whether each position point on the vehicle's driving path falls within a blind spot area; The marking unit marks the position coordinate data of the overlapping area as a target area to be avoided if the judgment result indicates that the vehicle's driving path overlaps with the blind spot area.

4. The intelligent vehicle path prediction system for autonomous driving according to claim 1, characterized in that: The avoidance path planning module includes: A path reference line acquisition unit, used to search for a global optimal path from a starting point to a destination based on a high-precision map, and use the global optimal path as a path reference line; A reference line smoothing unit, used for smoothing the path reference line to obtain a smoothed path reference line; An initial path acquisition unit is used to establish a Frenet coordinate system based on the smooth path reference line, perform raster sampling on the current road environment, and search for an initial path using a search algorithm; The path space acquisition unit is used to generate a drivable path space avoiding the blind spot area based on the initial path and the road environment; and obtain an avoidance path plan based on the drivable path space.

5. The intelligent vehicle path prediction system for autonomous driving according to claim 1, characterized in that: The path optimization module: A collision detection unit, used to determine whether there is a collision risk in the avoidance path scheme based on a collision detection algorithm, and select a safe avoidance path; The adjustment cost calculation unit is used to compare the safe avoidance path with the original driving path of the vehicle to obtain a path adjustment cost. When the path adjustment cost meets a preset threshold, the safe avoidance path is used as the optimal path.

6. The intelligent vehicle path prediction system for autonomous driving according to claim 1, characterized in that: The path optimization module includes: A traffic prediction unit is used to collect dynamic traffic information around the vehicle in real time based on vehicle-mounted sensors, and to fuse the dynamic traffic information with historical traffic data to obtain a trend of traffic condition changes; A waypoint optimization unit, used to discretize the optimal path into waypoints, and perform constraint optimization on each waypoint using a vehicle dynamics model and road curvature and slope; A motion state prediction unit, used to predict the motion state of the vehicle based on the optimized waypoints as reference tracks; The path and speed prediction unit is used to plan the vehicle's motion trajectory and speed in a short period of time based on the vehicle's motion state prediction results.

7. The intelligent vehicle path prediction system for autonomous driving according to claim 6, characterized in that: The waypoints include vehicle position, speed, and acceleration.

8. The intelligent vehicle path prediction system for autonomous driving according to claim 6, characterized in that: The process of constructing the vehicle dynamics model is: Based on the actual position vector and the projection vector of the vehicle, a vehicle velocity vector and a projection velocity vector are obtained; Calculating the lateral velocity of the vehicle in the Frenet coordinate system based on the vehicle velocity vector and the projected velocity vector; Obtaining an error set based on the lateral velocity, the vehicle yaw angle, and the center of mass sideslip angle; Based on the error set, a state quantity and a control quantity are obtained; The vehicle dynamics model is obtained based on the state quantity and the control quantity.

Images