Vehicle trajectory planning method based on other vehicle behavior prediction

Abstract

The invention relates to the field of automatic driving path planning, and discloses a vehicle trajectory planning method based on other vehicle behavior prediction, which is used for improving the passing efficiency and safety of an automatic driving vehicle. The vehicle trajectory planning method based on other vehicle behavior prediction comprises the steps of obtaining real-time state data of other vehicles in a current environment and line network data provided by a map; based on the real-time state data, the line network data and a preset positioning proximity principle, matching the current position of the other vehicle with a nearest neighbor path segment in the line network; according to the nearest neighbor path segment and the historical state data of the other vehicle, predicting a multi-modal trajectory of the other vehicle in a future preset time period, and generating a prediction intention set of the other vehicle; according to the prediction intention set, the reference line of the vehicle is adjusted to a feasible area with the lowest probability of conflict with the prediction path of the other vehicle in real time through a dynamic adsorption algorithm, so that a safety reference trajectory is generated; and controlling the own vehicle to run based on the safe reference trajectory.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving path planning technology, and in particular to a vehicle trajectory planning method based on prediction of other vehicle behavior. Background Technology

[0002] With the rapid development of intelligent transportation systems and autonomous driving technology, vehicle trajectory planning has become one of the core components for achieving safe and efficient autonomous driving.

[0003] Existing vehicle trajectory planning methods are mostly based on static environmental perception and fixed rule-based decision-making, such as planning local paths solely based on the instantaneous position and speed information of obstacles around the vehicle. The main technical drawback of such methods is that they treat other vehicles as dynamic obstacles rather than intelligent agents with intentions, making it impossible to effectively predict their future behavior. This results in the planning system only being able to passively react to conflicts at the present moment, lacking proactive avoidance capabilities. Static planning that relies solely on instantaneous states can easily lead to the vehicle generating overly conservative or aggressive trajectories, and may even result in path conflicts due to the inability to predict the intentions of other vehicles, severely restricting the safety and traffic efficiency of autonomous vehicles in real-world road environments. Summary of the Invention

[0004] This invention provides a vehicle trajectory planning method, apparatus, device, and storage medium based on prediction of other vehicle behavior, in order to solve the problem that the safety and traffic efficiency of autonomous vehicles in real road environments are low due to static planning that relies on instantaneous states in the prior art.

[0005] The first aspect of this invention provides a vehicle trajectory planning method based on the prediction of other vehicle behavior, comprising: acquiring real-time status data of other vehicles in the current environment and route network data provided by a map; matching the current location of the other vehicle with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and a preset location proximity principle; predicting the multimodal trajectory of the other vehicle within a preset time period in the future based on the nearest neighbor path segment and historical status data, thereby generating a set of predicted intentions for the other vehicle; adjusting the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of the other vehicle using a dynamic adsorption algorithm based on the set of predicted intentions, so as to generate a safe reference trajectory; and controlling the vehicle to drive based on the safe reference trajectory.

[0006] In one feasible implementation, the step of matching the current location of another vehicle with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and a preset location proximity principle includes: determining a search area based on the coordinates and heading angle of the other vehicle in the real-time status data; extracting all candidate path segments located within the search area based on the route network data; calculating the vertical distance from the current location of the other vehicle to the center line of each candidate path segment, matching the path segment with the smallest vertical distance as the nearest neighbor path segment, and vertically projecting the current location of the other vehicle onto the center line of the corresponding nearest neighbor path segment to determine the matched position of the other vehicle on the map.

[0007] In one feasible implementation, the step of predicting the multimodal trajectory of another vehicle within a future preset time period based on its nearest neighbor path segments and historical state data, and generating a set of predicted intentions for the other vehicle, includes: analyzing the driving characteristics of the other vehicle within the past preset time period based on its historical state data, the driving characteristics including speed change patterns, acceleration changes, and turning frequency and angle range; extracting relevant information of the corresponding nearest neighbor path segments from the route network data based on the other vehicle's nearest neighbor path segments, and generating network topology information based on the relevant information, the relevant information including the curvature, length, and connection relationship of the path segments with other path segments; constructing multiple possible driving scenario models based on the driving characteristics and the network topology information; and for each driving scenario model, using a preset trajectory prediction algorithm to generate multiple candidate trajectories for the other vehicle within the future preset time period, thereby obtaining a set of predicted intentions for the other vehicle.

[0008] In one feasible implementation, the step of constructing multiple possible driving scenario models based on the driving characteristics and the network topology information includes: determining all possible next path segments that other vehicles may enter based on the network topology information, as candidate intention targets; for each candidate intention target, calculating the probability weight of other vehicles selecting each candidate intention target by combining the speed change pattern and turning frequency in the driving characteristics; based on the probability weight, filtering out candidate intention targets with probabilities higher than a preset threshold, and constructing corresponding multiple possible driving scenario models based on the filtered candidate intention targets.

[0009] In one feasible implementation, based on the predicted intent set, a dynamic adsorption algorithm is used to adjust the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted paths of other vehicles, thereby generating a safe reference trajectory. This includes: extracting the spatial distribution of all predicted trajectories of other vehicles within a preset future time period based on the predicted intent set, and constructing a spatiotemporal conflict probability map; generating an initial reference line based on the vehicle's current position and a preset driving target, and calculating the lateral offset of each point on the initial reference line using a dynamic adsorption algorithm based on the spatiotemporal conflict probability map; and adjusting the initial reference line in real time to the feasible area with the lowest probability of conflict based on the lateral offset, thereby generating a safe reference trajectory.

[0010] In one feasible implementation, the step of extracting the spatial distribution of all predicted trajectories of other vehicles within a future preset time period based on the predicted intent set and constructing a spatiotemporal conflict probability map includes: discretizing the future preset time period into a series of time steps; associating each predicted trajectory in the predicted intent set with a probability weight representing its likelihood of occurrence; for each time step, calculating the union of the spatial regions occupied by all predicted trajectories of other vehicles in the corresponding time step to form a static occupation grid for the corresponding time step; weighting the static occupation grid for each time step based on the probability weight associated with each predicted trajectory to generate a probability grid map for the corresponding time step; and superimposing the probability grid maps of all time steps in chronological order to construct a spatiotemporal conflict probability map.

[0011] In one feasible implementation, the step of calculating the lateral offset of each point on the initial reference line using a dynamic adsorption algorithm based on the spatiotemporal conflict probability map includes: sampling at equal intervals along the initial reference line to obtain a series of path points; for each path point, querying the target probability value of its corresponding spatiotemporal position in the spatiotemporal conflict probability map, and using the target probability value as the magnitude of the repulsive force experienced by the corresponding path point; based on the magnitude of the repulsive force, and combined with vehicle dynamics constraints and road boundary constraints, solving for a set of lateral offsets that minimize the overall conflict probability using an optimization algorithm.

[0012] A second aspect of the present invention provides a vehicle trajectory planning device based on prediction of other vehicle behavior, comprising: an acquisition module for acquiring real-time status data of other vehicles in the current environment and route network data provided by a map; a matching module for matching the current location of other vehicles with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and a preset location proximity principle; a prediction module for predicting the multimodal trajectory of other vehicles within a preset time period based on the nearest neighbor path segment and historical status data of other vehicles, generating a set of predicted intentions for other vehicles; an adjustment module for adjusting the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles through a dynamic adsorption algorithm, based on the set of predicted intentions, to generate a safe reference trajectory; and a control module for controlling the vehicle to drive based on the safe reference trajectory.

[0013] In one feasible implementation, the matching module is specifically used to: determine the search area based on the coordinates and heading angle of other vehicles in the real-time status data; extract all candidate path segments located within the search area based on the route network data; calculate the vertical distance from the current position of other vehicles to the center line of each candidate path segment, match the path segment with the smallest vertical distance as the nearest neighbor path segment, and vertically project the current position of other vehicles onto the center line of the corresponding nearest neighbor path segment to determine the matched position of other vehicles on the map.

[0014] In one feasible implementation, the prediction module includes: an analysis unit, used to analyze the driving characteristics of another vehicle within a preset time period based on its historical state data, the driving characteristics including speed change patterns, acceleration changes, and turning frequency and angle range; an extraction unit, used to extract relevant information of the corresponding nearest neighbor path segments from the route network data based on the nearest neighbor path segments of the other vehicle, and generate network topology information based on the relevant information, the relevant information including the curvature, length, and connection relationship of the path segments with other path segments; a first construction unit, used to construct multiple possible driving scenario models based on the driving characteristics and the network topology information; and a generation unit, used to generate multiple candidate trajectories corresponding to the other vehicle within a preset time period in the future for each driving scenario model using a preset trajectory prediction algorithm, thereby obtaining a set of predicted intentions for the other vehicle.

[0015] In one feasible implementation, the first construction unit is specifically used to: determine all possible next path segments that other vehicles may enter based on the network topology information, as candidate intention targets; for each candidate intention target, combine the speed change pattern and turning frequency in the driving characteristics, and calculate the probability weight of other vehicles selecting each candidate intention target; based on the probability weight, filter out candidate intention targets with a probability higher than a preset threshold, and construct corresponding multiple possible driving scenario models based on the filtered candidate intention targets.

[0016] In one feasible implementation, the adjustment module includes: a second construction unit, configured to extract the spatial distribution of all predicted trajectories of other vehicles within a future preset time period based on the predicted intent set, and construct a spatiotemporal conflict probability map; a processing unit, configured to generate an initial reference line based on the current position of the vehicle and a preset driving target, and calculate the lateral offset of each point on the initial reference line using a dynamic adsorption algorithm based on the spatiotemporal conflict probability map; and an adjustment unit, configured to adjust the initial reference line in real time to the feasible area with the lowest conflict probability based on the lateral offset, thereby generating a safe reference trajectory.

[0017] In one feasible implementation, the second construction unit is specifically used to: discretize a future preset time period into a series of time steps; associate a probability weight representing the likelihood of occurrence for each predicted trajectory in the predicted intent set; for each time step, calculate the union of the spatial regions occupied by all other vehicle predicted trajectories in the corresponding time step to form a static occupancy grid for the corresponding time step; based on the probability weight associated with each predicted trajectory, weight the static occupancy grid for each time step to generate a probability grid map for the corresponding time step; and superimpose the probability grid maps of all time steps in chronological order to construct a spatiotemporal conflict probability map. In one feasible implementation, the processing unit is specifically used to: perform equally spaced sampling along the initial reference line to obtain a series of path points; for each path point, query the target probability value of its corresponding spatiotemporal location in the spatiotemporal conflict probability map, and use the target probability value as the magnitude of the repulsive force on the corresponding path point; based on the magnitude of the repulsive force, and in combination with vehicle dynamics constraints and road boundary constraints, solve a set of lateral offsets that minimize the overall conflict probability through an optimization algorithm.

[0018] A third aspect of the present invention provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to execute the above-described vehicle trajectory planning method based on prediction of other vehicle behavior.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described vehicle trajectory planning method based on prediction of other vehicle behavior.

[0020] The technical solution provided by this invention involves acquiring real-time status data of other vehicles in the current environment and route network data provided by a map; matching the current location of other vehicles with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and a preset proximity principle; predicting the multimodal trajectory of other vehicles within a preset time period based on the nearest neighbor path segment and historical status data, generating a set of predicted intentions for other vehicles; adjusting the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles using a dynamic adsorption algorithm, thereby generating a safe reference trajectory; and controlling the vehicle to drive based on the safe reference trajectory. In this embodiment of the invention, by predicting the behavioral intentions of other vehicles in advance through multimodal trajectory prediction, the vehicle can make proactive speed and path adjustments before potential conflicts occur, rather than reacting passively afterward, thus avoiding emergency braking. At the same time, the dynamic adsorption algorithm quantifies the conflict probability of each predicted path, enabling the vehicle to identify areas that are safe to pass through, significantly improving the efficiency and smoothness of path planning while ensuring safety, thereby improving the safety and traffic efficiency of the vehicle in real road environments. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of an embodiment of the vehicle trajectory planning method based on the prediction of other vehicle behavior in this invention. Figure 2 This is a schematic diagram of another embodiment of the vehicle trajectory planning method based on the prediction of other vehicle behavior in this invention. Figure 3 This is a schematic diagram of an embodiment of the vehicle trajectory planning device based on the prediction of other vehicle behavior in this invention. Figure 4 This is a schematic diagram of another embodiment of the vehicle trajectory planning device based on the prediction of other vehicle behavior in this invention. Figure 5 This is a schematic diagram of one embodiment of the electronic device in this invention. Detailed Implementation

[0022] This invention provides a vehicle trajectory planning method based on the prediction of other vehicle behavior. By analyzing the multimodal behavioral intentions of other vehicles and dynamically optimizing the trajectory of the vehicle in real time to avoid conflicts, the method significantly improves the traffic safety and efficiency of autonomous driving.

[0023] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] It is understood that the executing entity of this invention can be a vehicle trajectory planning device based on the prediction of other vehicle behavior, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0025] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the vehicle trajectory planning method based on the prediction of other vehicle behavior in this invention includes: 101. Obtain real-time status data of other vehicles in the current environment and route network data provided by the map; Utilizing various sensors onboard the vehicle, such as cameras and radar, the system performs real-time perception and data fusion of the surrounding environment to obtain real-time status data of other vehicles. This data primarily includes information such as the vehicle's position coordinates, speed, acceleration, and heading angle. Simultaneously, the system extracts structured route network data from high-precision maps or navigation systems. This route network data includes lane geometry, lane connections, and lane attributes.

[0026] 102. Based on real-time status data, route network data, and preset location proximity principles, match the current location of other vehicles with the nearest neighbor path segment in the route network; The real-time coordinates and heading angles of other vehicles are extracted from real-time status data. A reasonable search area is determined based on these coordinates and heading angles, and all candidate path segments within this area are selected from the route network data. Based on the principle of proximity, the vertical projection distance from the current position of the other vehicle to the centerline of each candidate path segment is calculated. Specifically, the centerline of each candidate path segment is modeled as a polyline composed of a series of continuous line segments. Each line segment is uniquely determined by the coordinates of its two adjacent path points. Each line segment of the polyline is traversed, and using the principle of vector projection, the vector formed by the current position of the other vehicle and the starting point of the line segment is projected onto the direction vector defined by the starting and ending points of the line segment. The projection scale coefficient is calculated to determine whether the projection point falls within the effective range of the line segment. If it falls within the effective range, the Euclidean distance from the projection point to the position of the other vehicle is directly calculated. If it falls outside the effective range, the Euclidean distances from the position of the other vehicle to the two endpoints of the line segment are calculated separately. Finally, the minimum value among the distances calculated from all candidate line segments is selected as the vertical projection distance from the other vehicle to the candidate path segment. This process is repeated by traversing all candidate path segments to complete all distance calculations. By traversing and comparing all these distances, the path segment with the smallest vertical distance is identified as the nearest neighbor path segment for the other vehicle. The current position of the other vehicle is then vertically projected onto the centerline of this nearest neighbor path segment, and this projected point is identified as its matched position on the high-precision map. This process effectively eliminates lateral errors present in GPS positioning and sensor observations, accurately binding the other vehicle's state to a specific lane, thus providing a reliable logical starting point for its possible future driving paths.

[0027] 103. Based on the nearest neighbor path segments and historical state data of other vehicles, predict the multimodal trajectory of other vehicles within a preset time period in the future, and generate a set of predicted intentions for other vehicles. The historical state data of other vehicles comes from the continuous monitoring and tracking of the surrounding environment by the autonomous vehicle's perception system. By fusing data from multiple sensors such as cameras, millimeter-wave radar, and lidar, the autonomous vehicle can detect and identify the presence of other vehicles in real time. Using a multi-target tracking algorithm, an independent tracking trajectory is established for each other vehicle target. The autonomous vehicle records and updates the state information of each other vehicle at continuous time stamps, including its global coordinates, velocity vector, acceleration, heading angle, and other key dynamic parameters. These state points arranged in time sequence are stored as a historical state queue, thus forming historical state data for analyzing the behavior patterns of other vehicles and predicting future trajectories.

[0028] Based on the nearest neighbor path segments of other vehicles, the road topology of their location is analyzed to list all possible driving paths, such as going straight, turning left, or turning right at an intersection. Each path choice represents a potential driving intention. Historical state data of other vehicles is analyzed to extract their speed profile, acceleration trend, and lateral motion characteristics to assess the probability of executing each intention, i.e., probability weights. For each candidate intention, a corresponding trajectory generator is invoked. This generator combines vehicle dynamics constraints and lane geometry constraints to calculate several possible future trajectories. Finally, the trajectories corresponding to all candidate intentions and their evaluated probability weights are packaged together to form a comprehensive set of predicted intentions that comprehensively describes future uncertainties.

[0029] 104. Based on the set of predicted intentions, the reference line of the vehicle is adjusted in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles through a dynamic adsorption algorithm in order to generate a safe reference trajectory. Based on lane information provided by the map and real-time perception of static obstacles, the feasible corridor area of ​​the vehicle is determined. A spatiotemporal conflict probability map is constructed based on the predicted intent set. This map discretizes the future time domain and integrates the spatial distribution of the predicted trajectories of all other vehicles and their corresponding probabilities at each time step, forming a three-dimensional probability field describing the collision risk at each point in the future spatiotemporal space. The initial reference line of the vehicle is obtained as the optimization basis, and a series of path points are obtained by dense sampling along the initial reference line. For each path point, the risk value of its position in the conflict probability map at the corresponding time point in the future is calculated, and this risk value is mapped to a lateral repulsive force perpendicular to the reference line. These repulsive forces, together with the constraint force generated by the feasible area corridor boundary and the inherent tension to maintain the smoothness of the trajectory, constitute a force balance model of the optimization system. Through numerical optimization algorithms such as sequential quadratic programming, under the constraint that the lateral position of all path points does not exceed the feasible area corridor boundary, a set of lateral offsets that minimize the overall cost function is solved. This cost function integrates the sum of conflict probabilities, the curvature change of the trajectory, and the degree of deviation from the original reference line, thereby smoothly and dynamically adjusting the initial reference line to the safe area and outputting the final safe reference trajectory.

[0030] 105. The vehicle is controlled to move based on a safety reference trajectory.

[0031] The vehicle's control system compares the vehicle's current actual state, including parameters such as speed and yaw angle, with the expected theoretical state on the safety reference trajectory in real time, calculates the error between the two, and then uses these errors as inputs. Through internally optimized algorithms, the control system calculates the specific control commands that need to be executed to eliminate these errors. These commands mainly include longitudinal control of the throttle and brake, as well as steering wheel angle control. The control system outputs these calculation results to the vehicle's drive-by-wire actuators, thereby realizing active intervention in the throttle, brake, and steering, and ultimately driving the vehicle smoothly, accurately, and safely along the planned safety reference trajectory.

[0032] In this embodiment of the invention, by acquiring the real-time status of other vehicles and high-precision map data, matching their path segments according to the proximity principle, and predicting and generating a set of multimodal trajectory intentions of other vehicles based on historical data, a dynamic adsorption algorithm is used to adjust the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted paths of other vehicles, generating a smooth and safe reference trajectory. By predicting the behavioral intentions of other vehicles in advance through multimodal trajectory prediction, the vehicle can make proactive speed and path adjustments before potential conflicts occur, rather than reacting passively afterward, thus avoiding emergency braking. At the same time, the dynamic adsorption algorithm quantifies the conflict probability of each predicted path, enabling the vehicle to identify areas that are safe to pass through. While ensuring safety, it significantly improves the efficiency and smoothness of path planning, thereby improving the safety and traffic efficiency of the vehicle in real road environments.

[0033] Please see Figure 2 Another embodiment of the vehicle trajectory planning method based on the prediction of other vehicle behavior in this invention includes: 201. Obtain real-time status data of other vehicles in the current environment and route network data provided by the map; 202. Based on real-time status data, route network data, and preset location proximity principles, match the current location of other vehicles with the nearest neighbor path segment in the route network; Based on the coordinates and heading angle of other vehicles in the real-time status data, the search area is determined; based on the route network data, all candidate path segments located within the search area are extracted; the vertical distance from the current position of other vehicles to the center line of each candidate path segment is calculated, the path segment with the smallest vertical distance is matched as the nearest neighbor path segment, and the current position of other vehicles is vertically projected onto the center line of the corresponding nearest neighbor path segment to determine the matched position of other vehicles on the map.

[0034] Based on the real-time coordinates and heading angle information of the other vehicle, a search area is formed by extending a set distance forward and backward along the heading direction, centered on its current position, and extending to both sides to a certain extent due to the uncertainty of the heading angle. Based on the route network data provided by the map, all path segments that intersect with this search area are extracted as candidate path segments. Using a geometric projection algorithm, the vertical projection distance from the current position of the other vehicle to the center line of each candidate path segment is calculated sequentially. By comparing all the calculated distance values, the path segment with the smallest vertical distance is selected as the nearest neighbor path segment. Finally, the current position of the other vehicle is vertically projected onto the center line of the nearest neighbor path segment, and the coordinates of the projected point and the path segment information are used as the precise matching position of the other vehicle on the high-precision map.

[0035] 203. Based on the historical status data of other vehicles, analyze the driving characteristics of other vehicles in the past preset time period. The driving characteristics include speed change pattern, acceleration change, and steering frequency and angle range. By performing time series analysis on historical state data of other vehicles collected continuously over a preset time period, the speed sequence is extracted and the mean, variance, and trend are calculated to characterize the speed change pattern. Based on the acceleration data sequence, the intensity, frequency, and duration of acceleration and deceleration behavior are statistically analyzed. At the same time, the heading angle sequence is analyzed, and the turning frequency and angle range are quantified by calculating the number of heading angle changes and the average change amplitude per unit time, thereby constructing a multi-dimensional driving feature vector that reflects the driving behavior characteristics of other vehicles.

[0036] 204. Based on the nearest neighbor path segments of other vehicles, extract relevant information of the corresponding nearest neighbor path segments from the network data, and generate network topology information based on the relevant information. The relevant information includes the curvature, length and connection relationship of the path segments with other path segments. Based on the nearest neighbor path segments of other vehicles, relevant information is extracted from the road network data to construct network topology information. First, for the identified nearest neighbor path segments, detailed geometric parameters are retrieved from the road network data provided by the map. By parsing specific fields in the road network data, the curvature information of the path segment is obtained, reflecting the degree of curvature at various locations. Simultaneously, the length data of the path segment is extracted to determine the possible distance range that other vehicles can travel on that path segment. The connection relationships between the path segment and other path segments are determined. By traversing the topology description section of the road network data, all downstream path segments directly connected to the nearest neighbor path segment are identified, and the identifiers, connection methods, and specific location information of the connection points of these connected path segments are recorded. Based on this, and using the extracted information such as path segment curvature, length, and connection relationships, a network topology information model that accurately reflects the characteristics of the road network in the area is constructed.

[0037] 205. Based on driving characteristics and combined with network topology information, construct models of multiple possible driving scenarios; Based on the network topology information, all possible next path segments that other vehicles may enter are identified as candidate intention targets. For each candidate intention target, the speed change pattern and turning frequency in the driving characteristics are combined, and the probability weight of other vehicles selecting each candidate intention target is calculated. Based on the probability weight, candidate intention targets with a probability higher than a preset threshold are selected, and corresponding models of multiple possible driving scenarios are constructed based on the selected candidate intention targets.

[0038] Based on network topology information, a set of candidate intent targets is formed by traversing all directly connected downstream path segments of the nearest neighbor path segment. For each candidate path segment, the initial probability of selecting each path segment is calculated using a Bayesian probability model, taking into account the current speed value and typical turning frequency of other vehicles. Geometric attributes of the path segment, such as curvature matching degree and connection type priority, are introduced to correct the probability, and candidate targets with a corrected probability higher than a set threshold are retained. Based on the candidate intent targets with probability weights higher than the set threshold, typical driving scenario models with different topological connection relationships, such as going straight, turning left, and turning right, are constructed.

[0039] 206. For each driving scenario model, a preset trajectory prediction algorithm is used to generate multiple candidate trajectories for other vehicles within a preset time period in the future, thereby obtaining a set of predicted intentions for other vehicles. For each selected driving scenario model, a trajectory prediction algorithm based on constant acceleration and constant steering rate is adopted. According to the current motion state of other vehicles and the geometric characteristics of the corresponding path segments, multiple possible trajectories in the future time period are generated under the constraints of vehicle dynamics. By simulating different combinations of acceleration and steering rate through Monte Carlo sampling, a candidate trajectory set covering the possible motion space is generated, and finally a complete set of prediction intentions containing each trajectory and its probability of occurrence is formed.

[0040] 207. Based on the set of predicted intentions, extract the spatial distribution of all predicted trajectories of other vehicles within a preset time period in the future, and construct a spatiotemporal conflict probability map; The project discretizes a future time period into a series of time steps. For each predicted trajectory in the set of predicted intentions, it associates a probability weight representing the likelihood of its occurrence. For each time step, it calculates the union of the spatial regions occupied by all other vehicle predicted trajectories in the corresponding time step, forming a static occupation grid for that time step. Based on the probability weight associated with each predicted trajectory, it weights the static occupation grid for each time step to generate a probability grid map for that time step. Finally, it overlays the probability grid maps of all time steps in chronological order to construct a spatiotemporal conflict probability map.

[0041] The project discretizes a predetermined future time period into a continuous sequence of time steps at fixed time intervals, with each time step corresponding to a specific future moment. A normalized probability weight is assigned to each predicted trajectory in the prediction intent set, reflecting the relative likelihood of that trajectory occurring. For each time step, a high-resolution environmental raster map is constructed, mapping the spatial occupancy range of each other vehicle's predicted trajectory within the corresponding time step onto the raster. The static occupancy raster for the corresponding time step is obtained by calculating the union of the occupied areas of all trajectories, where each raster cell records the binary state of being covered by any trajectory. Based on this, a probability weighting mechanism is introduced, accumulating the probabilities of the raster cells occupied by each trajectory according to its weight value. This results in the value of each raster cell representing the overall probability of that spatial location being occupied at the corresponding time step, thus generating a probability raster map for that time step. Finally, these chronologically ordered probability raster maps are superimposed along the time dimension to construct a three-dimensional spatiotemporal conflict probability map. The value of each spatiotemporal cell characterizes the probability intensity of a collision at the corresponding future moment and spatial location, providing a complete risk field representation for subsequent trajectory planning. This map takes into account both spatial occupancy and temporal evolution characteristics, and can accurately reflect the changes in risk distribution under dynamic environments.

[0042] 208. Based on the vehicle's current position and preset driving target, generate an initial reference line, and use a dynamic adsorption algorithm to calculate the lateral offset of each point on the initial reference line based on the spatiotemporal conflict probability map. Equal-interval sampling is performed along the initial reference line to obtain a series of path points. For each path point, the target probability value of its corresponding spatiotemporal location in the spatiotemporal conflict probability map is queried, and the target probability value is used as the magnitude of the repulsive force on the corresponding path point. Based on the magnitude of the repulsive force, and combined with the vehicle dynamics constraints and road boundary constraints, an optimization algorithm is used to solve for a set of lateral offsets that minimize the overall conflict probability.

[0043] Dense sampling is performed along the initial reference line at fixed longitudinal intervals to obtain a set of discrete pathpoint sequences. Each pathpoint contains its spatial coordinates and corresponding timestamp information. For each pathpoint, a 3D interpolation query is performed on the spatiotemporal conflict probability map to obtain the collision probability value corresponding to the spatial location of the pathpoint at a specific timestamp. This collision probability value is directly converted into the magnitude of the repulsive force perpendicular to the reference line; the higher the probability value, the stronger the repulsive force. An optimization problem with multi-objective constraints is established, where the objective function comprehensively considers the sum of the repulsive forces experienced by all pathpoints, the rate of change of trajectory curvature, and the degree of deviation from the original reference line. Hard constraints are also added, including dynamic constraints such as the maximum lateral acceleration limit of the vehicle and the maximum rate of change of steering angle, as well as feasible corridor constraints formed by road boundaries. A sequential quadratic programming algorithm is used to iteratively solve this optimization problem. Through multiple linearizations and quadratic approximations, a set of lateral offset sequences that minimize the objective function is calculated. This sequence, while satisfying all constraints, keeps the adjusted trajectory as far away from high-conflict probability areas as possible, while ensuring the smoothness and feasibility of the trajectory. The final lateral offset sequence will be used to generate a safe reference trajectory.

[0044] 209. Based on the lateral offset, the initial reference line is adjusted in real time to the feasible area with the lowest probability of conflict, generating a safe reference trajectory; A set of lateral offset sequences obtained from the optimization solution is mapped one-to-one with the path points sampled on the initial reference line. The calculated lateral offset is applied to each path point to obtain a series of adjusted new path point coordinates. A spline curve fitting algorithm is used to smoothly interpolate these discrete new path points, generating a continuous and smooth trajectory curve. This trajectory curve not only strictly follows the obstacle avoidance direction indicated by the lateral offset, but also satisfies the curvature continuity constraint of vehicle kinematics. During the fitting process, the relative position of the adjusted trajectory and the road boundary is monitored in real time to ensure that the trajectory always lies within the feasible corridor.

[0045] 210. The vehicle is controlled to move based on a safety reference trajectory.

[0046] In this embodiment of the invention, by analyzing the historical driving characteristics of other vehicles and combining them with the network topology, a multi-dimensional driving scenario model is constructed, generating a multi-modal trajectory prediction and spatiotemporal conflict probability map of other vehicles. Based on this, a dynamic adsorption algorithm is used to perform fine-grained lateral offset adjustment on each point of the vehicle's reference line, guiding the vehicle's trajectory in real time to the feasible area with the lowest conflict probability. This significantly enhances driving safety in real road environments. At the same time, through more accurate scenario-based prediction and trajectory optimization, sudden braking and redundant detours are effectively reduced, greatly improving the vehicle's traffic efficiency.

[0047] The vehicle trajectory planning method based on other vehicle behavior prediction in the embodiments of the present invention has been described above. The vehicle trajectory planning device based on other vehicle behavior prediction in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 3 One embodiment of the vehicle trajectory planning device based on other vehicle behavior prediction in this invention includes: The acquisition module 301 is used to acquire real-time status data of other vehicles in the current environment and route network data provided by the map; The matching module 302 is used to match the current location of another vehicle with the nearest neighbor path segment in the route network based on real-time status data, route network data and preset location proximity principle; The prediction module 303 is used to predict the multimodal trajectory of another vehicle within a preset time period based on the nearest neighbor path segment and historical state data of the other vehicle, and generate a set of predicted intentions for the other vehicle. The adjustment module 304 is used to adjust the vehicle reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles based on the predicted intention set and through a dynamic adsorption algorithm, so as to generate a safe reference trajectory. The control module 305 is used to control the vehicle to drive based on a safety reference trajectory.

[0048] In this embodiment of the invention, by acquiring the real-time status of other vehicles and high-precision map data, matching their path segments according to the proximity principle, and predicting and generating a set of multimodal trajectory intentions of other vehicles based on historical data, a dynamic adsorption algorithm is used to adjust the vehicle's reference line in real time to the feasible area with the lowest probability of conflict with the predicted paths of other vehicles, generating a smooth and safe reference trajectory. By predicting the behavioral intentions of other vehicles in advance through multimodal trajectory prediction, the vehicle can make proactive speed and path adjustments before potential conflicts occur, rather than reacting passively afterward, thus avoiding emergency braking. At the same time, the dynamic adsorption algorithm quantifies the conflict probability of each predicted path, enabling the vehicle to identify areas that are safe to pass through. While ensuring safety, it significantly improves the efficiency and smoothness of path planning, thereby improving the safety and traffic efficiency of the vehicle in real road environments.

[0049] Please see Figure 4 Another embodiment of the vehicle trajectory planning device based on other vehicle behavior prediction in this invention includes: The acquisition module 301 is used to acquire real-time status data of other vehicles in the current environment and route network data provided by the map; The matching module 302 is used to match the current location of another vehicle with the nearest neighbor path segment in the route network based on real-time status data, route network data and preset location proximity principle; The prediction module 303 is used to predict the multimodal trajectory of another vehicle within a preset time period based on the nearest neighbor path segment and historical state data of the other vehicle, and generate a set of predicted intentions for the other vehicle. The adjustment module 304 is used to adjust the vehicle reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles based on the predicted intention set and through a dynamic adsorption algorithm, so as to generate a safe reference trajectory. The control module 305 is used to control the vehicle to drive based on a safety reference trajectory.

[0050] Optionally, the matching module 302 can be specifically used for: Based on the coordinates and heading angle of other vehicles in the real-time status data, the search area is determined; based on the route network data, all candidate path segments located within the search area are extracted; the vertical distance from the current position of other vehicles to the center line of each candidate path segment is calculated, the path segment with the smallest vertical distance is matched as the nearest neighbor path segment, and the current position of other vehicles is vertically projected onto the center line of the corresponding nearest neighbor path segment to determine the matched position of other vehicles on the map.

[0051] Optionally, the prediction module 303 includes: The analysis unit 3031 is used to analyze the driving characteristics of other vehicles in the past preset time period based on the historical status data of other vehicles. The driving characteristics include speed change pattern, acceleration change, and steering frequency and angle range. The extraction unit 3032 is used to extract relevant information of the corresponding nearest neighbor path segment from the road network data based on the nearest neighbor path segment of other vehicles, and generate network topology information based on the relevant information. The relevant information includes the curvature, length and connection relationship of the path segment with other path segments. The first building unit 3033 is used to build multiple possible driving scenario models based on driving characteristics and network topology information; The generation unit 3034 is used to generate multiple candidate trajectories for other vehicles within a preset time period in the future, based on a preset trajectory prediction algorithm for each driving scenario model, thereby obtaining a set of predicted intentions for other vehicles.

[0052] Optionally, the first building unit 3033 can be specifically used for: Based on the network topology information, all possible next path segments that other vehicles may enter are identified as candidate intention targets. For each candidate intention target, the speed change pattern and turning frequency in the driving characteristics are combined, and the probability weight of other vehicles selecting each candidate intention target is calculated. Based on the probability weight, candidate intention targets with a probability higher than a preset threshold are selected, and corresponding models of multiple possible driving scenarios are constructed based on the selected candidate intention targets.

[0053] Optionally, adjustment module 304 includes: The second building unit 3041 is used to extract the spatial distribution of all predicted trajectories of other vehicles within a preset time period in the future based on the set of predicted intentions, and to build a spatiotemporal conflict probability map. The processing unit 3042 is used to generate an initial reference line based on the current position of the vehicle and the preset driving target, and to calculate the lateral offset of each point on the initial reference line based on the spatiotemporal conflict probability map using a dynamic adsorption algorithm. The adjustment unit 3043 is used to adjust the initial reference line in real time to the feasible area with the lowest probability of conflict based on the lateral offset, thereby generating a safe reference trajectory.

[0054] Optionally, the second building unit 3041 is specifically used for: The project discretizes a future time period into a series of time steps. For each predicted trajectory in the set of predicted intentions, it associates a probability weight representing the likelihood of its occurrence. For each time step, it calculates the union of the spatial regions occupied by all other vehicle predicted trajectories in the corresponding time step, forming a static occupation grid for that time step. Based on the probability weight associated with each predicted trajectory, it weights the static occupation grid for each time step to generate a probability grid map for that time step. Finally, it overlays the probability grid maps of all time steps in chronological order to construct a spatiotemporal conflict probability map. Optionally, the processing unit 3042 is specifically used for: Equal-interval sampling is performed along the initial reference line to obtain a series of path points. For each path point, the target probability value of its corresponding spatiotemporal location in the spatiotemporal conflict probability map is queried, and the target probability value is used as the magnitude of the repulsive force on the corresponding path point. Based on the magnitude of the repulsive force, and combined with the vehicle dynamics constraints and road boundary constraints, an optimization algorithm is used to solve for a set of lateral offsets that minimize the overall conflict probability.

[0055] In this embodiment of the invention, by analyzing the historical driving characteristics of other vehicles and combining them with the network topology, a multi-dimensional driving scenario model is constructed, generating a multi-modal trajectory prediction and spatiotemporal conflict probability map of other vehicles. Based on this, a dynamic adsorption algorithm is used to perform fine-grained lateral offset adjustment on each point of the vehicle's reference line, guiding the vehicle's trajectory in real time to the feasible area with the lowest conflict probability. This significantly enhances driving safety in real road environments. At the same time, through more accurate scenario-based prediction and trajectory optimization, sudden braking and redundant detours are effectively reduced, greatly improving the vehicle's traffic efficiency.

[0056] above Figure 3 and Figure 4 The vehicle trajectory planning device based on other vehicle behavior prediction in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The electronic device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0057] See Figure 5 As shown, the electronic device includes a processor 500 and a memory 501. The memory 501 stores machine-executable instructions that can be executed by the processor 500. The processor 500 executes the machine-executable instructions to implement the above-described vehicle trajectory planning method based on the prediction of other vehicle behavior.

[0058] Furthermore, Figure 5 The electronic device shown also includes a bus 502 and a communication interface 503. The processor 500, the communication interface 503 and the memory 501 are connected via the bus 502.

[0059] The memory 501 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 502 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0060] The processor 500 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 500 or by instructions in software form. The processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 501. The processor 500 reads the information in memory 501 and, in conjunction with its hardware, completes the method steps of the aforementioned embodiment.

[0061] The present invention also provides an electronic device, the computer device including a memory and a processor, wherein the memory stores computer-readable instructions, and when executed by the processor, the processor causes the processor to perform the steps of the vehicle trajectory planning method based on other vehicle behavior prediction described in the above embodiments. The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the computer causes the computer to perform the steps of the vehicle trajectory planning method based on other vehicle behavior prediction.

[0062] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0064] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle trajectory planning method based on prediction of other vehicle behavior, characterized in that, The vehicle trajectory planning method based on the prediction of other vehicle behavior includes: Obtain real-time status data of other vehicles in the current environment and route network data provided by the map; Based on the real-time status data, the route network data, and the preset location proximity principle, the current location of other vehicles is matched with the nearest neighbor path segment in the route network; Based on the nearest neighbor path segments and historical state data of other vehicles, predict the multimodal trajectory of other vehicles within a preset time period in the future, and generate a set of predicted intentions for other vehicles. Based on the predicted intent set, the vehicle reference line is adjusted in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles through a dynamic adsorption algorithm, so as to generate a safe reference trajectory. The vehicle is controlled to move based on the aforementioned safety reference trajectory.

2. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 1, characterized in that, The process of matching the current location of another vehicle with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and the preset location proximity principle includes: The search area is determined based on the coordinates and heading angle of other vehicles in the real-time status data; Based on the network data, extract all candidate path segments located within the search area; Calculate the vertical distance from the current location of the other vehicle to the center line of each candidate path segment, match the path segment with the smallest vertical distance as the nearest neighbor path segment, and vertically project the current location of the other vehicle onto the center line of the corresponding nearest neighbor path segment to determine the matched position of the other vehicle on the map.

3. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 1, characterized in that, The step of predicting the multimodal trajectory of another vehicle within a preset time period based on its nearest neighbor path segments and historical state data, and generating a set of predicted intentions for that vehicle, includes: Based on the historical status data of other vehicles, analyze the driving characteristics of other vehicles in the past preset time period, including speed change patterns, acceleration changes, and steering frequency and angle range; Based on the nearest neighbor path segments of other vehicles, relevant information of the corresponding nearest neighbor path segments is extracted from the network data, and network topology information is generated based on the relevant information. The relevant information includes the curvature, length and connection relationship of the path segments with other path segments. Based on the driving characteristics and the network topology information, various possible driving scenario models are constructed. For each driving scenario model, a preset trajectory prediction algorithm is used to generate multiple candidate trajectories for other vehicles within a preset time period in the future, thus obtaining a set of predicted intentions for other vehicles.

4. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 3, characterized in that, Based on the driving characteristics and combined with the network topology information, a variety of possible driving scenario models are constructed, including: Based on the network topology information, determine all possible next path segments that other vehicles may enter, as candidate intended targets; For each candidate intention target, the speed change pattern and turning frequency in the driving characteristics are combined, and the probability weight of other vehicles selecting each candidate intention target is calculated. Based on the probability weights, candidate intent targets with probabilities higher than a preset threshold are selected, and corresponding models of various possible driving scenarios are constructed based on the selected candidate intent targets.

5. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 1, characterized in that, Based on the predicted intent set, a dynamic adsorption algorithm is used to adjust the vehicle's reference line in real time to the feasible area with the lowest probability of conflicting with the predicted paths of other vehicles, in order to generate a safe reference trajectory, including: Based on the predicted intent set, the spatial distribution of all predicted trajectories of other vehicles within a preset time period in the future is extracted, and a spatiotemporal conflict probability map is constructed. Based on the vehicle's current position and preset driving target, an initial reference line is generated, and based on the spatiotemporal conflict probability map, a dynamic adsorption algorithm is used to calculate the lateral offset of each point on the initial reference line. Based on the lateral offset, the initial reference line is adjusted in real time to the feasible area with the lowest probability of conflict, thereby generating a safe reference trajectory.

6. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 5, characterized in that, The step of extracting the spatial distribution of all predicted trajectories of other vehicles within a preset future time period based on the predicted intent set, and constructing a spatiotemporal conflict probability map, includes: Discretize the future preset time period into a series of time steps; For each predicted trajectory in the set of predicted intentions, a probability weight representing the likelihood of its occurrence is associated; For each time step, calculate the union of the spatial regions occupied by all other vehicles' predicted trajectories at the corresponding time step, forming a static occupancy grid for the corresponding time step; Based on the probability weight associated with each predicted trajectory, the static occupied grid at each time step is weighted to generate a probability grid map for the corresponding time step. By overlaying the probability raster charts of all time steps in chronological order, a spatiotemporal conflict probability map is constructed.

7. The vehicle trajectory planning method based on other vehicle behavior prediction according to claim 5, characterized in that, The calculation of the lateral offset of each point on the initial reference line based on the spatiotemporal conflict probability map using a dynamic adsorption algorithm includes: A series of path points are obtained by sampling at equal intervals along the initial reference line; For each path point, query the target probability value of its corresponding spatiotemporal location in the spatiotemporal conflict probability map, and use the target probability value as the magnitude of the repulsive force experienced by the corresponding path point; Based on the magnitude of the repulsive force, and in conjunction with vehicle dynamics constraints and road boundary constraints, an optimization algorithm is used to solve for a set of lateral offsets that minimize the overall probability of conflict.

8. A vehicle trajectory planning device based on prediction of other vehicle behavior, characterized in that, The vehicle trajectory planning device based on prediction of other vehicle behavior includes: The acquisition module is used to acquire real-time status data of other vehicles in the current environment and route network data provided by the map; The matching module is used to match the current location of another vehicle with the nearest neighbor path segment in the route network based on the real-time status data, the route network data, and the preset location proximity principle. The prediction module is used to predict the multimodal trajectory of another vehicle within a preset time period based on the nearest neighbor path segments and historical state data of the other vehicle, and generate a set of predicted intentions for the other vehicle. The adjustment module is used to adjust the vehicle reference line in real time to the feasible area with the lowest probability of conflict with the predicted path of other vehicles based on the predicted intention set and through a dynamic adsorption algorithm, so as to generate a safe reference trajectory. The control module is used to control the vehicle to drive based on the safety reference trajectory.

9. An electronic device, characterized in that, The electronic device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to execute the vehicle trajectory planning method based on prediction of other vehicle behavior as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the vehicle trajectory planning method based on the prediction of other vehicle behavior as described in any one of claims 1-7.

Images