Path space-time corridor data construction method, electronic device, and storage medium

Abstract

Embodiments of the present application provide a path space-time corridor data construction method, an electronic device and a storage medium. The purpose is to solve the problem that the existing path constraint modeling is not complete in expressing the boundaries of static, dynamic and multi-type objects, and is not continuous in space-time organization, resulting in unstable planning constraints in a complex traffic scene. To this end, an initial space-time corridor is established based on the road boundary, the road curve coordinate system envelope box, the starting state of the ego vehicle and the vehicle geometric parameters, and a uniform time sampling sequence is determined. Then, the vehicle dynamic boundary data is generated in combination with the boundary information of static objects, dynamic objects and multi-type objects, and the object pre-decision information is fused to reconstruct the space-time constraints of the initial corridor, forming the target path space-time corridor data for autonomous driving. This scheme can improve the completeness, continuity and space-time consistency of path constraint modeling, and improve the smoothness, executability and safety of trajectory planning.

Description

Technical Field

[0001] This application relates to the field of autonomous vehicle path planning and vehicle automatic control, and in particular to a method for constructing path spatiotemporal corridor data, electronic devices, storage media, and program products. Background Technology

[0002] Autonomous driving path planning typically combines road boundaries, obstacle prediction trajectories, and planning constraint information to model and solve the drivable area of ​​a vehicle in complex traffic scenarios.

[0003] However, existing solutions rely heavily on predicted trajectories and safe distances for coarse-grained constraints on dynamic obstacles, lacking fine-grained boundary representations for different types of obstacles, which can easily lead to incomplete corridor boundary information. At the same time, the continuity of time-dimensional constraints is insufficient, making it difficult to provide stable and accurate constraint inputs for trajectory planning in high-speed or highly interactive scenarios.

[0004] Therefore, it is important to improve the spatiotemporal consistency of obstacle constraint modeling in complex traffic scenarios in order to enhance the safety and feasibility of trajectory planning. Summary of the Invention

[0005] This application provides a method for constructing path spatiotemporal corridor data, an electronic device, and a storage medium to solve the aforementioned technical problems. This method, geared towards autonomous driving scenarios, unifies and integrates the spatial range and temporal constraints of the path during vehicle operation, as well as the boundary and behavioral information of relevant objects in the scenario, to construct path spatiotemporal corridor data that can characterize the drivable area and spatiotemporal constraint relationships of the vehicle. This provides a more complete and stable constraint foundation for trajectory planning in complex traffic scenarios.

[0006] In a first aspect, embodiments of this application provide a method for constructing path spatiotemporal corridor data, including:

[0007] The first path spatiotemporal corridor data and time sampling point sequence of the vehicle are determined. The first path spatiotemporal corridor data includes static boundary data, road curve coordinate system envelope data, vehicle starting point status information and vehicle geometric parameters.

[0008] Determine the object boundary information, which includes at least one of static object boundary information, dynamic object boundary information, and multiple types of object boundary information;

[0009] Generate dynamic boundary data for the vehicle;

[0010] Acquire object pre-decision information, which includes at least one of following, detouring, and avoiding;

[0011] Based on the first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence, and object pre-decision information, second path spatiotemporal corridor data is constructed, which is used to assist in autonomous driving.

[0012] In one possible embodiment, determining the vehicle's first path spatiotemporal corridor data includes:

[0013] Acquire static boundary data, which includes lane lines, curbs, and non-crossable areas;

[0014] Based on static boundary data, the spatiotemporal corridor data of the first path is determined.

[0015] In one possible embodiment, determining the time sampling point sequence includes:

[0016] Based on static boundary data, road curve coordinate system envelope data, vehicle starting state information, and vehicle geometric parameters, a time sampling point sequence is generated.

[0017] In one possible embodiment, determining object boundary information includes:

[0018] Determine the sampling range and static object type information;

[0019] Within the sampling range of the vehicle, half-contour sampling is performed on objects of static object type information to obtain static object boundary information.

[0020] In one possible embodiment, determining object boundary information includes:

[0021] Determine the trajectory type information of dynamic objects;

[0022] Half-contour sampling is performed on objects with dynamic object trajectory type information to obtain dynamic object boundary information.

[0023] In one possible embodiment, it also includes:

[0024] The object contour in the boundary information of dynamic objects is expanded in both the front and back directions.

[0025] In one possible embodiment, determining object boundary information includes:

[0026] Obtain multi-type object type information, which includes at least one of the following: occupying grid objects, clustered objects, and movable curbs;

[0027] By sampling objects with multiple object type information, boundary information of multiple object types can be obtained.

[0028] In one possible embodiment, generating dynamic boundary data for the vehicle includes:

[0029] Obtain the first parameter, which includes at least one of the following: avoidance direction, avoidance buffer, avoidance gain, collision time, lateral offset, and time decay coefficient;

[0030] Based on the first parameter, the dynamic boundary data of the vehicle is generated.

[0031] In one possible embodiment, it also includes:

[0032] When the spatiotemporal consistency check of the dynamic boundary data fails, interpolation is performed to complete the dynamic boundary data.

[0033] In one possible embodiment, half-contour sampling is performed on the object with dynamic object trajectory type information to obtain dynamic object boundary information, including:

[0034] Determine the first time step based on the dynamic object trajectory type information;

[0035] Based on the first time step, half-contour sampling is performed on objects with dynamic object trajectory type information to obtain dynamic object boundary information.

[0036] In one possible embodiment, the dynamic object trajectory type information includes at least one of the following: dynamic trajectories of objects to the left and right of the vehicle, dynamic trajectories of vehicles cutting in, dynamic trajectories of vehicles playing against each other, pedestrian movement trajectories, bicycle movement trajectories, and dynamic trajectories of non-motorized vehicles.

[0037] In one possible embodiment, it also includes:

[0038] Obtain the density values ​​of objects within a preset range around the vehicle;

[0039] When the object density value is greater than the first threshold, the sampling point density of the half-contour sampling is increased.

[0040] Secondly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0041] The memory stores instructions that the computer executes;

[0042] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0043] Thirdly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0044] The path spatiotemporal corridor data construction method, electronic device, and storage medium provided in this application embodiment determine the first path spatiotemporal corridor data and time sampling point sequence of the vehicle, and construct the second path spatiotemporal corridor data by combining object boundary information, vehicle dynamic boundary data, and object pre-decision information. This can integrate the road boundary, vehicle's own state, constraints of different types of objects, and time dimension constraints, thereby improving the integrity and spatiotemporal consistency of path spatiotemporal corridor boundary modeling in complex traffic scenarios, and providing more stable, accurate, and executable constraint inputs for vehicle autonomous driving trajectory planning. Attached Figure Description

[0045] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0046] Figure 1 A flowchart illustrating the method for constructing path spatiotemporal corridor data provided in this application embodiment;

[0047] Figure 2 A schematic diagram of the structure of the path spatiotemporal corridor data construction device provided in the embodiments of this application;

[0048] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0049] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0050] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0051] Autonomous driving path planning falls under the field of vehicle automatic control and intelligent driving decision-making technology, and is typically applied to complex traffic scenarios such as urban roads, highways, intersections, and construction detours. In these scenarios, the system needs to integrate information such as road boundaries, obstacle trajectories, vehicle status, and vehicle dimensions to construct the vehicle's drivable area in the future time domain and provide a constraint basis for trajectory planning.

[0052] In existing autonomous driving solutions, the future movement space of the vehicle is often modeled based on road boundary information, predicted obstacle trajectories, and safe distance rules. The modeling results are then provided to the trajectory planning module for trajectory solving. The basic idea of ​​this type of solution is to use static boundaries as spatial constraints, transform the predicted results of dynamic obstacles into temporal constraints, and then combine them with the vehicle's current position, speed, and other state information to generate a driving corridor or constraint area that can be used by the planner for reference.

[0053] However, in complex traffic environments, the above-mentioned solutions typically provide a rather coarse representation of obstacle constraints, relying more on predicted trajectories and fixed safety margins. They lack a unified organization and detailed characterization of boundary information for static objects, dynamic objects, and multiple types of objects, which can easily lead to problems such as missing corridor boundaries and incomplete constraint coverage. Especially when there are temporary roadblocks, clustered objects, cutting-in vehicles, or vulnerable road users, a single constraint method is difficult to accurately reflect the actual drivable space.

[0054] Meanwhile, existing solutions often only apply discrete constraints to local moments in the time dimension, failing to continuously organize boundary information around a unified sequence of time sampling points. This results in discontinuous and unstable constraint inputs obtained during trajectory planning. In high-speed or highly interactive scenarios, this deficiency can further affect the smoothness, executability, and safety of the planning results, and may even cause vehicles to lag in responding to complex obstacle relationships.

[0055] In light of this, improving the completeness and spatiotemporal consistency of path constraint modeling has become a key issue in autonomous driving planning. To address this problem, a method for constructing path spatiotemporal corridor data is proposed. This method first determines the vehicle's first path spatiotemporal corridor data and time sampling point sequence, then determines object boundary information and generates the vehicle's dynamic boundary data. Combined with acquired object pre-decision information, second path spatiotemporal corridor data is constructed based on the first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence, and object pre-decision information. This provides a more stable and complete constraint input for autonomous driving.

[0056] In terms of implementation, this method can be deployed in a modular system architecture. The relevant system may include an input module, a static corridor construction module, a dynamic obstacle handling module, a spatiotemporal consistency verification module, and a trajectory optimization interface. Each module works together to complete the reception of boundary information, the organization of corridor data, and the output of planning constraints to adapt to the needs of autonomous driving applications in complex traffic scenarios.

[0057] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0058] Figure 1 This is a flowchart illustrating the method for constructing path spatiotemporal corridor data provided in this application embodiment. The method provided in this application embodiment can be applied to any electronic device. Figure 1 As shown, the method includes:

[0059] S101. Determine the first path spatiotemporal corridor data and time sampling point sequence of the vehicle. The first path spatiotemporal corridor data includes static boundary data, road curve coordinate system envelope data, vehicle starting point status information, and vehicle geometric parameters.

[0060] In this application, the first path spatiotemporal corridor data serves as the foundational constraint data for constructing the second path spatiotemporal corridor data, representing the initial feasible range of the vehicle in both the path and time dimensions. Static boundary data characterizes the static spatial constraints within the road; the road curve coordinate system envelope data can be boundary envelope information in a curve coordinate system established based on the reference road centerline or reference trajectory, used to convert the road geometry in the global Cartesian coordinate system into longitudinal and lateral coordinate constraints suitable for planning processing; the vehicle's starting state information includes at least one or more of the vehicle's position, heading angle, longitudinal velocity, lateral velocity, longitudinal acceleration, and yaw rate at the planning start time; the vehicle's geometric parameters include at least vehicle length, width, front overhang length, rear overhang length, wheelbase, and the vehicle's circumscribed rectangular or polygonal contour parameters, used to reflect the vehicle's physical dimensions during boundary modeling. The time sampling point sequence is used to discretize the future planning time domain, serving as a unified time index for subsequent dynamic boundary updates and spatiotemporal consistency processing.

[0061] In practice, the executing entity can be either the path constraint construction module deployed in the autonomous driving domain controller or the pre-planning processing module. This executing entity first receives static boundary data from the environmental perception module or the high-precision map module. Then, based on the current navigation path, local reference lines, or the target lane centerline given by the upper-level decision, the executing entity establishes a road curve coordinate system and projects the static boundary onto this system, forming the left and right envelope ranges distributed along the longitudinal stations. For the road curve coordinate system envelope box data, the reference lines can be discretized according to a preset method, and the permissible lateral traffic intervals can be recorded at the discretized stations. Vehicle starting point state information can be jointly provided by the positioning module, vehicle chassis state module, and inertial measurement unit, and time synchronization and coordinate alignment can be completed before entering this step. Vehicle geometric parameters can be read from the vehicle model configuration file and transformed into the planning coordinate system via the vehicle coordinate system, ensuring that the subsequent boundary expansion is consistent with the actual vehicle dimensions.

[0062] In one possible embodiment, the time sampling point sequence is generated according to the planned time domain length and sampling interval, thus obtaining a set of sampling times arranged in ascending order of time. After generating the time sampling point sequence, the executing entity associates each time sampling point with the vehicle's starting state information, so that the first path spatiotemporal corridor data not only includes static spatial boundaries but also has a clear index on the time axis. Based on the above analysis, it can be seen that by unifying the static boundaries, road curve coordinate system envelope data, vehicle starting state information, and vehicle geometric parameters into the first path spatiotemporal corridor data and establishing a unified time sampling point sequence, various dynamic obstacle constraints can be superimposed within the same coordinate frame and the same time frame, thereby enabling the corridor construction process to have basic integrity and computability. It should be understood that the above example is only for demonstration and not a limitation.

[0063] S102. Determine the object boundary information, which includes at least one of static object boundary information, dynamic object boundary information, and multi-type object boundary information.

[0064] In this application, object boundary information is used to describe the boundary constraints formed by various objects around the vehicle on the drivable space, and is an important input for constructing dynamic boundary data and second path spatiotemporal corridor data. Static object boundary information can correspond to object boundaries whose positions remain basically unchanged within the planning period; dynamic object boundary information can correspond to target boundaries whose positions and orientations change in the future; multi-type object boundary information can represent a unified set of boundaries formed after different categories of targets are expressed using differentiated boundary representations. This object boundary information can be generated in a Cartesian coordinate system and then transformed to a road curve coordinate system, or it can be directly organized in the road curve coordinate system according to longitudinal and lateral positions.

[0065] In practice, the executing entity receives a target list and target contour data from the perception fusion module. The target list may include target identifier, category, location, velocity, acceleration, heading, size, and confidence level; the target contour data may include point cloud clustering boundaries, image segmentation contour projection results, or radar tracking boxes. For static objects, the executing entity can determine the boundary information of static objects based on the perception results. For dynamic objects, the executing entity can determine the boundary information of dynamic objects based on the state information of dynamic objects. For multiple types of objects, the executing entity can determine the corresponding boundary information according to the target category. Subsequently, the executing entity maps all boundary information to a road curve coordinate system consistent with the first path spatiotemporal corridor data and records the longitudinal occupancy interval, lateral occupancy interval, and boundary contour index of each boundary at the corresponding time sampling point.

[0066] In one possible implementation, object boundary information is aligned one-to-one with the sequence of time sampling points, meaning a set of object boundaries is maintained for each time sampling point. Based on the above analysis, it can be seen that by unifying the organization of static, dynamic, and multi-type object boundaries, a more complete boundary representation can be formed in complex traffic scenarios. This allows subsequent corridor construction to no longer rely on a single obstacle constraint form, thereby reducing the occurrence of missing boundaries and incomplete constraint coverage. It should be understood that the above examples are merely illustrative and not limiting.

[0067] S103. Generate dynamic boundary data for the vehicle.

[0068] In this application, dynamic boundary data is used to characterize the boundary changes of a vehicle at various future sampling points, serving as the basis for dynamic constraints in the second path spatiotemporal corridor data. This dynamic boundary data does not merely represent the vehicle's own geometric contour, but rather is the vehicle's reachable or safe boundary data formed by the combined effects of the vehicle's starting state information, extrapolated vehicle motion results, vehicle geometric parameters, and boundary constraints of surrounding objects. Dynamic boundary data can be represented as a time-varying sequence of vehicle envelope polygons, a sequence of longitudinal and lateral feasible intervals in the road curve coordinate system, or a set of left and right boundary offsets organized according to sampling times.

[0069] In practice, the executing entity first reads the vehicle's starting point state information and geometric parameters from the first path spatiotemporal corridor data, and then makes a preliminary prediction of the vehicle's future attitude based on the time sampling point sequence. Subsequently, based on the predicted center point and predicted heading, the executing entity generates the vehicle's entity profile at each time sampling point, incorporating vehicle length, width, front overhang length, and rear overhang length. To enable this dynamic boundary to be directly used as a planning constraint, the executing entity further projects the vehicle entity profile onto the road curve coordinate system to obtain the longitudinal and lateral ranges occupied by the vehicle. Next, the executing entity introduces object boundary information to perform time-varying corrections on the vehicle boundary. Correction methods may include compressing the vehicle's feasible lateral offset range based on the minimum distance requirement between adjacent object boundaries and the vehicle boundary; for dynamic preceding vehicles, increasing the time-varying clearance distance on the vehicle's forward boundary based on the expected following distance; and for lateral targets, applying additional expansion on the corresponding lateral boundary. If the vehicle's predicted boundary overlaps with an object boundary at a certain time sampling point, the executing entity marks that sampling point as a restricted point and performs boundary recalculation or interpolation correction based on adjacent non-conflicting sampling points.

[0070] In one possible embodiment, the generation of dynamic boundary data employs an iterative approach using time-series sampling points. For each time-series sampling point, a nominal boundary is first obtained based on the vehicle's motion state, and then collision detection is performed using the set of object boundaries corresponding to that moment. When a boundary conflict is detected, the execution entity adjusts the boundary according to a preset priority. For cases where boundary gaps appear at multiple consecutive sampling points, spline interpolation or linear interpolation can be used to fill the gaps between the preceding and following valid boundaries, forming a continuous dynamic boundary curve. Based on the above analysis, it can be seen that by generating vehicle dynamic boundaries within a unified time-series sampling framework and combining them with time-series correction of object boundaries, the vehicle's own geometric dimensions, initial motion state, and relationships with surrounding obstacles can be synchronously incorporated into the constraint expression, thereby providing a continuous dynamic constraint foundation for subsequent spatiotemporal corridors. It should be understood that the above example is merely illustrative and not limiting.

[0071] S104. Obtain object pre-decision information, which includes at least one of following, detouring, and avoiding.

[0072] In this application, object pre-decision information is used to characterize the vehicle's pre-selective behavior towards surrounding target objects, in order to determine the constraint method of corresponding obstacles on the drivable space when constructing the second path spatiotemporal corridor data. "Following" indicates maintaining a longitudinal following relationship behind the target object; "Detour" indicates not following the target object's current trajectory but choosing to form a passageway from one side of the target object; "Avoidance" indicates setting the area corresponding to the target object as a restricted area that needs to be actively avoided. This pre-decision information can be generated for a single target or can form a combined decision set for multiple targets and be associated with target identifiers.

[0073] In practice, the executing entity receives object pre-decision information from the behavior decision module, scene understanding module, or upstream module of local planning. If the upstream module directly outputs behavior labels for each target, the executing entity reads the behavior label and establishes a mapping table from the target to the pre-decision type. If the upstream module does not directly output behavior labels, the executing entity can infer the pre-decision type based on object boundary information, target relative position, target speed, lane topology, and time sampling point sequence. For example, when a target is in front of the current vehicle in the lane, its speed is lower than the vehicle's, but the lanes on both sides are restricted, the executing entity marks the target as a following target and sets the corresponding longitudinal safety distance parameter. When the target is stationary and partially blocks the current lane, and there is continuous available lateral space in the adjacent lane, the executing entity marks the target as a detour target and determines the detour direction and the timing of entry. When the target is a pedestrian crossing, a vehicle rapidly cutting in, or a target that has a spatiotemporal conflict with the vehicle's predicted trajectory, the executing entity marks the target as an avoidance target and generates an avoidance area and avoidance duration. The specific data structure of pre-decision information may include target identifier, decision type, decision effective start time, decision effective end time, target priority, and constraint parameters corresponding to the decision. Constraint parameters may include minimum following distance, detour lateral interval, and avoidance time window.

[0074] In one possible implementation, the three types of pre-decision decisions—following, detouring, and yielding—are not mutually exclusive fixed values, but rather take effect separately at different time sampling points or for different targets. For example, a following constraint is applied to vehicles in the same lane ahead at the first few sampling points, but when the target moves to the side to slow down and the adjacent lane opens, a detouring constraint is switched to the next sampling point. Similarly, a yielding constraint is applied to pedestrian targets before they approach the conflict point, and the constraint is removed after the conflict is resolved. After obtaining the pre-decision information, the executing entity aligns it with the time sampling point sequence and generates a decision validity table for each sampling point, allowing the next step to select different boundary construction rules in the corresponding time slice. Based on the above analysis, it can be seen that after explicitly incorporating object pre-decision information into the corridor construction process, different obstacle targets are no longer uniformly processed only through a fixed safety margin, but different constraint generation logics are implemented based on specific interaction relationships, enabling the spatiotemporal corridor to reflect actual driving intentions such as following, detouring, and yielding. It should be understood that the above examples are merely illustrative and not limiting.

[0075] S105. Based on the first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence and object pre-decision information, construct the second path spatiotemporal corridor data. The second path spatiotemporal corridor data is used to assist the vehicle's autonomous driving.

[0076] In this application, the second path spatiotemporal corridor data is the final output of the path spatiotemporal constraint result. Within the static basis given by the first path spatiotemporal corridor data, it further integrates dynamic boundary data, object boundary information, unified time sampling point sequence, and object pre-decision information to form a spatiotemporal corridor that simultaneously possesses static road constraints, dynamic obstacle constraints, and behavioral decision constraints. The second path spatiotemporal corridor data can be represented as a set of passable intervals organized by time sampling points, or it can be represented as a two-dimensional corridor grid composed of longitudinal stations and time in the road curve coordinate system. Each grid cell records the passage permission status, left and right boundary values, and boundary source identifier at the corresponding time and corresponding path position.

[0077] In practice, the executing entity performs corridor fusion calculations for each time sampling point. First, it reads the static envelope range corresponding to that moment from the first path spatiotemporal corridor data as the basic feasible region. Then, it reads the dynamic boundary data for that moment, mapping the vehicle's geometric occupancy, motion extrapolation range, and necessary safety interval to the basic feasible region. Next, it reads the object boundary information for that moment, clipping and superimposing the static and dynamic obstacle boundaries, removing or compressing the passage areas overlapping with obstacles from the basic feasible region. Afterward, the executing entity corrects the behavioral constraints of the remaining region based on the object pre-decision information corresponding to that moment. When the target is a following vehicle, the executing entity retains the passage behind the target that meets the longitudinal safety distance requirements and marks the areas in front of the target and too close to the rear as inaccessible areas. When the target is a detour target, the executing entity selects a continuous lateral passable section on one side of the target and closes the blocked section on the other side, thus forming a directional detour channel. When the target is an avoidance target, the executing entity expands the restricted area around the target and cuts off the area overlapping with the target's spatiotemporal trajectory within the conflict time window. After the above processing, the executing entity obtains the candidate feasible region corresponding to each time sampling point, and then performs continuity verification on the boundary changes between adjacent time sampling points. Continuity verification may include threshold verification of the change rate of the left and right boundaries, topological consistency verification of feasible regions, and removal of isolated small regions; for boundary segments that fail the verification, the executing entity performs interpolation to complete, splice broken sections, or merge invalid regions based on the effective boundaries before and after, thereby obtaining continuous second path spatiotemporal corridor data in the time dimension.

[0078] In one possible embodiment, the construction of the second path spatiotemporal corridor data also includes establishing an output interface for the trajectory planner. The executing entity can encode the construction result as a time-indexed sequence of left and right boundaries, a spatial polygon sequence, or a constraint matrix, where each row of the constraint matrix corresponds to a time sampling point, and each column corresponds to the corridor boundary, longitudinal and lateral constraint values, obstacle category, and pre-decision marker. After receiving the second path spatiotemporal corridor data, the trajectory planning module can perform velocity planning, lateral offset planning, or joint trajectory optimization within its defined range. Based on the above analysis, this step integrates the static foundation provided by the first path spatiotemporal corridor data, the environmental obstacle constraints provided by the object boundary information, the time-varying boundary constraints provided by the dynamic boundary data, and the behavioral constraints provided by the object pre-decision information on a unified time sampling point sequence. It then outputs the second path spatiotemporal corridor data through continuity verification and boundary correction, enabling the planning module to obtain spatiotemporally consistent and constraint-complete input. This allows it to generate running trajectories that match the actual road boundaries and obstacle relationships in scenarios such as urban roads, highways, intersection traffic, and construction detours. It should be understood that the above example is merely illustrative and not limiting.

[0079] Based on the above analysis, this application provides a method for constructing path spatiotemporal corridor data, including determining the first path spatiotemporal corridor data and time sampling point sequence of a vehicle, determining object boundary information, generating dynamic boundary data of the vehicle, obtaining object pre-decision information, and constructing second path spatiotemporal corridor data based on the first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence, and object pre-decision information. In this application, a basic corridor is first established, containing static boundary data, road curve coordinate system envelope data, vehicle starting state information, and vehicle geometric parameters. Then, static objects, dynamic objects, and boundaries of multiple object types are organized around a unified time sampling point sequence. The dynamic boundaries and the final corridor are then corrected using pre-decision rules corresponding to following, detouring, and yielding, resulting in second path spatiotemporal corridor data oriented towards trajectory planning output. Therefore, the path constraint modeling process can cover road boundaries and multiple obstacle boundaries in the spatial dimension, maintain continuous organization and consistent updates in the temporal dimension, making the constraint input in complex traffic environments more complete and stable, and can be directly used to assist in trajectory solving during autonomous driving. It should be understood that the above examples are merely illustrative and not limiting.

[0080] In one possible implementation, determining the first path spatiotemporal corridor data for the vehicle includes: acquiring static boundary data, which includes lane lines, curbs, and non-crossable areas; and determining the first path spatiotemporal corridor data based on the static boundary data.

[0081] In practical implementation, the system first unifies the coordinates of lane lines, curbs, and uncrossable areas, transforming them to the road curve coordinate system or local Cartesian coordinate system currently used by the vehicle. Then, it projects and encloses the boundaries based on the vehicle's center point trajectory to obtain the left and right boundary ranges corresponding to the target path. Subsequently, considering vehicle length, width, and safety margin, the boundary ranges are expanded or contracted to ensure that the first path spatiotemporal corridor data covers the actual space occupied by the vehicle and the necessary safety intervals. When there are curvature changes in the road boundary, continuous interpolation can be performed along the road arc length direction to ensure the spatial continuity and consistency of the corridor boundary. After the first path spatiotemporal corridor data is output, it can be used as the basic constraint input for subsequent trajectory solving.

[0082] During the process, static boundary data first provides fixed road constraints for vehicles. Then, the first path spatiotemporal corridor data generated based on this boundary defines the drivable area of ​​the vehicle within a future period, enabling the planning module to obtain stable constraint boundaries in a static road environment. Since lane lines, curbs, and uncrossable areas are all uniformly incorporated into the boundary modeling, the resulting corridor data is consistent with the actual road morphology and can be directly used for subsequent path planning and trajectory optimization.

[0083] By adopting the above method, the spatiotemporal corridor data of the first path can form a continuous and clear spatial constraint input based on the static boundary of the road, reduce the situation of missing boundary expression, and enable vehicles to obtain more stable basic planning constraints in complex road environments, which facilitates subsequent dynamic boundary superposition and trajectory generation.

[0084] Based on the foregoing embodiments, further determining the time sampling point sequence includes:

[0085] Based on static boundary data, road curve coordinate system envelope data, vehicle starting state information, and vehicle geometric parameters, a time sampling point sequence is generated.

[0086] When generating the time sampling point sequence, the spatial range of the drivable area is first determined based on static boundary data. Then, the temporal coverage interval of the vehicle within this range is determined by combining the road curve coordinate system envelope data. Furthermore, the vehicle's motion evolution trend within the planned time domain is inferred based on the vehicle's initial state information. Subsequently, vehicle geometric parameters are introduced into the time-domain discretization process to constrain the sampling interval. This ensures that the sampling points cover the vehicle's state changes within curves, lane changes, or boundary contraction areas, while also meeting the matching requirements between the vehicle's outer contour and boundary constraints. The resulting time sampling point sequence can serve as a unified temporal reference, providing input for the subsequent synchronous organization of dynamic boundary data and object boundary information.

[0087] By adopting this method, the time sampling point sequence is consistent with the spatiotemporal reference of the static boundary, the road curve coordinate system envelope data, and the vehicle starting state information. Subsequent constraint data can be matched at the same time node, thereby improving the continuity and consistency of the path spatiotemporal corridor data and making the planning results more in line with vehicle geometric constraints and road spatial constraints.

[0088] Based on the foregoing embodiments, further determining object boundary information includes:

[0089] Determine the sampling range and static object type information;

[0090] Within the sampling range of the vehicle, half-contour sampling is performed on objects of static object type information to obtain static object boundary information.

[0091] After acquiring the environmental perception results around the vehicle, the sampling range is first determined based on the vehicle's current position, orientation, and a preset safe search distance. Within this range, target objects corresponding to static object type information are selected. Subsequently, when sampling the half-contour of the target object, only the contour points on the side facing the vehicle or the side related to the vehicle's planned travel direction are extracted. The collected contour points are then sorted and fitted according to the object's local coordinate system or the road curve coordinate system to form a continuous boundary representation. For elongated static objects, sampling can be performed at fixed intervals along their extension direction; for irregular static objects, boundary points can be extracted based on the relationship between the circumscribed polygon and the visible boundary, and the static object boundary information can be formed by connecting polylines or fitting splines.

[0092] In this application, semi-contour sampling refers to retaining only the portion of the boundary that has constraints on the vehicle's drivable space within the complete contour, thereby ensuring that the boundary representation is consistent with the actual traffic relationship of the vehicle. By combining the setting of the sampling range with the filtering of static object type information, irrelevant targets can be excluded, and then the valid targets are subjected to semi-contour extraction to output the static object boundary information for subsequent use in the construction of path spatiotemporal corridor constraints.

[0093] In this way, static object boundary information can accurately represent the spatial occupancy characteristics of static obstacles with fewer boundary points, while maintaining consistency with the vehicle sampling range. This makes the subsequent corridor constraint input more complete and stable. Since only the semi-contour portion with constraint significance is retained, the boundary extraction results are closer to the lateral passage requirements of vehicles, providing directly usable static spatial constraint data for autonomous driving planning.

[0094] Based on the foregoing embodiments, further determining object boundary information includes:

[0095] Determine the trajectory type information of dynamic objects;

[0096] Half-contour sampling is performed on objects with dynamic object trajectory type information to obtain dynamic object boundary information.

[0097] In the specific implementation, the trajectory type information of the dynamic object is first determined based on the target tracking results output by the perception module. The trajectory type information includes at least the representation quantities such as the direction of motion, the trend of velocity change, and the continuity of the trajectory. Subsequently, for targets that meet the trajectory type information, half-contour sampling is performed on the outer contour corresponding to the target's current posture, historical posture, or predicted posture. Only the contour point set facing the side of the vehicle's drivable space is extracted, and the sampled point set is sorted and stitched according to the target's orientation to form the boundary information of the dynamic object. The half-contour sampling can be adaptively set according to the target's length and width dimensions, heading angle, and pose changes, so that the sampling boundary covers the outer envelope of the space actually occupied by the target. The spacing between sampling points can be adjusted according to the target size and local curvature to ensure the continuity of the boundary representation.

[0098] During operation, the system first identifies the trajectory type of dynamic objects, then extracts the semi-contour of the object's outer contour corresponding to that type, and outputs dynamic object boundary information that can be used for corridor construction. This boundary information can describe the constraint range of dynamic targets on vehicles with fewer contour points and maintain the correspondence with the target's motion state.

[0099] By adopting this method, the boundaries of dynamic objects can more accurately reflect the space occupied by the target and its motion-related constraints, and the boundary expression is more complete. When used by the planning module, it can improve the continuity and stability of constraint input, thereby enhancing the quality of path spatiotemporal corridor construction in complex traffic scenarios.

[0100] One possible implementation also includes:

[0101] The object contour in the boundary information of dynamic objects is expanded in both the front and back directions.

[0102] In practical implementation, the object contour at the corresponding moment can be obtained first based on the dynamic object trajectory type information, and then mapped to the road curve coordinate system or vehicle coordinate system. Next, an expansion offset is applied to the front and rear boundaries of the contour along the longitudinal direction to obtain the expanded boundary point set or boundary polygon. For low-speed or stationary dynamic objects, the bidirectional expansion amount can be set to a small fixed value; for high-speed cutting-in, following, or lateral weaving dynamic objects, the bidirectional expansion amount can be adjusted according to collision risk, collision time, or time decay coefficient, so that the expanded boundary changes dynamically with time. In this process, bidirectional expansion can be achieved through geometric translation, envelope expansion, or grid occupancy expansion. The output is the expanded dynamic object boundary data, which is stored corresponding to the time sampling point sequence.

[0103] This processing method enables dynamic object boundaries to form more complete occupancy constraints based on the original contours. Subsequently, when constructing the second path spatiotemporal corridor data, the planning module can directly use the extended dynamic boundaries as constraints in trajectory solving. Because the boundaries are extended in both the forward and backward directions, vehicles are more conservative in judging the passability gaps of dynamic obstacles in complex interaction scenarios. The generated spatiotemporal corridor boundaries are more continuous and have more complete coverage, and can form a stable planning constraint input together with the object pre-decision information.

[0104] In one possible implementation, determining object boundary information includes: acquiring multi-type object type information, which includes at least one of grid-occupying objects, clustered objects, and movable curbs; and sampling the objects in the multi-type object type information to obtain multi-type object boundary information.

[0105] In practical implementation, the system can receive object information from LiDAR, cameras, or high-precision maps via a perception fusion module, classify and label the object types, and generate multi-type object information. For objects occupying grids, sampling points can be extracted based on the outer boundaries of continuous occupying cells in the grid. The sampling interval can be set to 0.1m to 0.5m according to the grid resolution to maintain the continuity of the boundary contour. For clustered objects, adjacent targets can be clustered first, and the outer contour can be obtained. Then, discrete sampling points can be extracted along the contour curve. The sampling point spacing can be set according to the target size and planning accuracy. For movable curbs, sampling points can be extracted based on their current position, normal direction, and edge polyline, and the changed boundary position is synchronously updated to the boundary information set. When multi-type object information contains multiple objects, the system samples each object separately, merges, removes duplicates, and sorts them to form unified multi-type object boundary information, which is then output to the spatiotemporal corridor construction module for subsequent constraint use. In practical applications, this module can also use other types of perception computing units or boundary extraction units; this application does not limit this.

[0106] This method samples grid objects, clustered objects, and movable curbs separately, enabling boundary objects of different shapes and origins to be uniformly expressed as boundary information that can be used for planning constraints, and ensuring that the boundary contours are consistent with the real obstacle shapes in the road scene.

[0107] By adopting this method, the boundary information of multiple object types can cover a richer range of road boundary sources, the boundary expression is more complete, and it is easier to use in conjunction with the time sampling point sequence, thereby giving the subsequently generated path spatiotemporal corridor a more stable constraint input.

[0108] Based on the foregoing embodiments, the dynamic boundary data of the vehicle is further generated by: obtaining a first parameter, the first parameter including at least one of avoidance direction, avoidance buffer, avoidance gain, collision time, lateral offset and time decay coefficient; and generating dynamic boundary data of the vehicle based on the first parameter.

[0109] In its implementation, the controller receives a first parameter from the path planning module or obstacle decision module and maps it to the spatiotemporal envelope of a dynamic boundary based on the current vehicle speed, vehicle geometry parameters, and the relative position of the target object. If the first parameter includes an avoidance direction, the vehicle's boundary centerline in the reference coordinate system is laterally offset according to the left or right avoidance direction marker. If it includes an avoidance buffer zone, a preset safety margin is superimposed on the outer side of the target object's outline. If it includes an avoidance gain, the boundary expansion coefficient is amplified or contracted. If it includes a collision time, boundary expansion is enabled only within the corresponding time window. If it includes a lateral offset, the boundary is translated laterally according to the offset. If it includes a time decay coefficient, the boundary expansion decreases or converges with the planning time domain. These parameters can be used individually or in combination, and the controller outputs dynamic boundary data corresponding to the time sampling point sequence. In practical applications, the controller can also select different parameter parsing modes, which are not limited in this application.

[0110] This dynamic boundary data is used to uniformly express avoidance intentions, risk temporal domain, and spatial margin into a constraint region that can be invoked by the planner. This allows the boundary shape to adjust according to scene changes and remain consistent with object boundary information and the time sampling point sequence. The dynamic boundary data generated based on the first parameter can refine the drivable space of vehicles in complex traffic environments and provide continuous constraint inputs for subsequent path solving.

[0111] By adopting the above method, the vehicle dynamic boundary no longer relies on a fixed width or static expansion rules, but can be adaptively generated based on different avoidance parameters, thereby improving the specificity, temporal continuity, and scene adaptability of the boundary representation. With the dynamic boundary data updated synchronously with the planning constraints, the generated results can more closely reflect actual avoidance requirements.

[0112] Based on the foregoing embodiments, it further includes:

[0113] When the spatiotemporal consistency check of the dynamic boundary data fails, interpolation is performed to complete the dynamic boundary data.

[0114] When the verification results indicate that the dynamic boundary data does not meet the spatiotemporal consistency requirements, the system performs interpolation to complete the boundary points corresponding to the missing time points based on the dynamic boundary data already generated at adjacent time sampling points. The interpolation can be calculated based on the lateral offset of the boundary points, the boundary contour coordinates, or the boundary width parameter, ensuring that the completed dynamic boundary data remains continuous with the preceding and following sampling points in the time dimension and maintains a smooth transition of the boundary shape in the spatial dimension. The interpolation method can be linear interpolation, spline interpolation, or other interpolation methods that meet the boundary continuity requirements. In practical applications, other forms of interpolation can also be selected, and this application does not limit this choice.

[0115] After interpolation and completion, the system can perform a spatiotemporal consistency check on the completed dynamic boundary data again to confirm that there are no obvious temporal discontinuities or spatial offsets between the completed result and the original dynamic boundary data. If the check passes, the completed dynamic boundary data is output to the second path spatiotemporal corridor data construction process; if it still fails, the corresponding boundary values ​​can continue to be corrected based on adjacent valid sampling points until dynamic boundary data that meets the constraints is formed.

[0116] Through the above processing, dynamic boundary data can be completed into continuous data in the case of missing sampling points, local anomalies, or boundary breaks, so that the subsequently constructed path spatiotemporal corridors are consistent in time and space, thereby making the planning constraint input more complete and stable, and enabling the output second path spatiotemporal corridor data to more accurately reflect the vehicle's drivable space.

[0117] Based on the aforementioned embodiments, further, half-contour sampling is performed on the object with dynamic object trajectory type information to obtain dynamic object boundary information, including: determining a first time step based on dynamic object trajectory type information; and performing half-contour sampling on the object with dynamic object trajectory type information based on the first time step to obtain dynamic object boundary information.

[0118] In implementation, after receiving the trajectory type information of the dynamic object, the controller first parses the type, calculates the velocity change rate and trajectory curvature of the target object based on the trajectory history point sequence, and generates a first time step according to a preset mapping relationship. Subsequently, the system discretely samples the outer contour of the target object at consecutive time intervals, retaining only the half-contour point set facing the vehicle's drivable space, and uniformly organizes the sampled points at each time step into dynamic object boundary information. For scenarios using a shorter first time step, the time interval between sampling points is denser, which can reflect boundary changes during cutting in, paralleling, or traversing; for scenarios using a longer first time step, unnecessary repeated sampling can be reduced, lowering the computational burden.

[0119] During operation, this structure first determines the sampling interval based on the trajectory type information of the dynamic object, and then extracts the semi-contour boundary according to the interval. This ensures that the boundary information is updated synchronously with the object's motion state, thus giving the output dynamic object boundary information continuity and temporal consistency. It can also be directly used as constraint input in the construction of path spatiotemporal corridors. Because the sampling granularity adjusts with the trajectory type, dynamic objects can form denser boundary descriptions in high-speed motion, sharp turns, or complex interactive scenes, while maintaining high computational efficiency in low-speed, stable scenes.

[0120] After adopting this implementation method, the dynamic object boundary information can more accurately reflect the occupied range of the target object at different times, the matching degree between the sampling results and the motion state is higher, thereby improving the integrity and spatiotemporal consistency of the subsequent corridor constraints, and making the obstacle expression on which the trajectory planning is based more stable.

[0121] Based on the aforementioned embodiments, the dynamic object trajectory type information further includes at least one of the following: dynamic trajectories of objects to the left and right of the vehicle, dynamic trajectories of vehicles cutting in, dynamic trajectories of vehicles playing against each other, pedestrian movement trajectories, bicycle movement trajectories, and dynamic trajectories of non-motorized vehicles.

[0122] When the dynamic object trajectory type information includes the dynamic trajectories of objects to the left and right of the vehicle or the dynamic trajectory of vehicles cutting in, the system will extract denser sampling points for the lateral and forward interaction boundaries of the vehicle body, and use the sampling results to form dynamic object boundary information. When it includes the dynamic trajectories of vehicles in opposing games, it will focus on representing the intersection boundaries in the opposing traffic space. When it includes the dynamic trajectories of pedestrians, bicycles, or non-motorized vehicles, it will combine their low speed, high randomness, and lateral swaying characteristics to construct boundary constraints that are closer to the actual motion envelope. By incorporating the trajectory categories of different traffic participants into unified trajectory type information, the dynamic object boundary information can cover richer road interaction scenarios and provide more complete dynamic constraint inputs for the subsequent construction of path spatiotemporal corridors.

[0123] Based on the foregoing embodiments, it further includes:

[0124] Obtain the density values ​​of objects within a preset range around the vehicle;

[0125] When the object density value is greater than the first threshold, the sampling point density of the half-contour sampling is increased.

[0126] In practical implementation, the object density value within a preset range around the vehicle can be calculated by the perception module based on target detection results, grid occupancy results, or multi-target tracking results. The controller counts the number of objects per unit area, the number of target center points, or the number of bounding boxes within this range and maps the statistical values ​​to density values. The sampling point density of half-contour sampling can be achieved by shortening the spacing between adjacent sampling points, increasing the number of sampling points on the same contour, or increasing the sampling frequency of curve interpolation. When the object density value exceeds a first threshold, the controller calls high-density sampling parameters to reduce the original sampling interval, so that the dynamic object boundary obtains a finer-grained contour expression in the dense target area around the vehicle. If the object density value around the vehicle is not greater than the first threshold, the original sampling point density is maintained to maintain a balance between computing resources and boundary accuracy.

[0127] In this implementation, the density value determination result directly drives the adjustment of the half-contour sampling accuracy. After completing the object density assessment, the controller updates the sampling configuration in real time and uses the adjusted set of sampling points for subsequent dynamic object boundary information generation, thereby making the boundary characterization in high-density scenes more continuous and complete. Since the sampling point density adapts to the degree of environmental congestion, the dynamic object boundary information can form more detailed constraint inputs in scenarios with nearby vehicles, cutting-in vehicles, or pedestrian gatherings, thereby improving the boundary consistency and constraint coverage of the second path spatiotemporal corridor data.

[0128] This application provides a method for constructing path spatiotemporal corridor data. During execution, static boundary data of the vehicle's driving environment is first acquired. This static boundary data includes lane lines, road edges, and uncrossable areas. Combined with road curve coordinate system envelope data, vehicle starting point state information, and vehicle geometric parameters, the first path spatiotemporal corridor data of the vehicle is determined. The first path spatiotemporal corridor data includes static boundary data, road curve coordinate system envelope data, vehicle starting point state information, and vehicle geometric parameters. Simultaneously, a sequence of time sampling points is generated based on the static boundary data, road curve coordinate system envelope data, vehicle starting point state information, and vehicle geometric parameters, ensuring that each time sampling point corresponds to a spatial range in the path curve coordinate system.

[0129] When determining environmental object information, the sampling range around the vehicle and the static object type information are first determined. Within the sampling range, half-contour sampling is performed on the objects corresponding to the static object type information to obtain static object boundary information. For dynamic objects, the trajectory type information of the dynamic objects is first determined. The dynamic object trajectory type information includes at least one of the following: the dynamic trajectory of objects to the left and right sides of the vehicle, the dynamic trajectory of vehicles cutting in, the dynamic trajectory of vehicles playing against each other, the movement trajectory of pedestrians, the movement trajectory of bicycles, and the dynamic trajectory of non-motorized vehicles. Then, a first time step is determined based on the dynamic object trajectory type information, and half-contour sampling is performed on the dynamic objects according to the first time step to obtain dynamic object boundary information. The object contours in the dynamic object boundary information are further expanded in both directions. During sampling, the object density value within a preset range around the vehicle is also obtained. When the object density value is greater than a first threshold, the sampling point density of the half-contour sampling is increased. For multi-type objects, multi-type object type information is obtained. The multi-type object type information includes at least one of the following: objects occupying grids, clustered objects, and movable curbs. These objects are sampled to obtain multi-type object boundary information.

[0130] During the vehicle boundary modeling phase, first parameters are acquired, including at least one of avoidance direction, avoidance buffer, avoidance gain, collision time, lateral offset, and time decay coefficient. Dynamic boundary data of the vehicle is then generated based on these first parameters. After generation, spatiotemporal consistency verification is performed on the dynamic boundary data. If the verification fails, interpolation is performed on the dynamic boundary data to ensure continuity of the dynamic boundary at each time sampling point.

[0131] Subsequently, object pre-decision information is acquired, which includes at least one of following, detouring, and yielding. The first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence, and object pre-decision information are input into the construction process. Static boundaries, vehicle dynamic boundaries, static object boundaries, dynamic object boundaries, and boundaries of multiple object types are uniformly constrained in the road curve coordinate system. The second path spatiotemporal corridor data of the vehicle is formed by combining the spatial occupancy relationship at different time sampling points.

[0132] Figure 2 This is a schematic diagram of the structure of the path spatiotemporal corridor data construction device provided in the embodiments of this application, as shown below. Figure 2 As shown, the path spatiotemporal corridor data construction device 20 provided in this embodiment includes a determination module 201, a processing module 202, and an acquisition module 203.

[0133] The determination module 201 is used to determine the first path spatiotemporal corridor data and time sampling point sequence of the vehicle. The first path spatiotemporal corridor data includes static boundary data, road curve coordinate system envelope data, vehicle starting point status information and vehicle geometric parameters.

[0134] The determination module 201 is also used to determine object boundary information, which includes at least one of static object boundary information, dynamic object boundary information, and multi-type object boundary information.

[0135] Processing module 202 is used to generate dynamic boundary data of the vehicle;

[0136] The acquisition module 203 is used to acquire object pre-decision information, which includes at least one of following, detouring, and avoiding.

[0137] The processing module 202 is also used to construct second path spatiotemporal corridor data based on first path spatiotemporal corridor data, dynamic boundary data, object boundary information, time sampling point sequence and object pre-decision information. The second path spatiotemporal corridor data is used to assist in autonomous driving of vehicles.

[0138] In one possible embodiment, determining the vehicle's first path spatiotemporal corridor data includes:

[0139] Acquire static boundary data, which includes lane lines, curbs, and non-crossable areas;

[0140] Based on static boundary data, the spatiotemporal corridor data of the first path is determined.

[0141] In one possible embodiment, determining the time sampling point sequence includes:

[0142] Based on static boundary data, road curve coordinate system envelope data, vehicle starting state information, and vehicle geometric parameters, a time sampling point sequence is generated.

[0143] In one possible embodiment, determining object boundary information includes:

[0144] Determine the sampling range and static object type information;

[0145] Within the sampling range of the vehicle, half-contour sampling is performed on objects of static object type information to obtain static object boundary information.

[0146] In one possible embodiment, determining object boundary information includes:

[0147] Determine the trajectory type information of dynamic objects;

[0148] Half-contour sampling is performed on objects with dynamic object trajectory type information to obtain dynamic object boundary information.

[0149] In one possible embodiment, it also includes:

[0150] The object contour in the boundary information of dynamic objects is expanded in both the front and back directions.

[0151] In one possible embodiment, determining object boundary information includes:

[0152] Obtain multi-type object type information, which includes at least one of the following: occupying grid objects, clustered objects, and movable curbs;

[0153] By sampling objects with multiple object type information, boundary information of multiple object types can be obtained.

[0154] In one possible embodiment, generating dynamic boundary data for the vehicle includes:

[0155] Obtain the first parameter, which includes at least one of the following: avoidance direction, avoidance buffer, avoidance gain, collision time, lateral offset, and time decay coefficient;

[0156] Based on the first parameter, the dynamic boundary data of the vehicle is generated.

[0157] In one possible embodiment, it also includes:

[0158] When the spatiotemporal consistency check of the dynamic boundary data fails, interpolation is performed to complete the dynamic boundary data.

[0159] In one possible embodiment, half-contour sampling is performed on the object with dynamic object trajectory type information to obtain dynamic object boundary information, including:

[0160] Determine the first time step based on the dynamic object trajectory type information;

[0161] Based on the first time step, half-contour sampling is performed on objects with dynamic object trajectory type information to obtain dynamic object boundary information.

[0162] In one possible embodiment, the dynamic object trajectory type information includes at least one of the following: dynamic trajectories of objects to the left and right of the vehicle, dynamic trajectories of vehicles cutting in, dynamic trajectories of vehicles playing against each other, pedestrian movement trajectories, bicycle movement trajectories, and dynamic trajectories of non-motorized vehicles.

[0163] In one possible embodiment, it also includes:

[0164] Obtain the density values ​​of objects within a preset range around the vehicle;

[0165] When the object density value is greater than the first threshold, the sampling point density of the half-contour sampling is increased.

[0166] The path spatiotemporal corridor data construction device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0167] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 30 provided in this embodiment includes at least one processor 301 and a memory 302. Optionally, the device 30 further includes a communication component 303. The processor 301, memory 302, and communication component 303 are connected via a bus.

[0168] In a specific implementation, at least one processor 301 executes computer execution instructions stored in memory 302, causing at least one processor 301 to perform the above-described method.

[0169] The specific implementation process of processor 301 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0170] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0171] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0172] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0173] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0174] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0175] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0176] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0177] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0178] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0180] If a function 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 this invention, or the part that contributes to the prior art, or a 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 of the various embodiments of this 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.

[0181] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0182] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for constructing path spatiotemporal corridor data, characterized in that, include: The first path spatiotemporal corridor data and time sampling point sequence of the vehicle are determined. The first path spatiotemporal corridor data includes static boundary data, road curve coordinate system envelope data, vehicle starting point state information and vehicle geometric parameters. Determine object boundary information, wherein the object boundary information includes at least one of static object boundary information, dynamic object boundary information, and multi-type object boundary information; Generate the dynamic boundary data of the vehicle; Obtain object pre-decision information, wherein the object pre-decision information includes at least one of following, detouring, and avoiding; Based on the first path spatiotemporal corridor data, the dynamic boundary data, the object boundary information, the time sampling point sequence, and the object pre-decision information, a second path spatiotemporal corridor data is constructed, which is used to assist the vehicle's autonomous driving.

2. The method according to claim 1, characterized in that, The spatiotemporal corridor data for determining the first path of the vehicle includes: Acquire static boundary data, which includes lane lines, road edges, and non-crossable areas; Based on the static boundary data, the spatiotemporal corridor data of the first path is determined.

3. The method according to claim 1, characterized in that, The determination of the time sampling point sequence includes: Based on the static boundary data, the road curve coordinate system envelope data, the vehicle starting point state information, and the vehicle geometric parameters, a time sampling point sequence is generated.

4. The method according to claim 1, characterized in that, The determination of object boundary information includes: Determine the sampling range and static object type information; Within the sampling range of the vehicle, half-contour sampling is performed on the object of the static object type information to obtain the static object boundary information.

5. The method according to claim 1, characterized in that, The determination of object boundary information includes: Determine the trajectory type information of dynamic objects; The object with the dynamic object trajectory type information is sampled in half contour to obtain the dynamic object boundary information.

6. The method according to claim 5, characterized in that, Also includes: The object contour in the dynamic object boundary information is expanded bidirectionally, both forward and backward.

7. The method according to claim 1, characterized in that, The determination of object boundary information includes: Obtain multi-type object type information, wherein the multi-type object type information includes at least one of grid-occupying objects, clustered objects, and movable curbs; The object type information of the multi-type object is sampled to obtain the boundary information of the multi-type object.

8. The method according to claim 1, characterized in that, The generation of the vehicle's dynamic boundary data includes: Obtain a first parameter, which includes at least one of the following: avoidance direction, avoidance buffer, avoidance gain, collision time, lateral offset, and time decay coefficient; Based on the first parameter, the dynamic boundary data of the vehicle is generated.

9. The method according to claim 1, characterized in that, Also includes: When the spatiotemporal consistency check of the dynamic boundary data fails, the dynamic boundary data is interpolated and padded.

10. The method according to claim 5, characterized in that, The step of sampling the half-contour of the object with the dynamic object trajectory type information to obtain the dynamic object boundary information includes: Based on the dynamic object trajectory type information, a first time step is determined; Based on the first time step, half-contour sampling is performed on the object whose trajectory type information is dynamic to obtain the boundary information of the dynamic object.

11. The method according to any one of claims 5 or 10, characterized in that, The dynamic object trajectory type information includes at least one of the following: the dynamic trajectory of objects to the left and right sides of the vehicle, the dynamic trajectory of vehicles cutting in, the dynamic trajectory of vehicles playing against each other, the movement trajectory of pedestrians, the movement trajectory of bicycles, and the dynamic trajectory of non-motorized vehicles.

12. The method according to any one of claims 4 or 5, characterized in that, Also includes: Obtain the density values ​​of objects within a preset range around the vehicle; When the object density value is greater than the first threshold, the sampling point density of the half-contour sampling is increased.

13. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-12.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-12.

Images