An obstacle avoidance trajectory parameter multiplexing method based on local environment geometry and motion constraint features
By constructing a local environmental feature descriptor and combining laser sensing and pose optimization results, the problems of initial condition sensitivity and poor adaptability of historical trajectory reuse in autonomous motion planning in complex environments are solved, thereby improving the stability and efficiency of trajectory planning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG UNIV OF TECH
- Filing Date
- 2026-03-27
- Publication Date
- 2026-06-02
AI Technical Summary
Existing autonomous motion planning methods are sensitive to initial conditions in complex environments, have limited environmental representation methods, poor adaptability to historical trajectory reuse, and fragmented information between the positioning and planning modules, leading to planning failures or decreased computational efficiency.
By constructing a multi-level descriptor of local environmental geometry and motion constraint features, and combining laser sensing and pose optimization results, environmental features are extracted and bound to historical trajectory parameters for use as initial conditions and optimization guidance for trajectory planning.
It improves the stability and success rate of trajectory planning, reduces sensitivity to initial conditions, enhances the generalization ability and robustness of trajectory planning, while maintaining computational efficiency and making it easy to integrate into existing autonomous navigation systems.
Smart Images
Figure CN122130088A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of autonomous navigation and motion planning, and particularly relates to a method for reusing obstacle avoidance trajectory parameters based on local environment geometry and motion constraint features. BACKGROUND
[0002] With the wide application of unmanned aerial vehicles, mobile robots and autonomous driving vehicles in complex environments, autonomous motion planning and obstacle avoidance technology has become one of the key technologies to ensure system safety and real-time performance. Especially in unknown or semi-structured environments, the vehicle needs to generate a feasible, smooth and safe motion trajectory in real time under the premise of meeting the dynamic constraints and obstacle avoidance constraints. Existing autonomous motion planning methods usually adopt an optimization-based trajectory planning framework, which solves the optimal trajectory that meets the constraint conditions by constructing a nonlinear optimization problem containing smoothness constraints, dynamic constraints and obstacle constraints. Such methods have strong generality in theory, but in practical applications, they have strong dependence on optimization initial conditions and environmental structure, especially in complex, narrow or significantly changing geometric structures, which can easily cause planning failure or reduced computational efficiency.
[0003] However, the prior art still has the following deficiencies: first, the trajectory planning is sensitive to initial conditions, and if the initial trajectory does not match the environmental structure, the optimization process can easily fall into local optimum or even fail to converge; second, the environmental representation method is single, and is mainly based on simple geometric distance or obstacle distribution, which is difficult to describe the constraint difference of the environment in different directions on the motion freedom of the vehicle; third, the historical trajectory reuse method has poor adaptability, and existing methods directly reuse the entire trajectory or path points, lack of abstraction of trajectory generation parameters, resulting in limited trajectory generalization ability; fourth, the information between the positioning and planning modules is fragmented, although a large amount of information reflecting the environmental structure is implicitly contained in the SLAM or direct positioning process, but the existing technology fails to effectively utilize these information to assist motion planning, resulting in low information utilization efficiency. SUMMARY
[0004] To solve the above technical problems, the present application provides a method for reusing obstacle avoidance trajectory parameters based on local environment geometry and motion constraint features to solve the above problems existing in the prior art.
[0005] To achieve the above purpose, the present application provides a method for reusing obstacle avoidance trajectory parameters based on local environment geometry and motion constraint features, comprising: obtaining a local point cloud to construct a local map, calculating a geometric residual error of a point to a plane based on the local map to obtain residual statistical features, calculating a plane normal distribution based on the local map to obtain structure direction features, and constructing an information matrix based on the Jacobian matrix of the direct method pose optimization to obtain motion constraint features; The residual statistical features, structural orientation features, and motion constraint features are combined into a local environment feature descriptor; The local environmental feature descriptor is bound to the trajectory generation parameters of the historical obstacle avoidance trajectory and stored in the database; Calculate the similarity between the local environmental feature descriptors of the current environment and the historical feature descriptors in the database, and select the corresponding historical trajectory generation parameters based on similarity matching; The historical trajectory generation parameters are used as the initial parameters for the current obstacle avoidance trajectory planning.
[0006] Optionally, the process of acquiring local point clouds to construct a local map includes: acquiring the point cloud collected by the lidar at the current moment, converting the point cloud to a map coordinate system based on the current pose estimation, and selecting the point cloud in a local area centered on the current carrier position and with a preset radius as the local map.
[0007] Optionally, the process of obtaining residual statistical characteristics by calculating the geometric residuals from points to a plane based on a local map includes: For each point in the local point cloud, search for the nearest neighbor set in the local map, and perform plane fitting on the nearest neighbor set based on the least squares method to obtain the local plane parameters. The algebraic distance from the point to the fitted plane is calculated based on the local plane parameters and used as the geometric residual. The geometric residuals of all points within a local region are calculated to obtain the residual mean, residual variance, and the proportion of large residuals as statistical characteristics of the residuals.
[0008] Optionally, the process of obtaining structural orientation features by calculating the planar normal distribution based on a local map includes: Construct the normal covariance matrix based on all valid plane normal vectors within the local region; The structural eigenvalues are obtained by performing eigenvalue decomposition on the normal covariance matrix. The normal concentration feature is calculated based on the ratio of the maximum structural eigenvalue to the sum of all structural eigenvalues, and is used as the structural orientation feature.
[0009] Optionally, the process of constructing the information matrix based on the Jacobian matrix of the direct pose optimization to obtain the motion constraint features includes: For each residual in the pose optimization process, calculate the Jacobian matrix with respect to the independent variables; A local information matrix is constructed based on all Jacobian matrices, and the submatrices related to the translation variables are extracted from the local information matrix. The submatrix is decomposed into eigenvalues to obtain motion eigenvalues. The number of environmental constraints is calculated based on the ratio of the maximum to the minimum motion eigenvalue as the motion constraint feature.
[0010] Optionally, the trajectory generation parameters of the historical obstacle avoidance trajectory include at least one of the following: polynomial trajectory control point parameters, trajectory time allocation parameters, obstacle constraint control point parameters, and collision segmentation structure parameters.
[0011] Optionally, the process of calculating the similarity between the local environmental feature descriptors of the current environment and the historical feature descriptors in the database includes: The feature distance between the current feature descriptor and the historical feature descriptors is calculated based on the preset feature weight diagonal matrix. The feature distance is compared with a preset threshold. When the feature distance is less than the preset threshold, it is determined that the environment is similar and the corresponding historical trajectory generation parameters are selected.
[0012] The present invention also provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described thereon.
[0013] The present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.
[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention overcomes the limitations of traditional methods that rely on a single geometric distance for environmental characterization by constructing a multi-level local environment feature descriptor that integrates geometric residual statistical features, structural orientation features, and motion constraint features. This achieves a precise characterization of the geometric consistency and motion constraint essence of the local environment. Based on this feature descriptor, similarity matching and reuse of historical trajectory parameters effectively reduces sensitivity to initial optimization conditions and solves the problem of optimization convergence difficulties or getting trapped in local optima due to mismatch between the initial trajectory and the environmental structure. Unlike existing methods that directly reuse the entire trajectory, this invention reuses trajectory generation parameters as initial optimization solutions or guiding information. While retaining the structural advantages of historical successful trajectories, it allows for local adjustments based on the current environment, significantly improving the generalization ability and robustness of trajectory planning. Furthermore, this method directly utilizes intermediate calculation results from laser sensing and pose optimization processes to construct features without introducing additional sensors or deep learning models. This improves planning success rate and real-time performance while maintaining controllable computational overhead, making it easy to integrate into existing autonomous navigation systems. Attached Figure Description
[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the technical solution of an embodiment of the present invention. Detailed Implementation
[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0018] Example 1 The purpose of this invention is to provide a method for reusing obstacle avoidance trajectory parameters based on local environmental features, so as to overcome the problems of sensitivity to initial conditions of trajectory planning, insufficient environmental constraint modeling, and poor adaptability of historical trajectory reuse in the prior art.
[0019] like Figure 1 As shown, this embodiment provides a method for reusing obstacle avoidance trajectory parameters based on local environmental geometry and motion constraint features. This method can extract features reflecting the environmental geometry and motion constraint characteristics from sensor data and the optimization process itself, and use them to guide trajectory planning and parameter initialization, thereby improving the stability, success rate, and real-time performance of trajectory planning. The method includes the following steps:
[0020] The process of acquiring local point clouds and constructing local maps includes: acquiring the point cloud collected by the lidar at the current moment, converting the point cloud to the map coordinate system based on the current pose estimation, and selecting the point cloud in a local area centered on the current carrier position with a preset radius as the local map.
[0021] At the present moment The point cloud acquired by the lidar is denoted as: ; in, This represents the coordinates of the i-th point in the carrier coordinate system, and N represents the total number of points. Representing a three-dimensional vector space, Let k be the point cloud acquired by the lidar at time k.
[0022] Based on current pose estimation Convert the point cloud to the map coordinate system: ; in, Let be the coordinates of the i-th point in the map coordinate system. In the map coordinate system, select a point centered at the current location of the carrier with a radius of . Local point cloud region: ; in, This represents a local point cloud region.
[0023] The process of calculating the geometric residuals from points to a plane based on a local map to obtain residual statistical features includes: searching for the nearest neighbor set for each point in the local point cloud in the local map; performing plane fitting on the nearest neighbor set using the least squares method to obtain local plane parameters; calculating the algebraic distance from the point to the fitted plane based on the local plane parameters as the geometric residual; and calculating the residual mean, residual variance, and proportion of large residuals for the geometric residuals of all points in the local region as residual statistical features. The specific implementation process is divided into the second and third steps.
[0024] The second step is to calculate the geometric residual from the point to the plane based on the local map.
[0025] For each point in the local point cloud Search for its nearest neighbor set in the local map: ; in, Let i be the set of nearest neighbors of the i-th point. Indicates the distance from the i-th point The nearest neighboring point.
[0026] Local plane parameters are obtained by fitting the nearest neighbor set using the least squares method: ; in, For unit normal vector, For planar offset, This represents a three-dimensional vector space. The geometric residual from a point to a plane is defined as: ; in, The geometric residual from the point to the plane.
[0027] The third step is to extract the statistical features of the residuals.
[0028] This embodiment includes an environmental complexity characterization based on point-to-plane geometric residual statistics. Specifically, it reflects the geometric consistency and complexity of the local environment by statistically analyzing the mean, variance, and proportion of large residuals of point-to-plane residuals in the local environment.
[0029] A set is formed for all valid residuals within the local region: ; Among them, the point-to-plane geometric residual This represents the algebraic distance from a laser point in a local environment to its local structural plane. Its magnitude reflects the degree to which the point conforms to the local planar model. When the local environment has a regular structure (such as a flat wall or ground), the residual value approaches zero; when the local environment has edges, corners, or complex obstacle structures, the residual value increases significantly.
[0030] Calculate the statistical characteristics of the residuals: (1) Mean of residuals.
[0031] ； The residual mean is used to describe the fitting bias of the overall geometric model within a local area. Its value reflects whether there is a systematic structural offset or unmodeled curvature change in the local map within the current area.
[0032] (2) Residual variance.
[0033] ; Residual variance is used to measure the consistency of geometry in a local environment. When the local area is a regular plane or a simple geometric structure, the residual distribution is concentrated and the variance is small; when the local area contains a variety of different structures (such as corners, columns, narrow passages), the residual distribution is discrete and the variance increases significantly.
[0034] (3) Large residual ratio.
[0035] ; in To preset the residual threshold, This is an indicator function. The large residual ratio is used to statistically determine the proportion of non-planar structural points within a local area, thereby characterizing the density of obstacle edges, abrupt structures, or discontinuous geometric shapes in the environment.
[0036] The above features are used to characterize the geometric complexity of the local environment.
[0037] Step 4: Calculate the plane normal distribution based on the local map to obtain the structural orientation features.
[0038] The process of obtaining structural orientation features by calculating the plane normal distribution based on a local map includes: constructing a normal covariance matrix based on all effective plane normal vectors within the local area; performing eigenvalue decomposition on the normal covariance matrix to obtain structural eigenvalues; and calculating the normal concentration feature as the structural orientation feature based on the ratio of the largest structural eigenvalue to the sum of all structural eigenvalues.
[0039] This embodiment includes environmental structural orientation feature extraction based on planar normal distribution. Specifically, by constructing the normal vector covariance matrix and analyzing its eigenvalue distribution, the consistency characteristics of the dominant structural orientation in the local environment are quantified.
[0040] Create a set of all valid plane normal vectors within the local region: ; Calculate the mean of the normal vector: ; Construct the normal covariance matrix: ; Perform eigenvalue decomposition on the covariance matrix: ; The characteristic of normal concentration is defined as follows: ; This feature characterizes the directional consistency of dominant structures in a local environment. The normal distribution covariance matrix reflects the distribution of the dominant directions of various geometric structures in the local environment. By analyzing its eigenvalues, it can be determined whether there are significant dominant structural directions in the local area. When the concentration eigenvalue is large, it indicates that the obstacle structures in the local environment have a high degree of directional consistency (such as corridors and tunnels); conversely, it indicates that the environmental structures exhibit multi-directional distribution (such as open areas or complex obstacle groups).
[0041] Step 5: Construct the information matrix based on the Jacobian matrix of the direct pose optimization to obtain the motion constraint features.
[0042] The process of constructing an information matrix based on the Jacobian matrix of pose optimization using the direct method to obtain motion constraint features includes: calculating the Jacobian matrix with respect to the independent variables for each residual in the pose optimization process; constructing a local information matrix based on all Jacobian matrices and extracting sub-matrices related to translation variables from the local information matrix; performing eigenvalue decomposition on the translation sub-matrices to obtain motion feature values; and calculating the number of environmental constraints based on the ratio of the maximum to the minimum motion feature values as motion constraint features.
[0043] This embodiment includes motion constraint observability modeling based on the direct method information matrix. Specifically, by analyzing the information matrix formed during pose optimization, sub-matrices related to translational degrees of freedom are extracted, and their eigenvalue conditions are used to quantify the strength of environmental constraints on the carrier's motion degrees of freedom.
[0044] In the direct pose optimization process, for each residual Calculate its pair as independent variables Jacobian matrix: ; Constructing a local information matrix: ; Extract the information matrix from the submatrix related to the translation variables: ; Perform eigenvalue decomposition on it: ; Define the number of environmental constraints as: ; This feature characterizes the degree of constraint imposed by the local environment on the carrier's motion degrees of freedom. The information matrix reflects the degree of observable constraint imposed by the current local environment on the carrier's motion state. By analyzing the eigenvalue distribution of its translation correlation submatrix, the strength of the environment's constraint on the motion degrees of freedom in different directions can be quantified. A large condition number indicates significant constraints in certain directions (e.g., narrow passages); a small condition number indicates weaker constraints on motion due to environmental structure (e.g., open areas).
[0045] Existing technologies often rely on point cloud density, obstacle distance, or manually defined geometric metrics to describe the environment, making it difficult to reflect the differences in the constraints imposed by the environment on the carrier's degrees of freedom of motion in different directions. This invention, based on the implementation steps one through five, analyzes the statistical characteristics of the point-to-plane geometric residuals, the distribution features of the plane normal, and the distribution of eigenvalues of the information matrix formed during pose optimization. It comprehensively models the local environment from three levels: geometric consistency, structural directionality, and motion observability, thus providing a more comprehensive and objective reflection of the inherent constraints imposed by the environment on the carrier's motion.
[0046] Step 6: Construct local environment feature vectors.
[0047] This embodiment includes a multi-level local environment feature combination modeling method. Specifically, geometric residual features, structural orientation features, and motion constraint features are combined to form a local environment feature descriptor for trajectory reuse decision-making.
[0048] Combining the above features forms a local environmental feature descriptor: ； Local environmental features include, but are not limited to, one or more combinations of the following features: statistical features based on point-to-plane geometric residuals, structural features based on plane normal distribution, and motion constraint features based on information matrices.
[0049] Step 7: Define historical trajectory parameters.
[0050] For a successfully generated obstacle avoidance trajectory without collision, its trajectory parameter set is defined as follows: ; in: These are the control point parameters for the polynomial trajectory; Assign trajectory time as a parameter; The parameters for the obstacle-constrained control points are as follows: The constraint for each control point is... , and Used as a reference point and obstacle avoidance direction; These are the parameters for the collision segmentation structure.
[0051] The trajectory parameters include, but are not limited to, trajectory polynomial control point parameters, time allocation parameters, obstacle constraint direction parameters, and obstacle avoidance trajectory segmentation structure parameters.
[0052] Step 8: Complete storage structure of historical trajectory entries.
[0053] The local environmental feature descriptors and trajectory parameters mentioned above are bound and stored in the database: ; in, For local environmental feature descriptors (residuals, normals, information matrices, etc.). The corresponding obstacle avoidance trajectory parameters.
[0054] Step 9: Calculate the similarity of environmental features.
[0055] The process of calculating the similarity between the local environmental feature descriptor of the current environment and the historical feature descriptor in the database includes: calculating the feature distance between the current feature descriptor and the historical feature descriptor based on the preset feature weight diagonal matrix; comparing the feature distance with a preset threshold, and determining that the environments are similar when the feature distance is less than the preset threshold and selecting the corresponding historical trajectory generation parameters.
[0056] The first in the historical database For each trajectory, calculate the feature distance: ; in, Feature weight diagonal matrix. The calculation method for feature distance is not limited to weighted Euclidean distance; Mahalanobis distance, normalized distance, or equivalent similarity measures can also be used.
[0057] Step 10: Matching determination.
[0058] If there exists i such that If the current environment is similar to the historical environment, select the corresponding trajectory parameters to make... .
[0059] Existing trajectory database methods typically rely on simple geometric distance, scene labels, or manually defined regions for matching, making it difficult to distinguish environments with similar geometric appearances but significantly different motion constraints. This can easily lead to the misuse of inappropriate trajectories. This invention, through steps six to ten, constructs a local environment feature descriptor by fusing geometric residual statistical features, structural orientation features, and motion constraint features. Based on this feature, environment similarity is determined, enabling more accurate differentiation of environments under different motion constraints, thereby improving the reliability of trajectory reuse.
[0060] Step 11: Methods for reusing historical trajectory parameters.
[0061] This embodiment includes a trajectory parameter-level reuse mechanism based on environmental similarity. Specifically, under similar environmental conditions, the trajectory generation parameters of historically successful obstacle avoidance trajectories are reused, rather than directly reusing the trajectories themselves, thereby improving trajectory generalization ability and planning robustness.
[0062] Historical parameters can be used as initial parameters, prior constraints, or search guidance information for planners or optimizers. The reuse of historical trajectory parameters is not limited to direct assignment as initial solutions; they can also be used as soft constraints, search space pruning conditions, or optimization direction guidance information.
[0063] Unlike existing methods that directly reuse the entire trajectory or discrete path points, this invention reuses trajectory generation parameters from historically successful obstacle avoidance trajectories, including control points, time allocation, and obstacle avoidance structure parameters. This approach retains the advantages of the original trajectory structure while allowing the planner to make local adjustments based on the current environment, significantly improving the trajectory's adaptability to environmental changes.
[0064] In complex or confined environments, nonlinear trajectory optimization problems often have multiple local extrema. Existing methods are prone to getting trapped in local optima or even failing to converge if the initial solution is not chosen properly. This invention introduces historical successful trajectory parameters as initial parameters, prior constraints, or search guidance information under similar environmental conditions, which effectively reduces the search space complexity of the optimization problem and improves the optimization convergence speed and planning success rate.
[0065] In addition, the geometric residuals, normal information and information matrix required by this invention can be directly derived from existing laser sensing and pose optimization processes, without the need to introduce additional deep learning models or additional sensors. The computational overhead is controllable and it is easy to integrate into existing autonomous navigation systems.
[0066] This invention significantly reduces the initialization difficulty of nonlinear trajectory optimization, improves the planning convergence speed, reduces the probability of getting trapped in local optima, and enhances the planning stability and real-time performance of the system in complex environments, all while maintaining the safety of trajectory planning. By reducing the number of iterations and the search space of nonlinear optimization, this invention helps to reduce the computation time for online trajectory generation while ensuring obstacle avoidance safety.
[0067] The present invention also provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described thereon.
[0068] The present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.
[0069] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for reusing obstacle avoidance trajectory parameters based on local environmental geometry and motion constraint features, characterized in that, Includes the following steps: A local map is constructed by acquiring local laser point clouds. Based on the local map, the geometric residuals from points to planes are calculated to obtain residual statistical features. Based on the local map, the plane normal distribution is calculated to obtain structural orientation features. Based on the Jacobian matrix optimized by the direct method pose, an information matrix is constructed to obtain motion constraint features. The residual statistical features, structural orientation features, and motion constraint features are combined into a local environment feature descriptor; The local environmental feature descriptor is bound to the trajectory generation parameters of the historical obstacle avoidance trajectory and stored in the database; Calculate the similarity between the local environmental feature descriptors of the current environment and the historical feature descriptors in the database, and select the corresponding historical trajectory generation parameters based on similarity matching; The historical trajectory generation parameters are used as the initial parameters for the current obstacle avoidance trajectory planning.
2. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The process of acquiring local point clouds and constructing a local map includes: acquiring the point cloud collected by the lidar at the current moment, converting the point cloud to the map coordinate system based on the current pose estimation, and selecting the point cloud in a local area centered on the current carrier position with a preset radius as the local map.
3. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The process of calculating the geometric residuals from points to a plane based on a local map to obtain the statistical characteristics of the residuals includes: For each point in the local point cloud, search for the nearest neighbor set in the local map, and perform plane fitting on the nearest neighbor set based on the least squares method to obtain the local plane parameters. The algebraic distance from the point to the fitted plane is calculated based on the local plane parameters and used as the geometric residual. The geometric residuals of all points within a local region are calculated to obtain the residual mean, residual variance, and the proportion of large residuals as statistical characteristics of the residuals.
4. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The process of obtaining structural orientation features by calculating the planar normal distribution based on a local map includes: Construct the normal covariance matrix based on all valid plane normal vectors within the local region; The structural eigenvalues are obtained by performing eigenvalue decomposition on the normal covariance matrix. The normal concentration feature is calculated based on the ratio of the maximum structural eigenvalue to the sum of all structural eigenvalues, and is used as the structural orientation feature.
5. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The process of constructing the information matrix based on the Jacobian matrix of the direct pose optimization method to obtain motion constraint features includes: For each residual in the pose optimization process, calculate the Jacobian matrix with respect to the independent variables; A local information matrix is constructed based on all Jacobian matrices, and the submatrices related to the translation variables are extracted from the local information matrix. The submatrix is decomposed into eigenvalues to obtain motion eigenvalues. The number of environmental constraints is calculated based on the ratio of the maximum to the minimum motion eigenvalue as the motion constraint feature.
6. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The trajectory generation parameters of the historical obstacle avoidance trajectory include at least one of the following: polynomial trajectory control point parameters, trajectory time allocation parameters, obstacle constraint control point parameters, and collision segmentation structure parameters.
7. The obstacle avoidance trajectory parameter reuse method based on local environmental geometry and motion constraint features according to claim 1, characterized in that, The process of calculating the similarity between the local environmental feature descriptors of the current environment and the historical feature descriptors in the database includes: The feature distance between the current feature descriptor and the historical feature descriptors is calculated based on the preset feature weight diagonal matrix. The feature distance is compared with a preset threshold. When the feature distance is less than the preset threshold, it is determined that the environment is similar and the corresponding historical trajectory generation parameters are selected.
8. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in claim 1.
9. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in claim 1.