Ground-air obstacle avoidance method guided by time-varying vector field

Abstract

The invention provides a ground-air obstacle avoidance method guided by a time-varying vector field, and the method comprises the steps: carrying out the clustering of obstacles based on the difference between the motion feature information of all obstacles at any moment, and determining the types of the obstacles based on a clustering result; based on the obstacle type of the obstacle and the movement feature information of the obstacle, determining the threat degree of the obstacle to the position of the mobile intelligent agent at any moment; based on the threat degree of each obstacle, a global time-varying vector field is determined, a navigation vector field of the mobile agent is determined based on the global time-varying vector field, and the global time-varying vector field represents the comprehensive risk of the position of the mobile agent at any moment; and carrying out obstacle avoidance path planning based on the navigation vector field of the mobile intelligent agent. According to the method provided by the invention, the motion features of different types of obstacles are mapped to a unified feature vector field space in a dimensionality reduction manner, the obstacle avoidance problem in a complex dynamic environment can be efficiently solved, and the safe operation of the mobile intelligent agent is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic obstacle avoidance, and particularly to a ground-air obstacle avoidance method guided by a time-varying vector field. Background Art

[0002] In rescue and industrial scenarios, mobile robots often need to face complex and changing environments, and may encounter projectile obstacles falling from high places and dynamic obstacles on the ground simultaneously. In this case, how to design a reasonable and efficient obstacle avoidance strategy to ensure the safe operation of mobile robots is the focus and difficulty of current research.

[0003] Facing a complex dynamic environment, it is crucial to design an efficient obstacle avoidance strategy. Global path planning methods such as A* and D* can perform global optimal planning for paths in static environments, while the rapidly-exploring random tree (RRT) and its variants based on random sampling are suitable for dealing with path planning problems in high-dimensional complex spaces. Secondly, local planning methods such as the dynamic window approach (DWA) and the artificial potential field method (APF) pay more attention to real-time performance and can quickly generate obstacle avoidance trajectories, but are suitable for local adjustments in simple dynamic environments.

[0004] However, many traditional algorithms lack the ability to quickly adapt to complex dynamic environments. Especially in scenarios where the states of obstacles change frequently, the planning results may lag behind the changes in the actual environment. For example, when DWA and APF face large-scale dynamic obstacles or obstacles moving at high speeds, their obstacle avoidance performance will significantly decline; moreover, existing research mainly focuses on the obstacle avoidance problem of ground dynamic obstacles, and there is little research on the handling of aerial obstacles. This is mainly because the dynamic characteristics of aerial obstacles (such as projectile motion or flying obstacles) are more complex, and their threat ranges and motion trajectories are difficult to accurately model.

[0005] When dealing with complex scenarios where ground-air obstacles coexist, the high-dimensional and redundant information in the feature space will lead to a significant increase in computational complexity. For example, when APF performs path planning in three-dimensional space-time, calculating the potential energy of each point requires a large amount of computational resources, resulting in a decrease in real-time performance. As the number and dynamic changes of obstacles increase, it is difficult for APF to balance real-time performance and computational resources. Summary of the Invention

[0006] The present invention provides a ground-air obstacle avoidance method guided by a time-varying vector field to solve the defects of high computational complexity and poor real-time performance in the prior art when dealing with complex scenarios where ground-air obstacles coexist, and to achieve efficient and real-time processing of obstacle avoidance problems in complex dynamic environments, ensuring the safe operation of mobile intelligent agents. The present invention provides a ground-air obstacle avoidance method guided by a time-varying vector field, including: Cluster the obstacles based on the differences between the motion characteristic information of each obstacle at any moment, and determine the obstacle type of the obstacle based on the clustering result; Based on the obstacle type of the obstacle and the motion characteristic information of the obstacle, determine the threat degree of the obstacle to the position of the mobile agent at any moment; Based on the threat degrees of each obstacle, determine a global time-varying vector field, and determine a navigation vector field of the mobile agent based on the global time-varying vector field. The global time-varying vector field represents the comprehensive risk of the position of the mobile agent at any moment; Perform obstacle avoidance path planning based on the navigation vector field of the mobile agent.

[0007] According to the time-varying vector field-guided ground-air obstacle avoidance method provided by the present invention, the determining the threat degree of the obstacle to the position of the mobile agent at any moment based on the obstacle type of the obstacle and the motion characteristic information of the obstacle includes: Based on the obstacle type of the obstacle and the motion characteristic information of the obstacle, determine the initial threat degree of the obstacle to the position of the mobile agent at any moment; If the obstacle type of the obstacle is a ground obstacle, use the initial threat degree as the threat degree of the obstacle; If the obstacle type of the obstacle is an air obstacle, perform height compensation on the initial threat degree to obtain the threat degree of the obstacle.

[0008] According to the time-varying vector field-guided ground-air obstacle avoidance method provided by the present invention, the determining the initial threat degree of the obstacle to the position of the mobile agent at any moment based on the obstacle type of the obstacle and the motion characteristic information of the obstacle includes: Based on the posterior probability of the obstacle type to which the obstacle belongs, the priority weight of the belonging obstacle type, and the predicted position of the obstacle at any moment, determine the initial threat degree of the obstacle to the position of the mobile agent at any moment.

[0009] According to the time-varying vector field-guided ground-air obstacle avoidance method provided by the present invention, the performing height compensation on the initial threat degree to obtain the threat degree of the obstacle if the obstacle type of the obstacle is an air obstacle includes: If the obstacle type of the obstacle is an air obstacle, perform height compensation on the initial threat degree based on the vertical height difference between the mobile agent and the obstacle, and use the threat degree after height compensation as the threat degree of the obstacle.

[0010] According to the ground-air obstacle avoidance method guided by a time-varying vector field provided by the present invention, determining the navigation vector field of a mobile agent based on the global time-varying vector field includes: Based on the negative gradient direction of the global time-varying vector field, determining the obstacle avoidance repulsive force field of the mobile agent; Based on the target position of the mobile agent, determining the target attraction force field of the mobile agent; Superposing the obstacle avoidance repulsive force field of the mobile agent and the target attraction force field to obtain the navigation vector field of the mobile agent.

[0011] According to the ground-air obstacle avoidance method guided by a time-varying vector field provided by the present invention, performing obstacle avoidance path planning based on the navigation vector field of the mobile agent includes: Based on the motion constraints of the mobile agent, iteratively moving along the direction of the navigation vector field to obtain a discrete path point sequence; Extracting the obstacle avoidance key points in the discrete path point sequence to obtain the control points of a spline curve; Initializing a spline curve based on the control points of the spline curve to obtain an initialized spline curve; Optimizing the initialized spline curve to obtain the obstacle avoidance path of the mobile agent.

[0012] According to the ground-air obstacle avoidance method guided by a time-varying vector field provided by the present invention, optimizing the initialized spline curve to obtain the obstacle avoidance path of the mobile agent includes: Based on at least one of the path smoothness, path length, obstacle avoidance penalty term, and curvature constraint of the initialized spline curve, determining an optimization objective function; Optimizing the initialized spline curve based on the optimization objective function to obtain the obstacle avoidance path of the mobile agent.

[0013] The present invention also provides a ground-air obstacle avoidance device guided by a time-varying vector field, including: A clustering unit, configured to cluster the obstacles based on the differences between the motion feature information of each obstacle at any moment, and determine the obstacle type of the obstacles based on the clustering result; A threat degree determination unit, configured to determine the threat degree of the obstacles to the position of the mobile agent at any moment based on the obstacle type of the obstacles and the motion feature information of the obstacles; A navigation vector field determination unit, configured to determine a global time-varying vector field based on the threat degrees of the respective obstacles, and determine the navigation vector field of the mobile agent based on the global time-varying vector field, where the global time-varying vector field characterizes the comprehensive risk of the position of the mobile agent at any moment; A path planning unit for performing obstacle avoidance path planning based on the navigation vector field of the mobile agent.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method for avoiding ground and air obstacles guided by the time-varying vector field as described in any one of the above.

[0015] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for avoiding ground and air obstacles guided by the time-varying vector field as described in any one of the above.

[0016] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method for avoiding ground and air obstacles guided by the time-varying vector field as described in any one of the above.

[0017] The method for avoiding ground and air obstacles guided by the time-varying vector field provided by the present invention models the differences between the motion characteristic information of each obstacle, quantifies the matching degree between the obstacle and each clustering center, thereby identifying the type of each obstacle, and can effectively unify the threat levels of air obstacles and ground obstacles, realizing efficient modeling of a complex dynamic environment. Secondly, by reducing the dimension of the motion characteristics of different types of obstacles and mapping them to a unified feature vector field space, the computational complexity is significantly reduced, enabling the algorithm to achieve real-time planning with limited computational resources. This method can efficiently handle the obstacle avoidance problem in a complex dynamic environment and ensure the safe operation of the mobile robot. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic flowchart of the method for avoiding ground and air obstacles guided by the time-varying vector field provided by the present invention.

[0020] Figure 2 It is a schematic diagram of a complex multi-obstacle scenario provided by the present invention.

[0021] Figure 3 It is a schematic diagram of the differences in different motion modes provided by the present invention.

[0022] Figure 4It is the threat degree distribution diagram of different time steps provided by the present invention.

[0023] Figure 5 It is the effect diagram of the obstacle avoidance path of the B-spline curve provided by the present invention.

[0024] Figure 6 It is the structural schematic diagram of the ground-air obstacle avoidance device guided by the time-varying vector field provided by the present invention.

[0025] Figure 7 It is the structural schematic diagram of the electronic device provided by the present invention. Specific embodiments

[0026] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Aiming at the problems of high computational complexity, poor real-time performance, and poor path quality in dealing with complex scenarios where ground-air obstacles coexist in the related art, the embodiments of the present invention propose a ground-air obstacle avoidance method guided by a time-varying vector field. In this method, by modeling the differences between the motion feature information of each obstacle, the matching degree between the obstacle and each clustering center is quantified, so as to identify the types of each obstacle, and thus the threat degrees of air obstacles and ground obstacles can be effectively expressed uniformly, realizing efficient modeling of complex dynamic environments; secondly, by reducing the dimension of the motion features of different types of obstacles and mapping them to a unified feature vector field space, the computational complexity is significantly reduced, enabling the algorithm to achieve real-time planning with limited computational resources. This method can efficiently handle the obstacle avoidance problem in complex dynamic environments and ensure the safe operation of mobile robots.

[0028] The embodiments of the present invention can be applied to obstacle avoidance scenarios in dynamic complex environments. For example, they can be widely applied to dynamic obstacle avoidance tasks in fields such as industrial robots, rescue robots, and driverless vehicles, which is of great significance for improving the safety and reliability of mobile robots in complex dynamic environments. The execution subject of this method can be an electronic device such as a terminal device (such as a mobile intelligent agent), a computer, a server, a server cluster, or a specially designed obstacle avoidance device, or an obstacle avoidance device set in the electronic device, and the obstacle avoidance device can be implemented by software, hardware, or a combination of both.

[0029] In the description of the embodiments of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the embodiments of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0030] Figure 1 FIG. is a schematic flowchart of a ground-air obstacle avoidance method guided by a time-varying vector field provided by the present invention. As Figure 1 shown, taking the application of this method to a mobile intelligent agent as an example, the mobile intelligent agent can be, for example, a wheeled mobile robot. This method includes the following steps: Step 110: Cluster the obstacles based on the differences between the motion feature information of each obstacle at any moment, and determine the obstacle type based on the clustering result. Step 120: Determine the threat degree of the obstacle to the position of the mobile intelligent agent at this moment based on the obstacle type and the motion feature information of the obstacle. Step 130: Determine the global time-varying vector field based on the threat degrees of each obstacle. The global time-varying vector field represents the comprehensive risk of the position of the mobile intelligent agent at this moment, and determine the navigation vector field of the mobile intelligent agent based on the global time-varying vector field. Step 140: Perform obstacle avoidance path planning based on the navigation vector field of the mobile intelligent agent.

[0031] Specifically, taking a wheeled mobile robot as the carrier and installing sensors on its top to sense the different obstacle feature information in the robot coordinate system. These obstacles include static ground obstacles (such as houses), dynamic ground obstacles (such as pedestrians and vehicles), and aerial obstacles (such as a football flying in the air), thus forming a Figure 2 complex multi-obstacle scenario as shown in Figure 2 FIG. is a schematic diagram of the complex multi-obstacle scenario provided by the present invention.

[0032] Through the sensor sensing data, obtain the motion feature information of each obstacle at any moment. The motion feature information may include the position, speed, acceleration, etc. of the obstacle. In addition, the sensor can also collect basic information such as time stamps, obstacle types, obstacle IDs, etc., thereby constituting a feature vector set of the obstacles , where represents the th obstacle at time 's -dimensional feature vector.

[0033] After obtaining the motion feature information of each obstacle, a Gaussian mixture model (GMM) can be constructed. Through the linear combination of multivariate Gaussian distributions, the diversity of the obstacle motion patterns is modeled. Figure 3 It is a schematic diagram of the differences in different motion patterns provided by the present invention.

[0034] The probability density function of GMM can be expressed as: (1) where are the model parameters, including the mixing coefficient , the mean vector , the covariance matrix , is the preset number of clusters.

[0035] The online expectation maximization (EM) algorithm is used to dynamically update the GMM parameters, specifically including: (1) E-step: Calculate the posterior probability , that is, the probability that the th obstacle belongs to the th class. Its role is to quantify the matching degree between the obstacle and each cluster center and provide weights for parameter update.

[0036] (2) (2) M-step: Introduce the time decay factor , whose role is to improve the adaptability of the model to the dynamic environment and avoid historical data from interfering with the current plan, and update the parameters: (3) (4) (5) The role is to reduce the weight of historical data, make the model pay more attention to recent observations, and adapt to the changes in the dynamic environment.

[0037] The clustering result provides obstacle type labels for the subsequent threat degree mapping. The obstacle types include ground static, ground dynamic, or aerial obstacles.

[0038] After obtaining the obstacle types, spatio-temporal threat degree mapping can be carried out. Reducing the dimensionality of the motion features of different types of obstacles and mapping them to a unified feature vector field space can effectively unify the expression of the threat degrees of aerial obstacles and ground obstacles.

[0039] Here, the threat degrees of obstacles with different motion patterns and different types to the position of the mobile intelligent agent are different.

[0040] For the threat level of a single obstacle, first, based on the posterior probability of the obstacle type to which the obstacle belongs, the priority weight of the obstacle type, and the predicted position of the obstacle at that moment, determine the initial threat level of the obstacle to the position of the mobile agent at that moment.

[0041] For the th obstacle , at time the initial threat level to the position of the mobile robot is: (6) represents the priority weight of the obstacle type. For example, an airborne obstacle has a higher priority than a ground dynamic obstacle; represents the predicted position of the obstacle at time . For a ground obstacle, it is a two-dimensional coordinate, and for an airborne obstacle, it is a three-dimensional coordinate;

[0042] If the obstacle type of the obstacle is a ground obstacle, then take the initial threat level as the threat level of the obstacle. The threat levels of ground stationary obstacles and ground moving obstacles can both be obtained through formula (6).

[0043] If the obstacle type of the obstacle is an airborne obstacle, then perform height compensation on the initial threat level. For height compensation, it can be achieved by performing height compensation on the initial threat level based on the vertical height difference between the mobile agent and the obstacle, and then take the threat level after height compensation as the threat level of the airborne obstacle.

[0044] For the processing of height compensation for airborne obstacles, if is an airborne obstacle, its threat level needs to be corrected according to the height diffusion characteristics. The threat level after height compensation can be expressed as (7) where represents the vertical height difference between the mobile robot and the airborne obstacle, represents the threat diffusion coefficient in the height direction, which controls the speed of threat height attenuation (the larger the value, the wider the threat range).

[0045] After obtaining the threat level of each single obstacle to the position of the mobile agent at that moment, the global time-varying vector field can be determined. Here, the global time-varying vector field characterizes the comprehensive risk of the position of the mobile agent at that moment.

[0046] The global time-varying vector field function can be expressed as: (8) where is the time decay coefficient, which is used to reduce the weight of long-term prediction. According to the maximum movement speed of the carrier Dynamic adjustment: (9) where is the preset safety distance threshold.

[0047] For determining the navigation vector field of a mobile agent based on a global time-varying vector field, the obstacle avoidance repulsive force field of the mobile agent can be determined through the negative gradient direction of the global time-varying vector field; the target attraction force field of the mobile agent can be determined based on the target position of the mobile agent; and the navigation vector field of the mobile agent can be obtained by superimposing the obstacle avoidance repulsive force field and the target attraction force field of the mobile agent.

[0048] (1) Calculation of the obstacle avoidance repulsive force field. The negative gradient direction of the global time-varying vector field is the obstacle avoidance guiding vector: (10) Gradient analytical formula (two-dimensional scenario): (11) It indicates that the gradient direction points to the path with the fastest decline of the global time-varying vector field, and the negative gradient guides the robot away from high-risk areas.

[0049] (2) Calculation of the target attraction force field. The target attraction force field guides the robot to move towards the target point Move: (12) where is the attraction gain coefficient, which is adaptively adjusted according to the distance between the mobile robot and the target point (the farther the distance, the greater the gain).

[0050] (3) Generation of the synthetic vector field. The obstacle avoidance repulsive force field and the target point attraction force field are superimposed to obtain the final navigation vector field: (13) The direction of the synthetic vector field indicates the movement direction of the mobile robot, and the intensity reflects the balance between the environmental risk and the proximity to the target, providing a continuous direction guidance for path planning. Figure 4 is the threat degree distribution diagram at different time steps provided by the present invention. Figure 4 shows the threat degree distribution at different time steps ( to ). The red represents static ground obstacles, the green represents the trajectory of dynamic ground obstacles, the blue represents aerial obstacles, and the yellow pentagram represents the target point.

[0051] Based on obtaining the navigation vector field, obstacle avoidance path planning is carried out based on the navigation vector field of the mobile agent. Step 140 specifically includes: Step 141, iteratively move along the direction of the navigation vector field based on the motion constraints of the mobile agent to obtain a discrete path point sequence; Step 142, extract the obstacle avoidance key points in the discrete path point sequence to obtain the control points of the spline curve; Step 143, initialize the spline curve based on the control points of the spline curve to obtain an initialized spline curve; Step 144, optimize the initialized spline curve to obtain the obstacle avoidance path of the mobile agent.

[0052] Specifically, starting from the starting point and moving iteratively along the direction of the composite vector field with the step size determined by the motion constraints of the mobile robot: (14) represents the maximum motion speed of the mobile robot, a preset maximum step size to prevent the path point spacing from being too large, and the iterative formula: (15) Generate a discrete path point sequence , where is close to the target point, and the termination condition is reaching the target point or exceeding the maximum number of iterations .

[0053] Convert the path point sequence into B-spline control points , retaining the obstacle avoidance characteristics and reducing redundancy. Use the Douglas-Peucker algorithm to eliminate redundant points and retain the curvature mutation points (such as obstacle avoidance inflection points and boundary approach points). Thus, the obstacle avoidance key points in the discrete path point sequence are extracted to obtain the control points of the spline curve.

[0054] The distance from a point to a line in the Douglas-Peucker algorithm is: (16) where represents the line segment endpoints, represents the point to be evaluated, represents the vector cross product, represents the vector norm, calculating the maximum vertical distance of all points on the line segment ; if , If the tolerance is preset, keep the point and split the line segment, otherwise discard the middle point and only keep the endpoint.

[0055] Then the spline curve is initialized based on the control points of the spline curve to obtain the initialized spline curve. Control Points and a node vector , connecting these control points in sequence can form Step ( The expression of the B-spline curve satisfies . is defined at the node vector On indivual Sub-B-spline basis functions.

[0056] Any point on the B-spline curve The expression is: (17) The calculation of B-spline basis functions is recursive. When the initial condition is: (18) when When , the recursive calculation formula is: (19) After obtaining the initialization spline curve, the initialization spline curve is optimized to obtain the obstacle avoidance path of the mobile agent. In some embodiments, step 144 specifically includes: Determine an optimization objective function based on at least one of path smoothness, path length, obstacle avoidance penalty, and curvature constraint of the initialized spline curve; The initialization spline curve is optimized based on the optimization objective function to obtain the obstacle avoidance path of the mobile agent.

[0057] Specifically, the overall optimization objective function consists of four weighted sub-objectives, and its mathematical form is: (20) is the weight, and the weight satisfies , dynamically adjusted according to the scene, such as dynamic environment needs to be increased , narrow areas need to be enlarged , They correspond to the cost functions of path smoothness, path length, obstacle avoidance penalty, and curvature constraint respectively. The sub-goals are defined as follows: (1) Path smoothness is to minimize the curvature change rate, ensure that the path is continuous and smooth, and avoid jitter or sudden turns of the robot due to sudden changes in the path.

[0058] (21) The second derivative of the B-spline reflects the curvature change rate, where is the parametric expression of the B-spline curve, and

[0059] (22) (2)The path length is to shorten the total path length and reduce the energy consumption and travel time of the mobile robot, where is the discretized path point, and (23) (3)The obstacle avoidance penalty term is to penalize the situation where the distance between the path point and the obstacle is less than the preset safety distance to ensure that the path is far from the threat area, where represents the total number of obstacles, is the th obstacle position. The obstacle avoidance penalty term is expressed by the formula: (24) (4)The curvature constraint is to limit the path curvature not to exceed the steering ability of the robot and avoid the situation of being unable to track the path, where is the curvature of the path at , is the maximum allowable curvature of the mobile robot, which is determined by the kinematic constraints. The curvature constraint is expressed by the formula: (25) where the curvature calculation formula is: (26) The above multi-objective optimization problem can be solved by advanced algorithms such as genetic algorithm, particle swarm optimization algorithm, L-BFGS, etc. Figure 5 is the obstacle avoidance path effect diagram of the B-spline curve provided by the present invention. As Figure 5 shown, the red is the static ground obstacle, the green is the trajectory of the dynamic ground obstacle, the blue is the aerial obstacle, and the yellow pentagram is the target point.

[0060] Based on any of the above embodiments, a ground-air obstacle avoidance method guided by a time-varying vector field is provided. The method includes: S1, constructing a time-varying vector field. S1 specifically includes: S11. Capture the differences in motion patterns through the Gaussian mixture model. Based on the differences between the motion feature information of each obstacle at any moment, cluster the obstacles, and determine the obstacle types of the obstacles based on the clustering results.

[0061] S12. Spatiotemporal threat degree mapping. Based on the obstacle types of the obstacles and the motion feature information of the obstacles, determine the threat degree of the obstacles to the position of the mobile agent at any moment. S12 specifically includes: Determine the initial threat degree of the obstacle to the position of the mobile agent at any moment based on the posterior probability of the obstacle type to which the obstacle belongs, the priority weight of the obstacle type to which it belongs, and the predicted position of the obstacle at any moment; if the obstacle type of the obstacle is a ground obstacle, then use the initial threat degree as the threat degree of the obstacle; if the obstacle type of the obstacle is an aerial obstacle, then based on the vertical height difference between the mobile agent and the obstacle, perform height compensation on the initial threat degree, and use the threat degree after height compensation as the threat degree of the obstacle.

[0062] S13. Generate the global time-varying vector field. Based on the threat degrees of each obstacle, determine the global time-varying vector field.

[0063] S14. Determine the navigation vector field of the mobile agent based on the global time-varying vector field, that is, calculate the gradient vector field. S14 specifically includes: Determine the obstacle avoidance repulsive force field of the mobile agent based on the negative gradient direction of the global time-varying vector field; determine the target attraction force field of the mobile agent based on the target position of the mobile agent; superimpose the obstacle avoidance repulsive force field of the mobile agent and the target attraction force field to obtain the navigation vector field of the mobile agent.

[0064] S2. Based on the navigation vector field of the mobile agent, perform obstacle avoidance path planning. S2 specifically includes: S21. Generate a sequence of path points. Based on the motion constraints of the mobile agent, iteratively move along the direction of the navigation vector field to obtain a discrete sequence of path points.

[0065] S22. Extract key points. Extract the obstacle avoidance key points in the discrete sequence of path points to obtain the control points of the spline curve.

[0066] S23. B-spline initialization. Initialize the spline curve based on the control points of the spline curve to obtain the initialized spline curve.

[0067] S23. Multi-objective optimization. Optimize the initialized spline curve to obtain the obstacle avoidance path of the mobile agent. S23 specifically includes: Determine the optimization objective function based on at least one of the path smoothness, path length, obstacle avoidance penalty term, and curvature constraint of the initialized spline curve; optimize the initialized spline curve based on the optimization objective function to obtain the obstacle avoidance path of the mobile agent.

[0068] The experimental results show that, compared with the prior art, the present invention has the following advantages: (1) High computational efficiency: In a complex noise environment, by analyzing the motion feature information of obstacles and mapping it to a unified threat degree feature vector field, effective dimensionality reduction is achieved, reducing the data dimensions that need to be processed, enabling the path planning algorithm to complete calculations in a shorter time.

[0069] (2) Strong adaptability: The time-varying vector field can quickly respond to environmental changes, identify ground and air obstacles with various different motion patterns, and dynamically adjust the path to ensure real-time performance and effectiveness in a complex environment.

[0070] (3) Good path quality: By combining B-spline curves and multi-objective comprehensive optimization, the smoothness and continuity of the path are ensured, meeting the motion constraints of the mobile robot, and improving the overall quality and reliability of the path.

[0071] The method provided by the embodiment of the present invention proposes a method for constructing a time-varying vector field for extracting obstacle features through motion pattern differences, which can effectively unify the expression of the threat degrees of air obstacles and ground obstacles, and realizes efficient modeling of complex dynamic environments; secondly, by reducing the dimensionality of the motion features of different types of obstacles and mapping them to a unified feature vector field space, the computational complexity is significantly reduced, enabling the algorithm to achieve real-time planning under limited computational resources; thirdly, by combining B-spline curves with multi-objective optimization, both the smoothness and feasibility of the path are ensured, and rapid optimization and adjustment of the path are realized. This method can efficiently handle the obstacle avoidance problem in complex dynamic environments and ensure the safe operation of the mobile robot.

[0072] The following describes the ground-air obstacle avoidance device guided by the time-varying vector field provided by the present invention. The ground-air obstacle avoidance device guided by the time-varying vector field described below can be mutually referred to the time-varying vector field-guided ground-air obstacle avoidance method described above.

[0073] Based on any of the above embodiments, Figure 6 is a schematic structural diagram of the ground-air obstacle avoidance device guided by the time-varying vector field provided by the present invention, as Figure 6 shown, the device includes: A clustering unit 610, configured to cluster the obstacles based on the differences between the motion feature information of each obstacle at any moment, and determine the obstacle type of the obstacles based on the clustering result; A threat degree determination unit 620, configured to determine the threat degree of the obstacle to the position of the mobile intelligent body at any moment based on the obstacle type of the obstacle and the motion feature information of the obstacle; A navigation vector field determination unit 630, configured to determine a global time-varying vector field based on the threat levels of the respective obstacles, where the global time-varying vector field characterizes the comprehensive risk of the position of the mobile intelligent agent at any moment, and determine the navigation vector field of the mobile intelligent agent based on the global time-varying vector field; A path planning unit 640, configured to perform obstacle avoidance path planning based on the navigation vector field of the mobile intelligent agent.

[0074] The device provided by the embodiment of the present invention models the differences between the motion feature information of each obstacle, quantifies the matching degree between the obstacle and each cluster center, thereby identifying the types of each obstacle, and can effectively unify the expression of the threat levels of aerial obstacles and ground obstacles, realizing efficient modeling of a complex dynamic environment; secondly, by reducing the dimension of the motion features of different types of obstacles and mapping them to a unified feature vector field space, the computational complexity is significantly reduced, enabling the algorithm to achieve real-time planning with limited computational resources. The device can efficiently handle the obstacle avoidance problem in a complex dynamic environment and ensure the safe operation of the mobile robot.

[0075] Based on any of the above embodiments, the threat level determination unit is specifically configured to: Determine the initial threat level of the obstacle to the position of the mobile intelligent agent at any moment based on the obstacle type of the obstacle and the motion feature information of the obstacle; If the obstacle type of the obstacle is a ground obstacle, use the initial threat level as the threat level of the obstacle; If the obstacle type of the obstacle is an aerial obstacle, perform altitude compensation on the initial threat level to obtain the threat level of the obstacle.

[0076] Based on any of the above embodiments, the threat level determination unit is specifically configured to: Determine the initial threat level of the obstacle to the position of the mobile intelligent agent at any moment based on the posterior probability of the obstacle type to which the obstacle belongs, the priority weight of the belonging obstacle type, and the predicted position of the obstacle at any moment.

[0077] Based on any of the above embodiments, the threat level determination unit is specifically configured to: If the obstacle type of the obstacle is an aerial obstacle, perform altitude compensation on the initial threat level based on the vertical altitude difference between the mobile intelligent agent and the obstacle, and use the threat level after altitude compensation as the threat level of the obstacle.

[0078] Based on any of the above embodiments, the navigation vector field determination unit is specifically configured to: Determine the obstacle avoidance repulsive force field of the mobile intelligent agent based on the negative gradient direction of the global time-varying vector field; Determine the target attraction field of the mobile agent based on the target position of the mobile agent; Superimpose the obstacle avoidance repulsive field of the mobile agent and the target attraction field to obtain the navigation vector field of the mobile agent.

[0079] Based on any of the above embodiments, the path planning unit is specifically configured to: Iteratively move along the direction of the navigation vector field based on the motion constraints of the mobile agent to obtain a discrete path point sequence; Extract the obstacle avoidance key points in the discrete path point sequence to obtain the control points of the spline curve; Perform spline curve initialization based on the control points of the spline curve to obtain an initialized spline curve; Optimize the initialized spline curve to obtain the obstacle avoidance path of the mobile agent.

[0080] Based on any of the above embodiments, the path planning unit is specifically configured to: Determine an optimization objective function based on at least one of the path smoothness, path length, obstacle avoidance penalty term, and curvature constraint of the initialized spline curve; Optimize the initialized spline curve based on the optimization objective function to obtain the obstacle avoidance path of the mobile agent.

[0081] Figure 7 An example of the physical structure diagram of an electronic device is shown in Figure 7 As shown, the electronic device may include: a processor (processor) 710, a communication interface (Communications Interface) 720, a memory (memory) 730, and a communication bus 740. Among them, the processor 710, the communication interface 720, and the memory 730 complete mutual communication through the communication bus 740. The processor 710 can call the logical instructions in the memory 730 to execute the ground-air obstacle avoidance method guided by the time-varying vector field. The method includes: clustering the obstacles based on the differences between the motion feature information of each obstacle at any moment, and determining the obstacle type of the obstacles based on the clustering result; determining the threat degree of the obstacles to the position of the mobile agent at any moment based on the obstacle type of the obstacles and the motion feature information of the obstacles; determining the global time-varying vector field based on the threat degrees of the respective obstacles, where the global time-varying vector field represents the comprehensive risk of the position of the mobile agent at any moment, and determining the navigation vector field of the mobile agent based on the global time-varying vector field; performing obstacle avoidance path planning based on the navigation vector field of the mobile agent.

[0082] In addition, when the logical instructions in the above-mentioned memory 730 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, 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. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0083] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the time-varying vector field-guided ground-air obstacle avoidance method provided by the above-mentioned various methods. The method includes: clustering the obstacles based on the differences between the motion characteristic information of each obstacle at any moment, and determining the obstacle type of the obstacles based on the clustering result; determining the threat degree of the obstacles to the position of the mobile agent at any moment based on the obstacle type of the obstacles and the motion characteristic information of the obstacles; determining a global time-varying vector field based on the threat degrees of each of the obstacles, where the global time-varying vector field represents the comprehensive risk of the position of the mobile agent at any moment, and determining a navigation vector field of the mobile agent based on the global time-varying vector field; and performing obstacle avoidance path planning based on the navigation vector field of the mobile agent.

[0084] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the time-varying vector field-guided ground-air obstacle avoidance method provided by the above-mentioned various methods. The method includes: clustering the obstacles based on the differences between the motion characteristic information of each obstacle at any moment, and determining the obstacle type of the obstacles based on the clustering result; determining the threat degree of the obstacles to the position of the mobile agent at any moment based on the obstacle type of the obstacles and the motion characteristic information of the obstacles; determining a global time-varying vector field based on the threat degrees of each of the obstacles, where the global time-varying vector field represents the comprehensive risk of the position of the mobile agent at any moment, and determining a navigation vector field of the mobile agent based on the global time-varying vector field; and performing obstacle avoidance path planning based on the navigation vector field of the mobile agent.

[0085] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0086] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A ground-to-air obstacle avoidance method guided by a time-varying vector field, characterized in that: include: Clustering the obstacles based on the differences between the motion feature information of the obstacles at any moment, and determining the obstacle types of the obstacles based on the clustering results; Determine, based on the obstacle type of the obstacle and the motion characteristic information of the obstacle, the threat degree of the obstacle to the position of the mobile agent at any time; Determining a global time-varying vector field based on the threat level of each obstacle, and determining a navigation vector field of the mobile agent based on the global time-varying vector field, wherein the global time-varying vector field represents a comprehensive risk of the position of the mobile agent at any time; Obstacle avoidance path planning is performed based on the navigation vector field of the mobile agent.

2. The ground-to-air obstacle avoidance method guided by a time-varying vector field according to claim 1, characterized in that: The determining, based on the obstacle type of the obstacle and the motion feature information of the obstacle, the threat degree of the obstacle to the position of the mobile agent at any time, includes: Determine, based on the obstacle type of the obstacle and the motion characteristic information of the obstacle, the initial threat degree of the obstacle to the position of the mobile agent at any time; If the obstacle type of the obstacle is a ground obstacle, the initial threat level is used as the threat level of the obstacle; If the obstacle type of the obstacle is an aerial obstacle, the initial threat level is height compensated to obtain the threat level of the obstacle.

3. The ground-to-air obstacle avoidance method guided by a time-varying vector field according to claim 2, characterized in that: The determining, based on the obstacle type of the obstacle and the motion feature information of the obstacle, the initial threat degree of the obstacle to the position of the mobile agent at any time, includes: Based on the posterior probability of the obstacle type to which the obstacle belongs, the priority weight of the obstacle type, and the predicted position of the obstacle at any time, the initial threat degree of the obstacle to the position of the mobile agent at any time is determined.

4. The ground-to-air obstacle avoidance method guided by a time-varying vector field according to claim 2, characterized in that: If the obstacle type of the obstacle is an aerial obstacle, performing height compensation on the initial threat level to obtain the threat level of the obstacle includes: If the obstacle type of the obstacle is an aerial obstacle, the initial threat level is height compensated based on the vertical height difference between the mobile agent and the obstacle, and the threat level after height compensation is used as the threat level of the obstacle.

5. The method for avoiding ground-to-air obstacles guided by a time-varying vector field according to claim 1, characterized in that: The determining of the navigation vector field of the mobile agent based on the global time-varying vector field comprises: Determining an obstacle avoidance repulsion field of the mobile agent based on the negative gradient direction of the global time-varying vector field; Determining a target attraction field of the mobile agent based on the target position of the mobile agent; The obstacle avoidance repulsion field of the mobile agent is superimposed on the target attraction field to obtain a navigation vector field of the mobile agent.

6. The method for avoiding ground-to-air obstacles guided by a time-varying vector field according to any one of claims 1 to 5, characterized in that: The obstacle avoidance path planning based on the navigation vector field of the mobile agent includes: Based on the motion constraints of the mobile agent, iteratively move along the direction of the navigation vector field to obtain a discrete path point sequence; Extracting key obstacle avoidance points in the discrete path point sequence to obtain control points of the spline curve; Initializing the spline curve based on the control points of the spline curve to obtain an initialized spline curve; The initialized spline curve is optimized to obtain an obstacle avoidance path for the mobile agent.

7. The method for avoiding ground-to-air obstacles guided by a time-varying vector field according to claim 6, characterized in that: The step of optimizing the initialization spline curve to obtain an obstacle avoidance path for the mobile agent includes: Determining an optimization objective function based on at least one of path smoothness, path length, obstacle avoidance penalty term, and curvature constraint of the initialized spline curve; The initialization spline curve is optimized based on the optimization objective function to obtain an obstacle avoidance path for the mobile agent.

8. A ground-to-air obstacle avoidance device guided by a time-varying vector field, characterized in that: include: A clustering unit, configured to cluster the obstacles based on the difference between the motion feature information of the obstacles at any moment, and determine the obstacle type of the obstacles based on the clustering result; A threat level determination unit, configured to determine the threat level of the obstacle to the position of the mobile agent at any time based on the obstacle type of the obstacle and the motion feature information of the obstacle; A navigation vector field determination unit, configured to determine a global time-varying vector field based on the threat level of each obstacle, and determine a navigation vector field of the mobile agent based on the global time-varying vector field, wherein the global time-varying vector field represents a comprehensive risk of the position of the mobile agent at any time; The path planning unit is used to perform obstacle avoidance path planning based on the navigation vector field of the mobile agent.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the ground-to-air obstacle avoidance method guided by the time-varying vector field as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the ground-to-air obstacle avoidance method guided by a time-varying vector field as described in any one of claims 1 to 7 is implemented.

Images