Path planning method and device based on potential energy landscape and multi-state operator

Abstract

The application relates to the field of automation, in particular to a path planning method, device and system based on a potential energy landscape and a multi-state operator and an electronic device. The method comprises the following steps: constructing a potential energy landscape function and a corresponding potential energy gradient field of a path planning area based on environment obstacle data; calculating an operator contribution parameter corresponding to each set sampling operator based on a current search graph, calculating an operator selection probability corresponding to each set sampling operator based on the operator contribution parameter, and obtaining a set sampling operator corresponding to a maximum value of the operator selection probability as a target sampling operator; generating a target candidate sample point based on the target sampling operator, the potential energy landscape function and the potential energy gradient field, and outputting a current optimal discrete path based on the target candidate sample point. The path planning method, device and system based on the potential energy landscape and the multi-state operator and the electronic device can realize the target of giving consideration to characteristics of multiple different path planning stages and high planning efficiency.

Description

Technical Field

[0001] This application relates to the field of automation, specifically to a path planning method, apparatus and system, and electronic equipment based on potential energy landscape and multimorphic operators. Background Technology

[0002] The working environments of mobile robots, drones, and multi-degree-of-freedom robotic arms are becoming increasingly complex. The main purpose of path planning is to find an optimal collision-free feasible path from the starting point to the target point within a bounded space containing obstacles, and further generate an executable trajectory that satisfies kinematic and dynamic constraints.

[0003] Existing sampling-based motion planning methods, such as the Fast Expanded Random Tree algorithm, the Asymptotically Optimal Fast Expanded Random Tree algorithm, the Batch Informed Tree algorithm, and their improved forms, generally exhibit significant shortcomings when dealing with geometric environments containing narrow passages, cavity obstacles, U-shaped obstacles, and complex topological disconnected regions. Traditional methods typically rely on global uniform sampling or local heuristic sampling, whose effective hit probability in narrow passages decreases rapidly as the passage width decreases. This results in a large number of sampling points falling into obstacle regions or irrelevant free regions, significantly increasing the computational cost of collision detection.

[0004] On the other hand, in environments with loop-shaped obstacles, local dead ends, and complex topological traps, traditional local expansion mechanisms lack the ability to actively cross disconnected regions. They often repeatedly attempt to cross these regions, making it difficult to establish cross-regional connections in a timely manner, resulting in excessively long search times or even search failures. Furthermore, existing planning methods typically separate sampling, graph search, and trajectory optimization into independent stages, making it difficult to balance front-end planning efficiency, theoretical convergence properties, and back-end control executability.

[0005] Therefore, there is an urgent need to propose a path planning method that can take into account the characteristics of multiple different path planning stages in complex environments and improve the efficiency of path planning. Summary of the Invention

[0006] In view of this, it is necessary to provide a path planning method, device and system, and electronic equipment based on potential energy landscape and multi-state operators, so as to achieve the goal of taking into account the characteristics of multiple different path planning stages and achieving high planning efficiency.

[0007] To achieve the aforementioned technical effects, firstly, this application provides a path planning method based on potential energy landscape and multimorphic operators, including: Obtain environmental obstacle data for the path planning area, and construct the potential energy landscape function and corresponding potential energy gradient field for the path planning area based on the environmental obstacle data; Multiple preset sampling operators are provided. Based on the current search graph, the operator contribution parameters corresponding to each preset sampling operator are calculated. Based on the operator contribution parameters, the operator selection probability corresponding to each preset sampling operator is calculated. The preset sampling operator corresponding to the maximum value of the operator selection probability is obtained as the target sampling operator. Based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field, target candidate sample points are generated, and the current search graph is updated based on the target candidate sample points; Output the current optimal discrete path based on the updated current search graph.

[0008] In one possible embodiment, constructing the potential energy landscape function and corresponding potential energy gradient field of the path planning area based on the environmental obstacle data includes: The environmental obstacle data is spatially discretized to construct free space and obstacle space; Obtain the minimum distance between any position in the free space and the obstacle space to form a distance field; The potential energy landscape function is constructed based on the distance field and a set cutoff threshold. The potential energy gradient field is obtained by performing a difference calculation on the potential energy landscape function.

[0009] In one possible embodiment, constructing the potential landscape function based on the distance field and a set cutoff threshold includes: Based on formula Construct the potential energy landscape function; in, Let be the potential energy landscape function. To set the potential energy scaling factor, For the distance field, To set the cutoff threshold, To set the decay index, To avoid the positive small amount of singularity.

[0010] In one possible embodiment, the operator contribution parameters include the set sampling operator. Total number of triggers , number of effective collision-free samples Contribution to the reduction of optimal path cost And a score for contribution to escape behavior. The step of calculating the operator selection probability corresponding to each of the set sampling operators based on the operator contribution parameters includes: Based on formula group ; ; The set sampling operator is calculated. Operator selection probability ; in, To update the current search graph, The number of the sampling operator is set. The total number of the set sampling operators, and These are the weight coefficients corresponding to the contributions of path cost reduction, effective collision-free sampling, and escape behavior, respectively. and These are the regularization parameters used to avoid the denominator being zero. The cost of the current optimal path corresponding to the current search graph. The temperature parameter at the k-th update. γ is the initial temperature parameter, and γ is the temperature decay coefficient.

[0011] In one possible embodiment, the set sampling operator includes a thermal fluctuation operator, the operator selection probability includes the thermal fluctuation operator selection probability corresponding to the thermal fluctuation operator, and the path planning method based on potential energy landscape and multi-state operators further includes: If the selection probability of the thermal fluctuation operator is less than the set correction probability, the selection probability of the thermal fluctuation operator is adjusted to the set correction probability; The selection probabilities of other operators are adjusted based on the set correction probability.

[0012] In one possible embodiment, the target sampling operator is a quantum tunneling operator, and the generation of target candidate sample points based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field includes: Obtain seed nodes in the current search graph where the number of local expansion failures is greater than the failure count threshold and the potential gradient is greater than the gradient threshold. Based on the potential energy gradient field, the potential energy gradient vector of the seed point is obtained, and an orthogonal projection matrix is ​​constructed based on the potential energy gradient vector. A random vector following a zero-mean Gaussian distribution is provided, and candidate sampling points are determined based on the potential energy gradient vector, the orthogonal projection matrix, and the random vector; The candidate sampling points that pass the validity test are used as the target candidate sampling points.

[0013] In one possible embodiment, the target sampling operator is a quantum entanglement operator, and the generation of target candidate sample points based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field includes: Obtain the identifiers of the connected components of the nodes maintained during the graph search process of the current search graph; Calculate the distance between boundary nodes of different connected components and extract the nearest disconnected boundary point; The geometric midpoint and unit direction vector are calculated based on the disconnected boundary points. Construct an adaptive covariance matrix whose principal axis direction is the same as the unit direction vector; The target candidate sample points are generated based on the geometric midpoint and the adaptive covariance matrix.

[0014] Secondly, this application provides a path planning device based on potential energy landscape and multimorphic operators, comprising: An environmental modeling module is used to acquire environmental obstacle data of the path planning area and construct the potential energy landscape function and the corresponding potential energy gradient field of the path planning area based on the environmental obstacle data. The operator determination module is used to provide multiple set sampling operators, calculate the operator contribution parameters corresponding to each set sampling operator based on the current search graph, calculate the operator selection probability corresponding to each set sampling operator based on the operator contribution parameters, and obtain the set sampling operator corresponding to the maximum value of the operator selection probability as the target sampling operator. The current search graph update module is used to generate target candidate sample points based on the target sampling operator, the potential energy landscape function and the potential energy gradient field, and update the current search graph based on the target candidate sample points; The path planning module is used to output the current optimal discrete path based on the updated current search graph.

[0015] Thirdly, this application provides a path planning system based on potential energy landscape and multimorphic operators, including: An environmental data acquisition device, automated equipment, and a path planning device based on potential energy landscape and multi-state operators connected to the environmental data acquisition device; The path planning device based on potential energy landscape and multi-state operators is used to execute the steps in the aforementioned path planning method based on potential energy landscape and multi-state operators to obtain the current optimal discrete path. The automated equipment is navigated based on the current optimal discrete path.

[0016] Fourthly, this application provides an electronic device, including: an input serial bus, an output serial bus, a memory, and a processor, wherein... The input serial bus is used to acquire environmental obstacle data; The memory is used to store the program and the environmental obstacle data; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the aforementioned design method for time-frequency high-resolution acoustic signals and obtain the current optimal discrete path. The output serial bus is used to output the current optimal discrete path to the automation equipment for path navigation.

[0017] The beneficial effects of this application are: Compared with related technologies, the path planning method, apparatus, system, and electronic device based on potential energy landscape and multi-state operators provided in this application, in the path planning process, first acquires environmental obstacle data of the path planning area, constructs the potential energy landscape function and corresponding potential energy gradient field of the path planning area based on the environmental obstacle data, realizes obstacle position modeling of the path planning area, further provides multiple set sampling operators, calculates the operator contribution parameters corresponding to each set sampling operator based on the current search graph, calculates the operator selection probability corresponding to each set sampling operator based on the operator contribution parameters, and obtains the set sampling operator corresponding to the maximum value of the operator selection probability. By using a fixed sampling operator as the target sampling operator, the most suitable target sampling operator that contributes the most to the current search graph can be selected. Therefore, the target candidate sample points generated subsequently based on the target sampling operator, the potential landscape function, and the potential gradient field are also more compatible with the current search graph. The current search graph is further updated based on the target candidate sample points. The current optimal discrete path is output based on the updated current search graph, which allows the current optimal discrete path to select a more suitable fixed sampling operator as the current search graph changes. This can take into account the characteristics of different path planning stages corresponding to multiple fixed sampling operators, thereby improving the efficiency of path planning. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the path planning method based on potential energy landscape and multimorphic operators provided in this application embodiment; Figure 2 This is a schematic diagram illustrating the process of constructing the potential landscape function and the corresponding potential gradient field of the path planning region in the path planning method based on potential landscape and multi-state operators provided in the embodiments of this application. Figure 3 This is a schematic diagram of the process for generating target candidate sample points based on the quantum tunneling operator in the path planning method based on potential energy landscape and multi-state operators provided in the embodiments of this application. Figure 4 This is a schematic diagram of the process for generating target candidate sample points based on quantum entanglement operators in the path planning method based on potential energy landscape and multi-state operators provided in the embodiments of this application. Figure 5 A schematic diagram of the path planning device based on potential energy landscape and multi-state operators provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application; Figure 7 This is a schematic diagram of a path planning system based on potential energy landscape and polymorphic operators provided in an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0022] The terms "first," "second," etc., used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] This application provides a path planning method, apparatus, system, and electronic device based on potential energy landscape and multimorphic operators, which will be described below.

[0025] Please refer to Figure 1 The path planning method based on potential energy landscape and multimorphic operators provided in this application includes: Step S101: Obtain environmental obstacle data of the path planning area, and construct the potential energy landscape function and corresponding potential energy gradient field of the path planning area based on the environmental obstacle data.

[0026] In this step, environmental obstacle data for the path planning area is obtained by integrating 3D detection devices such as LiDAR, depth cameras, millimeter-wave radar, and cameras. The environmental obstacle data for the path planning area is obtained by combining the sensor data collected by these devices. For example, LiDAR acquires 3D point clouds to depict the geometric contours of objects, depth cameras provide dense depth information, millimeter-wave radar reliably detects target positions and velocities unaffected by adverse weather conditions, and cameras capture rich semantic features. After time-stamp alignment and intrinsic / extrinsic parameter calibration, the sensor data undergoes voxel filtering downsampling. A random sampling consensus algorithm is used to segment and remove ground points. Subsequently, Euclidean clustering or deep learning methods are used to aggregate non-ground points into independent obstacle instances. Obstacle-related data, such as the obstacle's 3D bounding box, obstacle core, and obstacle velocity, are extracted as environmental obstacle data.

[0027] After obtaining environmental obstacle data, please refer to Figure 2 In this embodiment, constructing the potential energy landscape function and the corresponding potential energy gradient field of the path planning region specifically includes the following steps: Step S201: Spatial discretization processing is performed on the environmental obstacle data to construct free space and obstacle space. In this step, spatial discretization is performed within the path planning area based on environmental obstacle data to locate obstacles, thereby dividing the entire path planning area into free space. With obstacle space Among them, obstacle space The area where obstacles are distributed includes the area occupied by the obstacles themselves and the area occupied by the expanding regions of the obstacles, such as rectangles or spheres. This is the free space. That is, obstacle space Other areas besides these.

[0028] Step S202: Obtain the minimum distance between any position in free space and the obstacle space to form a distance field.

[0029] In this step, the fast travel method is used to solve the Eikonal equation: , obtain the distance field Distance field The data corresponding to each location point is specifically the minimum distance between that location point and the obstacle space.

[0030] Step S203: Construct a potential landscape function based on the distance field and a set truncation threshold.

[0031] In this step, the specific formula is used. Construct a potential energy landscape function; where, Let the potential energy landscape function be... To set the potential energy scaling factor, For distance field, To set the cutoff threshold, To set the decay index, To avoid the positive small amount of singularity.

[0032] Step S204: Perform differential calculation on the potential energy landscape function to obtain the potential energy gradient field.

[0033] In this step, the specific method is through the formula. For potential landscape function The potential energy gradient field is obtained through differential calculation and stored as a lookup table. Here, x, y, and z represent the three coordinate directions in the spatial coordinate system. x、 y、 z represents the discrete step lengths in the x, y, and z directions, respectively. When the path planning region is a two-dimensional space, the difference term corresponding to the z direction can be omitted.

[0034] Step S102: Provide multiple set sampling operators, calculate the operator contribution parameters corresponding to each set sampling operator based on the current search graph, calculate the operator selection probability corresponding to each set sampling operator based on the operator contribution parameters, and obtain the set sampling operator corresponding to the maximum value of the operator selection probability as the target sampling operator.

[0035] In this step, the search graph is a discretized representation of the path planning region during the path planning process, consisting of nodes and weighted edges. A node represents a possible position and orientation of the path planning device, even including vectors with dimensions such as velocity and time. Edges represent feasible direct transitions between nodes, with associated costs (such as distance, energy consumption, and risk). The search graph can be, for example, a pre-built topology map or landmark map (such as a route map generated by PRM), but more commonly it is a grid or state grid formed during the path planning process. For example, a two-dimensional occupancy grid map is itself a cost map; the search graph uses the center of each passable grid as a node, with edges connecting nodes to their neighbors to form a graph structure.

[0036] The sampling operators are defined as rules applied to nodes. These operators are used to generate all legal neighbor nodes of a given node and calculate the transition cost, which precisely describes "starting from a certain state, where the next step can be taken and at what cost." In grid maps, common operators include four-connected or eight-connected movement, with each movement costing the grid resolution or diagonal distance. In dynamic environments, operators check if neighboring grids are occupied by obstacles or expansion layers, generating only collision-free nodes. In this step, several sampling operators specifically include the quantum tunneling operator, the quantum entanglement operator, and the thermal fluctuation operator. The "quantum tunneling operator" refers to a candidate sample generation mechanism based on potential energy gradient backward propagation and gradient orthogonal subspace perturbation; the "quantum entanglement operator" refers to a candidate sample generation mechanism based on the identification of disconnected components in the search graph and directional bridging distribution; and the "thermal fluctuation operator" refers to a basic random sampling mechanism used to maintain global exploration coverage.

[0037] In this step, after initializing the search graph, the operator contribution parameters corresponding to each set sampling operator are calculated based on the current search graph. The operator contribution parameters specifically include: the set sampling operator... Total number of triggers , number of effective collision-free samples Contribution to the reduction of optimal path cost And a score for contribution to escape behavior. .

[0038] Based on this, the operator selection probability corresponding to each set sampling operator is calculated based on the operator contribution parameter, specifically including: Based on formula group ; ; The set sampling operator is calculated. Operator selection probability ;in, To update the current search graph count, To set the number of the sampling operator, To set the total number of sampling operators, , and These are the weight coefficients corresponding to the contributions of path cost reduction, effective collision-free sampling, and escape behavior, respectively. and These are the regularization parameters used to avoid the denominator being zero. The cost of the current optimal path corresponding to the current search graph. The temperature parameter at the k-th update. γ is the initial temperature parameter, and γ is the temperature decay coefficient.

[0039] Finally, the operator selection probabilities corresponding to each set sampling operator are... Sort the samples by size, select the maximum value as the target operator selection probability, and use the sampling operator corresponding to the target operator selection probability as the target sampling operator.

[0040] Furthermore, since the thermal fluctuation operator, as a fundamental exploration operator, is typically used to maintain global search coverage, in some embodiments of this application, if the selection probability of the thermal fluctuation operator is less than a set correction probability, the selection probability of the thermal fluctuation operator is adjusted to the set correction probability; the selection probabilities of other operators are adjusted based on the set correction probability. Specifically, the adjustment of the selection probabilities of other operators is achieved by normalizing the ratio of their operator selection probabilities to the selection probability of the thermal fluctuation operator. For example, if the selection probability of the thermal fluctuation operator is 0.1, the selection probability of the quantum tunneling operator is 0.3, and the selection probability of the quantum entanglement operator is 0.6, and the set correction probability is 0.25, then the selection probability of the thermal fluctuation operator (0.1) is less than the set correction probability (0.25). Therefore, the selection probability of the thermal fluctuation operator is adjusted to the set correction probability of 0.25. Simultaneously, based on the ratio of the selection probability of the quantum tunneling operator (0.3) to the selection probability of the quantum entanglement operator (0.6) (1:2), and combined with the normalization process of the selection probability of the thermal fluctuation operator (0.25), the corrected operator selection probabilities are: thermal fluctuation operator selection probability 0.25, quantum tunneling operator selection probability 0.25, and quantum entanglement operator selection probability 0.5. That is, the ratio of the selection probability of the quantum tunneling operator to the selection probability of the quantum entanglement operator remains 1:2.

[0041] The adaptive scheduling mechanism based on probabilistic lower bound constraints in this application can dynamically balance exploration and utilization based on real-time feedback, while maintaining the global exploration capability.

[0042] Step S103: Generate target candidate sample points based on the target sampling operator, potential landscape function and potential gradient field, and update the current search graph based on the target candidate sample points.

[0043] In this step, please refer to Figure 3 The specific methods for generating target candidate sample points based on quantum tunneling operators include: Step S301: Obtain seed nodes in the current search graph where the number of local expansion failures is greater than the failure count threshold and the potential gradient is greater than the gradient threshold.

[0044] In this step, for each node in the current search graph, the number of consecutive local expansion failures during search graph updates is recorded. Simultaneously, the potential gradient of each node is recorded based on the potential landscape function and the potential gradient field. Then, seed nodes with a local expansion failure count greater than a failure count threshold and a potential gradient greater than a gradient threshold are selected. .

[0045] Furthermore, seed nodes At least one of the following conditions must be met: (1) The seed point is located within the distance interval [d] from the boundary of the obstacle. min ,d max ]Inside; (2) The proportion of collision failures of the leaf node corresponding to the seed point in the most recent M expansions exceeds the threshold ρ. fail ; (3) The mean potential energy in the neighborhood of the seed point is higher than the threshold U high .

[0046] Step S302: Obtain the potential gradient vector of the seed point based on the potential gradient field, and construct an orthogonal projection matrix based on the potential gradient vector.

[0047] In this step, based on the formula Obtain the potential gradient vector of the seed point Then based on the formula Construct the orthogonal projection matrix. Where I is the identity matrix. This represents the transpose of the potential energy gradient vector g. This is a regularization parameter used to avoid zero denominators, with the aim of ensuring that the formula remains stable when the gradient is close to zero.

[0048] Step S303: Provide a random vector that follows a zero-mean Gaussian distribution, and determine candidate sampling points based on the potential gradient vector, the orthogonal projection matrix, and the random vector.

[0049] In this step, based on the formula Generate a random vector whose zero-mean random perturbation behavior follows a zero-mean Gaussian distribution.

[0050] Furthermore, based on the formula Determine candidate sampling points, where λ is the step size coefficient that propagates in the opposite direction of the potential energy gradient. The regularization parameter is used to avoid the gradient norm being zero. Let I be the perturbation scale parameter of the quantum tunneling operator, and I be the identity matrix.

[0051] In this step, please refer to Figure 4 The specific methods for generating target candidate sample points based on quantum entanglement operators include: Step S401: Obtain the identifiers of the connected components of the nodes maintained during the graph search process of the current search graph. Step S402: Calculate the distance between boundary nodes of different connected components and extract the nearest disconnected boundary point. Step S403: Calculate the geometric midpoint and unit direction vector based on the disconnected boundary points.

[0052] In this step, the specific formula is used. The geometric midpoint is calculated, where The nearest disconnected boundary point is extracted in step S402 above.

[0053] Furthermore, based on the formula The unit direction vector is calculated, where, This is a regularization parameter used to avoid computational instability caused by the distance between two disconnected boundary points being zero or too small.

[0054] Furthermore, after calculating the geometric midpoint, the geometric midpoint is further corrected based on the following method: (1) Perform a weighted offset on the geometric midpoint based on the target point direction; (2) Correct the geometric midpoint based on the local visibility score; (3) Correct the geometric midpoint based on the proportion of local free space.

[0055] Step S404: Construct an adaptive covariance matrix whose principal axis direction is the same as the unit direction vector. Step S405: Generate target candidate sample points based on the geometric midpoint and the adaptive covariance matrix.

[0056] In this step, based on the formula Generate target candidate sample points, where This refers to the adaptive covariance matrix generated in step S404 above.

[0057] In this step, candidate sampling points that pass the validity test are also used as target candidate sampling points. That is, after obtaining candidate sampling points, a validity test is performed on each candidate sampling point; if a candidate sampling point is invalid, the sampling parameters are adjusted and the aforementioned steps are repeated to determine new candidate sampling points; if a candidate sampling point is valid, it is used as the target candidate sampling point.

[0058] Furthermore, the target candidate sampling points are added to the current search graph to obtain a new search graph, and the current search graph is updated based on the target candidate sampling points.

[0059] Step S104: Output the current optimal discrete path based on the updated current search graph.

[0060] In this step, firstly, the converged discrete path points are extracted from the updated current search graph, and a safe corridor composed of several overlapping convex polygons is generated in free space; then, the trajectory control points are constrained to the corresponding safe corridor, and an optimization problem with trajectory smoothness index as the objective is constructed; finally, after solving the optimization problem, the continuous trajectory that satisfies the maximum velocity and maximum acceleration constraints is output as the current optimal discrete path.

[0061] In narrow passage regions, the proportion of free space is extremely low, and uniform sampling is unlikely to hit traversable areas, resulting in a significant decrease in search efficiency. In U-shaped obstacles, cavity obstacles, and complex environments with multiple connections, local expansion mechanisms are prone to falling into topological traps, making it difficult to quickly establish bridges between disconnected regions. In this application, quantum tunneling operators are used to improve the directional penetration capability in narrow passage regions. The quantum tunneling mechanism based on gradient orthogonal projection can reduce ineffective expansion towards the obstacle normal and improve the effective sampling rate in narrow passages. Quantum entanglement operators are used to enhance the bridging capability of disconnected regions. The quantum entanglement mechanism based on disconnected component identification and directional bridging distribution can actively cross disconnected regions, improving the search capability for feasible solutions in complex topological environments. Thermal fluctuation operators are used to maintain the sufficiency of global exploration. An adaptive scheduling mechanism based on probabilistic lower bound constraints can dynamically balance exploration and utilization based on real-time feedback and maintain the global exploration capability. The most suitable operator is flexibly selected through the current search graph, thereby achieving asymptotic optimization planning of path planning.

[0062] Compared with related technologies, the path planning method based on potential energy landscape and multi-state operators provided in this application first acquires environmental obstacle data of the path planning area during the path planning process. Based on the environmental obstacle data, a potential energy landscape function and corresponding potential energy gradient field of the path planning area are constructed to model the obstacle positions in the path planning area. Furthermore, multiple preset sampling operators are provided. Based on the current search graph, the operator contribution parameters corresponding to each preset sampling operator are calculated. Based on the operator contribution parameters, the operator selection probability corresponding to each preset sampling operator is calculated, and the preset sampling operator corresponding to the maximum operator selection probability is obtained. As a target sampling operator, the most suitable target sampling operator that contributes the most to the current search graph can be selected. Therefore, the target candidate sample points generated subsequently based on the target sampling operator, potential landscape function, and potential gradient field are more compatible with the current search graph. The current search graph is further updated based on the target candidate sample points. The current optimal discrete path is output based on the updated current search graph, which allows the current optimal discrete path to select a more suitable set sampling operator as the current search graph changes. This can take into account the characteristics of different path planning stages corresponding to multiple set sampling operators, thereby improving the efficiency of path planning.

[0063] To better implement the path planning method based on potential energy landscape and multimorphic operators in the embodiments of this application, based on the path planning method based on potential energy landscape and multimorphic operators, correspondingly, as follows: Figure 5 As shown in the figure, this application embodiment also provides a path planning device based on potential energy landscape and multimorphic operators. The path planning device based on potential energy landscape and multimorphic operators includes: The environmental modeling module 501 is used to acquire environmental obstacle data of the path planning area and construct the potential energy landscape function and the corresponding potential energy gradient field of the path planning area based on the environmental obstacle data. The operator determination module 502 is used to provide multiple set sampling operators, calculate the operator contribution parameters corresponding to each set sampling operator based on the current search graph, calculate the operator selection probability corresponding to each set sampling operator based on the operator contribution parameters, and obtain the set sampling operator corresponding to the maximum value of the operator selection probability as the target sampling operator. The search graph update module 503 is used to generate target candidate sample points based on the target sampling operator, potential energy landscape function and potential energy gradient field, and update the current search graph based on the target candidate sample points. The path planning module 504 is used to output the current optimal discrete path based on the updated current search graph.

[0064] The path planning device based on potential energy landscape and multi-state operators provided in the above embodiments can realize the technical solutions described in the above embodiments of the path planning method based on potential energy landscape and multi-state operators. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the path planning method based on potential energy landscape and multi-state operators, which will not be repeated here.

[0065] Please refer to Figure 6 This application also provides an electronic device 600. The electronic device 600 includes an input serial bus 601, an output serial bus 602, a memory 603, and a processor 604, wherein... The system includes an input serial bus 601 for acquiring environmental obstacle data, a memory 603 for storing the program and the environmental obstacle data input via the serial bus 601, and a processor 604 coupled to the memory 603 for executing the path planning program based on potential energy landscape and multimorphic operators stored in the memory 603. This processor implements the steps in the path planning method based on potential energy landscape and multimorphic operators provided in the aforementioned method embodiments, obtaining the current optimal discrete path. An output serial bus 602 is used to send the obtained current optimal discrete path to automated equipment such as robots and drones for path navigation.

[0066] In some embodiments, processor 604 may be a central processing unit (CPU), microprocessor, or other data processing chip for running program code stored in memory 603 or processing data. In some embodiments, processor 604 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 604 may be local or remote. In some embodiments, processor 604 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-cloud, multi-cloud, or any combination thereof.

[0067] In one embodiment, when the processor 604 executes the fixed-space gas leak location program in the memory 603, the following steps can be implemented: Obtain environmental obstacle data for the path planning area, and construct the potential energy landscape function and corresponding potential energy gradient field for the path planning area based on the environmental obstacle data; Multiple preset sampling operators are provided. The operator contribution parameters corresponding to each preset sampling operator are calculated based on the current search graph. The operator selection probability corresponding to each preset sampling operator is calculated based on the operator contribution parameters. The preset sampling operator corresponding to the maximum value of the operator selection probability is obtained as the target sampling operator. Target candidate sample points are generated based on the target sampling operator, potential energy landscape function and potential energy gradient field, and the current search graph is updated based on the target candidate sample points. Output the current optimal discrete path based on the updated current search graph.

[0068] It should be understood that when the processor 604 executes the fixed space gas leak location program in the memory 603, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0069] Please refer to Figure 7 This application also provides a path planning system based on potential energy landscape and polymorphic operators, including: an environmental acquisition device 701, an automated device 702, and a path planning device 703 based on potential energy landscape and polymorphic operators connected to the environmental acquisition device 701; the path planning device 703 is used to execute the steps in the aforementioned path planning method based on potential energy landscape and polymorphic operators to obtain the current optimal discrete path; and to perform path navigation for the automated device 702 based on the current optimal discrete path.

[0070] The path planning method, apparatus, electronic device, and storage medium based on potential energy landscape and multimorphic operators provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A path planning method based on potential energy landscape and multimorphic operators, characterized in that, include: Obtain environmental obstacle data for the path planning area, and construct the potential energy landscape function and corresponding potential energy gradient field for the path planning area based on the environmental obstacle data; Multiple preset sampling operators are provided. Based on the current search graph, the operator contribution parameters corresponding to each preset sampling operator are calculated. Based on the operator contribution parameters, the operator selection probability corresponding to each preset sampling operator is calculated. The preset sampling operator corresponding to the maximum value of the operator selection probability is obtained as the target sampling operator. Based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field, target candidate sample points are generated, and the current search graph is updated based on the target candidate sample points; Output the current optimal discrete path based on the updated current search graph.

2. The path planning method based on potential energy landscape and multi-state operators according to claim 1, characterized in that, The construction of the potential energy landscape function and corresponding potential energy gradient field of the path planning area based on the environmental obstacle data includes: The environmental obstacle data is spatially discretized to construct free space and obstacle space; Obtain the minimum distance between any position in the free space and the obstacle space to form a distance field; The potential energy landscape function is constructed based on the distance field and a set cutoff threshold. The potential energy gradient field is obtained by performing a difference calculation on the potential energy landscape function.

3. The path planning method based on potential energy landscape and multimorphic operators according to claim 2, characterized in that, The construction of the potential landscape function based on the distance field and a set cutoff threshold includes: Based on formula Construct the potential energy landscape function; in, Let be the potential energy landscape function. To set the potential energy scaling factor, For the distance field, To set the cutoff threshold, To set the decay index, To avoid the positive small amount of singularity.

4. The path planning method based on potential energy landscape and multimorphic operators according to claim 2, characterized in that, The operator contribution parameters include the set sampling operator. Total number of triggers , number of effective collision-free samples Contribution to the reduction of optimal path cost And a score for contribution to escape behavior. The step of calculating the operator selection probability corresponding to each of the set sampling operators based on the operator contribution parameters includes: Based on formula group ; ; The set sampling operator is calculated. Operator selection probability ; in, To update the current search graph, The number of the sampling operator is set. The total number of the set sampling operators, , and These are the weight coefficients corresponding to the contributions of path cost reduction, effective collision-free sampling, and escape behavior, respectively. and These are the regularization parameters used to avoid the denominator being zero. The cost of the current optimal path corresponding to the current search graph. The temperature parameter at the k-th update. γ is the initial temperature parameter, and γ is the temperature decay coefficient.

5. The path planning method based on potential energy landscape and multimorphic operators according to claim 2, characterized in that, The set sampling operator includes a thermal fluctuation operator, the operator selection probability includes the thermal fluctuation operator selection probability corresponding to the thermal fluctuation operator, and the path planning method based on potential energy landscape and multi-state operators further includes: If the selection probability of the thermal fluctuation operator is less than the set correction probability, the selection probability of the thermal fluctuation operator is adjusted to the set correction probability; The selection probabilities of other operators are adjusted based on the set correction probability.

6. The path planning method based on potential energy landscape and multimorphic operators according to claim 1, characterized in that, The target sampling operator is a quantum tunneling operator, and the generation of target candidate sample points based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field includes: Obtain seed nodes in the current search graph where the number of local expansion failures is greater than the failure count threshold and the potential gradient is greater than the gradient threshold. Based on the potential energy gradient field, the potential energy gradient vector of the seed point is obtained, and an orthogonal projection matrix is ​​constructed based on the potential energy gradient vector. A random vector following a zero-mean Gaussian distribution is provided, and candidate sampling points are determined based on the potential energy gradient vector, the orthogonal projection matrix, and the random vector; The candidate sampling points that pass the validity test are used as the target candidate sampling points.

7. The path planning method based on potential energy landscape and multimorphic operators according to claim 1, characterized in that, The target sampling operator is a quantum entanglement operator, and the generation of target candidate sample points based on the target sampling operator, the potential energy landscape function, and the potential energy gradient field includes: Obtain the identifiers of the connected components of the maintenance nodes during the graph search process of the current search graph; Calculate the distance between boundary nodes of different connected components and extract the nearest disconnected boundary point; The geometric midpoint and unit direction vector are calculated based on the disconnected boundary points. Construct an adaptive covariance matrix whose principal axis direction is the same as the unit direction vector; The target candidate sample points are generated based on the geometric midpoint and the adaptive covariance matrix.

8. A path planning device based on potential energy landscape and multimorphic operators, characterized in that, include: An environmental modeling module is used to acquire environmental obstacle data of the path planning area and construct the potential energy landscape function and the corresponding potential energy gradient field of the path planning area based on the environmental obstacle data. The operator determination module is used to provide multiple set sampling operators, calculate the operator contribution parameters corresponding to each set sampling operator based on the current search graph, calculate the operator selection probability corresponding to each set sampling operator based on the operator contribution parameters, and obtain the set sampling operator corresponding to the maximum value of the operator selection probability as the target sampling operator. The current search graph update module is used to generate target candidate sample points based on the target sampling operator, the potential energy landscape function and the potential energy gradient field, and update the current search graph based on the target candidate sample points; The path planning module is used to output the current optimal discrete path based on the updated current search graph.

9. A path planning system based on potential energy landscape and multimorphic operators, characterized in that, include: An environmental data acquisition device, automated equipment, and a path planning device based on potential energy landscape and multi-state operators connected to the environmental data acquisition device; The path planning device based on potential energy landscape and multi-mode operator is used to execute the steps in the path planning method based on potential energy landscape and multi-mode operator as described in any one of claims 1 to 7, and obtain the current optimal discrete path; The automated equipment is navigated based on the current optimal discrete path.

10. An electronic device, characterized in that, include: The system consists of an input serial bus, an output serial bus, a memory, and a processor. The input serial bus is used to acquire environmental obstacle data; The memory is used to store the program and the environmental obstacle data; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the design method of time-frequency high-resolution acoustic signal according to any one of claims 1 to 7, and to obtain the current optimal discrete path; The output serial bus is used to output the current optimal discrete path to the automation equipment for path navigation.

Images