Air-ground multi-unmanned system cooperative scheduling and task planning method and device

Abstract

The present application relates to the technical field of geographic information and multi-robot cooperative operation, in particular to a method and device for air-ground multi-unmanned system cooperative scheduling and task planning. The present application divides a space region on a work layer into a plurality of task regions; then predicts the benefit cost data corresponding to each task region executed by an unmanned device, to obtain a task exclusive region corresponding to each unmanned device; then plans a global path required by the unmanned device for passing; and plans a local path for the unmanned device to pass within the task exclusive region based on a shared local map. The present application first divides the task regions based on the accessibility topology model of the entire work layer, so that the unmanned device can pass within the task region; and since the shared local map can provide geographic information within the task exclusive region that affects the passing of the unmanned device, the local path planned by the unmanned device based on the shared local map is executed to improve the passing efficiency, and further improve the task execution efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geographic information and multi-robot cooperative operation, in particular to an air-ground multi-unmanned system cooperative scheduling and task planning method and device. BACKGROUND

[0002] The unmanned operation task of urban underground space, plant pipe gallery, multi-storey building and post-disaster environment needs both unmanned equipment on the ground operation layer and unmanned equipment on the air operation layer. The unmanned equipment on the ground operation layer and the unmanned equipment on the air operation layer together constitute a multi-unmanned system. The existing unmanned equipment plans a path based on the local map perceived by the unmanned equipment. However, the map information that can be perceived by a single unmanned equipment is limited, so that the task execution efficiency is reduced based on the path.

[0003] In summary, the prior art results in low task execution efficiency of unmanned equipment.

[0004] Therefore, the prior art still needs to be improved and enhanced. SUMMARY

[0005] To solve the above technical problems, the present application provides an air-ground multi-unmanned system cooperative scheduling and task planning method and device, which solves the problem of low task execution efficiency of unmanned equipment caused by the prior art.

[0006] To achieve the above purpose, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides an air-ground multi-unmanned system cooperative scheduling and task planning method, which comprises:

[0008] A reachability topology model of the operation layer is constructed, the reachability topology model is used to represent the space region that can be reached by the unmanned equipment in a three-dimensional mode, and the operation layer includes a ground operation layer and an air operation layer;

[0009] Based on the reachability topology model, the space region on the operation layer that allows the unmanned equipment to pass through is divided into a plurality of task regions;

[0010] The benefit cost data of the unmanned equipment executing each task region is determined, and based on the benefit cost data, a plurality of task regions are allocated to the corresponding unmanned equipment to obtain the task exclusive region corresponding to each unmanned equipment;

[0011] The global path of the unmanned equipment between the task exclusive regions is planned;

[0012] When the unmanned device reaches the task area along the global path, a local shared map shared by the unmanned devices on each of the job layers is obtained, and a local path of the unmanned device inside the task-specific area is planned based on the local shared map.

[0013] In an implementation manner, a reachability topology model of a job layer is constructed, including:

[0014] Three-dimensional space data of the job layer is obtained;

[0015] Based on the three-dimensional space data, key nodes in the space region on the job layer are obtained, and the key nodes are used to represent nodes on the job layer that affect the passage of the unmanned device;

[0016] Device constraint data of the unmanned device is constructed;

[0017] Node space features between the key nodes are determined;

[0018] Based on the device constraint data and the node space features, reachable edges between the key nodes are determined, and the reachable edges are used to represent that the unmanned device is allowed to pass between the key nodes;

[0019] Based on the reachable edges, a reachability topology model of the job layer is constructed.

[0020] In an implementation manner, the device constraint data of the unmanned device is constructed, including:

[0021] The maximum safe passage speed of the unmanned device is obtained, the maximum attitude inclination angle allowed for the unmanned device is obtained, and the minimum safe distance allowed between the unmanned device and the obstacle is obtained;

[0022] Based on the maximum safe passage speed, the maximum attitude inclination angle, and the minimum safe distance, the device constraint data of the unmanned device is constructed.

[0023] In an implementation manner, based on the reachability topology model, the space region on the job layer that allows the unmanned device to pass is divided into a plurality of task areas, including:

[0024] The forbidden space on the job layer is obtained, and based on the forbidden space and the reachability topology model, the free space on the job layer that allows the unmanned device to pass is obtained;

[0025] A space unit scale is set, and the free space is divided into a plurality of task areas according to the space unit scale.

[0026] In an implementation manner, the benefit-cost data of each of the task areas performed by the unmanned device is determined, including:

[0027] The distance between the unmanned device and the task area is obtained, the energy consumption consumed by the unmanned device to reach the task area is obtained, and the visibility of the task area is obtained, the visibility being used to represent the degree of line of sight of the unmanned device to the task area;

[0028] The cost data of the unmanned device performing the task area is obtained according to the distance, the energy consumption and the visibility;

[0029] The task completion quality evaluation of the task area and the information gain corresponding to the task area are obtained, the task completion quality evaluation being used to represent the influence of the task area on the completeness of the task performed by the unmanned device;

[0030] The benefit data corresponding to the task area is obtained according to the task completion quality evaluation and the information gain corresponding to the task area and the energy consumption;

[0031] The benefit-cost data of each of the task areas performed by the unmanned device is determined according to the cost data and the benefit data.

[0032] In an implementation manner, a plurality of the task areas are allocated to corresponding unmanned devices based on the benefit-cost data, to obtain a task exclusive area corresponding to each of the unmanned devices, including:

[0033] The benefit-cost data of a plurality of the task areas corresponding to each of the unmanned devices is sorted in descending order to obtain a task area sequence corresponding to each of the unmanned devices;

[0034] The task exclusive area corresponding to each of the unmanned devices is obtained based on the task area sequence.

[0035] In an implementation manner, the task exclusive area corresponding to each of the unmanned devices is obtained based on the task area sequence, including:

[0036] A task uniqueness constraint that any of the task areas is only allocated to one of the unmanned devices is constructed;

[0037] The task exclusive area corresponding to each of the unmanned devices is obtained based on the task area sequence and the task uniqueness constraint.

[0038] In an implementation manner, a global path of the unmanned device between the task exclusive areas is planned, including:

[0039] constructing a cumulative energy consumption constraint of the unmanned device;

[0040] determining, based on the cumulative energy consumption constraint, a path corresponding to a minimum total travel time for the unmanned device to patrol the task-specific area, and taking the path corresponding to the minimum total travel time as a global path.

[0041] In an implementation manner, planning a local path of the unmanned device inside the task-specific area based on the local shared map comprises:

[0042] acquiring a local map collected by a sensor of the unmanned device itself;

[0043] registering the local shared map based on the local map;

[0044] fusing the local map and the local shared map after registration to obtain a fused map;

[0045] planning a local path of the unmanned device inside the task-specific area based on the fused map.

[0046] In a second aspect, the embodiments of the present application further provide an air-ground multi-unmanned system cooperative scheduling and task planning device, wherein the device comprises the following components:

[0047] a topological model construction module configured to construct a reachability topological model of a work layer, the reachability topological model being configured to represent a space region reachable by an unmanned device in a three-dimensional mode, the work layer comprising a ground work layer and an air work layer;

[0048] a region division module configured to divide, based on the reachability topological model, a space region on the work layer that is allowed to be traveled by the unmanned device into a plurality of task regions;

[0049] a task allocation module configured to determine a benefit-cost data of the unmanned device performing each of the task regions, and allocate a plurality of the task regions to corresponding unmanned devices based on the benefit-cost data to obtain a task-specific area corresponding to each of the unmanned devices;

[0050] a global path planning module configured to plan a global path of the unmanned device between the task-specific areas;

[0051] a local path planning module configured to, when the unmanned device reaches the task region along the global path, acquire a local shared map shared by the unmanned devices on each of the work layers, and plan a local path of the unmanned device inside the task-specific area based on the local shared map.

[0052] Beneficial effects: firstly, the application divides the space area allowing unmanned equipment to pass on the operation layer into several task areas based on the reachability topological model representing the global map of the operation layer; then, the application predicts the benefit cost data of each task area corresponding to the unmanned equipment, allocates several task areas to the corresponding unmanned equipment, and obtains the task exclusive area corresponding to each unmanned equipment; then, the application plans the global path required for the unmanned equipment to pass between the task exclusive areas; and based on the local map shared between the unmanned equipment on the aerial operation layer and the unmanned equipment on the ground operation layer, the application plans the local path of the unmanned equipment when passing within the task exclusive area based on the shared local map. From the above analysis, firstly, the application divides the task area based on the reachability topological model of the entire operation layer, so that the unmanned equipment can pass within the task area; and since the shared local map can provide geographic information within the task exclusive area affecting the passing of the unmanned equipment, the unmanned equipment plans the local path based on the shared local map to perform the task, so as to improve the passing efficiency and further improve the task execution efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is the overall flowchart of the application;

[0054] Figure 2 is the flowchart of the multi-task scheduling of the unmanned equipment in the embodiment of the application by using the ant colony optimization algorithm;

[0055] Figure 3 is the schematic diagram of the stable feature map and the cross-map attention fusion network in the embodiment of the application;

[0056] Figure 4 is the structure diagram of the air-ground multi-unmanned system cooperative scheduling and task planning device provided by the application. DETAILED DESCRIPTION

[0057] The technical solutions in the application are described clearly and completely in combination with the embodiments and the drawings of the specification. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0058] It is found through research that the unmanned operation tasks of the urban underground space, the factory pipe gallery, the multi-storey building and the post-disaster environment require both the unmanned equipment on the ground operation layer and the unmanned equipment on the aerial operation layer, and the unmanned equipment on the ground operation layer and the unmanned equipment on the aerial operation layer together constitute a multi-unmanned system. The existing unmanned equipment plans a path based on the local map perceived by the unmanned equipment, and the map information that can be perceived by a single unmanned equipment is limited, so that the task execution efficiency is reduced based on the path.

[0059] To solve the above technical problems, the application provides a method and device for cooperative scheduling and task planning of an air-ground multi-unmanned system, which solves the problem of low task execution efficiency of unmanned equipment caused by the prior art.

[0060] The method for cooperative scheduling and task planning of the air-ground multi-unmanned system of the embodiment can be applied to a terminal device, which can be a terminal product with a data processing function, such as a computer. Figure 1

[0061] S100, a reachability topological model of a work layer is constructed, the reachability topological model being used to represent a space region reachable by unmanned equipment in a three-dimensional mode, the work layer including a ground work layer and an air work layer;

[0062] S200, based on the reachability topological model, a space region on the work layer that allows the unmanned equipment to pass through is divided into a plurality of task regions;

[0063] S300, the unmanned equipment is determined to execute a benefit cost data of each task region, and based on the benefit cost data, the plurality of task regions are allocated to corresponding unmanned equipment, to obtain a task exclusive region corresponding to each unmanned equipment;

[0064] S400, a global path of the unmanned equipment between the task exclusive regions is planned;

[0065] S500, when the unmanned equipment reaches the task region along the global path, a local shared map shared by the unmanned equipment on each work layer is obtained, and a local path of the unmanned equipment in the task exclusive region is planned based on the local shared map.

[0066] The specific process realized by steps S100 to S500 is as follows:

[0067] First, based on existing three-dimensional space data (i.e., basic geographic information data of the work layer in three-dimensional form), a reachability topological model is constructed in a unified coordinate system according to the ground work layer and the air work layer. Through topological simplification and consistency optimization, a global map for subsequent task division and path planning is formed.

[0068] The passable space is extracted on the global map and region division is performed, to generate a task unit set with attributes such as center coordinates, work difficulty, and information gain. A task map is constructed according to the reachable relationship between regions, to provide a unified task model for cooperative scheduling of multiple robots.

[0069] ​In combination with the movement ability and endurance constraints of the aerial unmanned devices and various ground heterogeneous unmanned devices, the executable task-specific areas are screened for each unmanned device on the task graph. A centralized or distributed scheduling strategy is used to solve the task allocation scheme, so that the overall task completion time or energy consumption index is optimal.

[0070] On the basis of the task allocation result, a task sequence for each unmanned device to access the task area is generated, and a global path skeleton between areas is planned on the reachability topology model. The planning takes into account constraints such as floor switching, narrow passages and forbidden area detours, etc., to ensure that the path is globally feasible.

[0071] When each unmanned device executes the task along the global path, the laser radar is used to collect the environment point cloud in real time, and a local map of the local area is constructed. The system generates and evaluates candidate trajectories on the local map, and selects trajectories that meet the safety distance and smoothness requirements for navigation and obstacle avoidance control.

[0072] The unmanned devices periodically output local key frame point clouds during the operation, which are preprocessed and feature-extracted for mutual matching and registration. Through heterogeneous point cloud fusion, the local maps of each unmanned device are unified in the same global coordinate system, realizing the continuous updating and sharing of local point cloud maps.

[0073] Real-time monitoring of the task progress, energy consumption status and map quality of each unmanned device, when path blocking, energy consumption overrun or map degradation occurs, triggers local re-planning or task rescheduling instructions. Only the affected tasks and paths are adjusted locally to reduce the disturbance to the global plan.

[0074] Through the above steps, global task organization based on reachability topology model, local intelligent navigation and multi-source point cloud consistency mapping in complex restricted environment can be realized. Multi-modal heterogeneous unmanned devices can realize efficient collaborative operation and stable environmental perception under a unified spatial reference.

[0075] The reachability topology model in step S100 is used to represent which internal space regions of the operation layer are allowed to be passed through by the unmanned devices, and which space regions between the unmanned devices are allowed to be passed through. The operation layer includes a ground operation layer and an aerial operation layer. The unmanned devices on the ground operation layer are robot dogs or tracked robots, and the unmanned devices on the aerial operation layer are unmanned aerial vehicles. The space regions on the aerial operation layer that do not allow unmanned aerial vehicles to pass through include the space above densely populated areas, the space above railways, the space above highways, the space above high-voltage power lines, and the space along key infrastructure such as oil and gas pipelines.

[0076] The construction of the reachability topology model of the operation layer in step S100 includes the following specific steps S101, S102, S103, S104, S105, S106 and S107:

[0077] S101, Obtain the three-dimensional spatial data of the working layer.

[0078] Three-dimensional spatial data includes basic geographic information data for the operational layer.

[0079] S102, Based on the three-dimensional spatial data, key nodes within the spatial region on the operation layer are obtained, and the key nodes are used to characterize the nodes on the operation layer that affect the passage of the unmanned equipment.

[0080] When the working level is the ground level, the space area can be any of the floors.

[0081] Key nodes include key access nodes and cross-level topology nodes. Key access nodes include passageway intersections, stairwells, and platform corners. Cross-level topology nodes include staircases, ramps, shafts, and platform edges between working levels. Shafts refer to elevator shafts, ventilation shafts, pipe shafts, and stairwells. The existence of shafts allows unmanned equipment to move between different working levels. Platform edges refer to the boundaries of the planar areas within a working level that are accessible for passage or work, such as corridor edges, stairwell edges, and equipment platform edges.

[0082] S103, obtain the maximum safe passage speed of the unmanned equipment, obtain the maximum allowed attitude tilt angle of the unmanned equipment, and obtain the minimum allowed safe distance between the unmanned equipment and the obstacle.

[0083] use Representing the The first spatial region The maximum safe passage speed corresponding to each unmanned device, using Representing the The first spatial region The maximum attitude tilt angle corresponding to each unmanned device, using Representing the The first spatial region The minimum safe distance for each unmanned device.

[0084] S104, Based on the maximum safe passage speed, the maximum attitude tilt angle, and the minimum safe distance, construct the equipment constraint data of the unmanned equipment.

[0085] Use sets The representative is located at the The first spatial region The equipment constraint data corresponding to each unmanned device.

[0086] ;

[0087] S105, Determine the node spatial characteristics between the key nodes.

[0088] Node spatial features include the slope angle corresponding to the path segment between two key nodes, and the minimum distance between that path segment and the nearest obstacle, expressed as follows: Representing the The key node and the first The slope angle corresponding to the path segment between two key nodes is used as... Representing the The key node and the first The minimum distance between the path segment and the nearest obstacle between each key node.

[0089] S106, Based on the device constraint data and the node space characteristics, determine the reachable edges between the key nodes, and the reachable edges are used to represent the unmanned equipment that is allowed to pass between the key nodes.

[0090] When two critical nodes satisfy the following constraints, a reachable edge is established between the two critical nodes, which means that the reachable edge represents the permission for unmanned equipment to travel between the two critical nodes.

[0091] This constraint is , .

[0092] By applying this constraint, the three-dimensional topology network model (which is the aforementioned three-dimensional spatial data) retains only staircases, ramps, and passageways that are feasible for unmanned equipment, thus avoiding the creation of unexecutable paths in subsequent planning stages.

[0093] S107. Based on the reachable edges, construct the reachability topology model of the job layer.

[0094] After obtaining the reachable edges, continue to optimize the model parameters of the 3D topological network model. and the number of boundary lines until the energy function The minimum value is obtained to arrive at the final 3D topology network model, which is then used as the reachability topology model for the job layer. The model parameters are as follows: Model parameters are used to describe the geometric representation of the same planar primitive in three-dimensional space. It includes the planar normal direction and its spatial position parameters, and is a continuous geometric abstraction of a local point cloud set. The optimization can achieve unified fitting and geometric consistency constraints of multi-source point clouds at the planar level under the condition of noise.

[0095] ;

[0096] In the formula, This represents the distance deviation between the actual distance and the theoretical shortest distance between key nodes. This represents the threshold that allows for distance deviation. for The weight, This represents the deviation between the theoretical tilt angle of the unmanned device and the actual tilt angle of the unmanned device allowed by the path. This represents the allowable threshold for attitude tilt angle deviation. Represents geometric terms, for The weight, Represents shape constraints. for The weights are defined by the geometric term, which characterizes the geometric consistency relationships that spatial primitives in the topological model should satisfy. This consistency originates from the structural priors in the existing 3D model of the factory area, such as the parallelism, orthogonality, and relative positional stability between room envelope structures. When the angles or relative relationships between planes deviate from the actual building geometry due to segmentation errors or incomplete local data during topological modeling, the geometric term corrects the mathematical surface parameters by introducing these constraints, ensuring that the topological model maintains reasonable spatial geometric relationships in its overall structure.

[0097] Shape constraints are used to limit the complexity of primitive structures and boundary representations during topology optimization. This constraint prevents unnecessary excessive splitting, redundant boundaries, or non-physical shapes from occurring during topology reconstruction or energy minimization, ensuring that the resulting spatial primitives maintain simplicity and interpretability at the topological level, consistent with the actual building configuration.

[0098] The reachability edges on the reachability topology model reveal which key nodes in the operational layer allow unmanned equipment to pass through. These key nodes are called connected key nodes, and the area covered by these connected key nodes is the spatial region where unmanned equipment can pass. Step S200 divides the spatial region where unmanned equipment can pass into several task regions, including the following specific steps:

[0099] S201, obtain the restricted space on the operation layer, and based on the restricted space and the reachability topology model, obtain the free space on the operation layer that allows the unmanned equipment to pass through.

[0100] FreeSpace represents free space. In this embodiment, FreeSpace can be filtered out from the global map represented by the accessibility topology model based only on the restricted space on the working layer (restricted space is the space where unmanned equipment is prohibited from entering), or it can be filtered out from the global map based on obstacles, restricted areas, safety buffer zones, and the traversable range of unmanned equipment.

[0101] S202, Set the spatial unit scale, and divide the free space into several task areas according to the spatial unit scale.

[0102] use This represents a spatial unit scale, which is the area of ​​each task region. In this embodiment, the scale is set according to the sensing range, operational accuracy requirements, and management unit scale of the unmanned equipment. Size.

[0103] use based on Divide FreeSpace into several task regions. That is, the space partitioning function. Based on After obtaining the task area, clustering is performed by combining topological structures such as floor segmentation and passage skeleton to optimize each task area, so that the optimized task area is spatially compact and topologically connected.

[0104] Determining the revenue-cost data for each task area performed by the unmanned equipment in step S300 includes the following specific steps S301, S302, S303, S304, and S305:

[0105] S301, obtain the distance between the unmanned device and the task area, obtain the energy consumption required for the unmanned device to reach the task area, and obtain the visibility of the task area, wherein the visibility is used to characterize the extent to which the line of sight of the unmanned device reaches the task area.

[0106] S302, based on the distance, energy consumption, and visibility, obtain the cost data for the unmanned equipment to perform the task in the designated area.

[0107] use Representing the Cost data corresponding to each task region The calculation formula is as follows:

[0108] ;

[0109] In the formula, Represents the current position of any unmanned device and the number The distance between each task area Representing the The visibility of the task area, which is the detection range of the unmanned equipment can cover. The extent of each task area Represents any unmanned device in the The energy consumed by each task area to execute a task. , , They are , , The weight.

[0110] This embodiment can also replace distance with the distance from any unmanned device to the current location. The time required for each task area.

[0111] S303, obtain the task completion quality evaluation of the task area and the information gain corresponding to the task area. The task completion quality evaluation is used to characterize the impact of the task area on the completeness of the unmanned equipment's task execution.

[0112] S304. Based on the task completion quality evaluation, information gain, and energy consumption corresponding to the task area, the revenue data corresponding to the task area is obtained.

[0113] use Representing the Revenue data corresponding to each task area The calculation formula is as follows:

[0114] ;

[0115] In the formula, Representing the The task completion quality evaluation for each task area refers to the clarity or completeness of the task area on the map. Representing the Information gain refers to the amount of effective information that an unmanned device can collect if it inspects that task area. , , Represent , , The weight.

[0116] S305, based on the cost data and the revenue data, determine the revenue and cost data for the unmanned equipment to perform each of the task areas.

[0117] That is, calculating revenue data. With cost data ratio And this ratio will be used as the basis for determining the first time any unmanned device performs its first operation. The revenue and cost data corresponding to each task area.

[0118] ;

[0119] That is, use represent the selection of one of the unmanned devices to perform the work task of the first task area.

[0120] The step S300 of allocating a plurality of task areas to the corresponding unmanned device based on the benefit cost data obtains the task exclusive area corresponding to each unmanned device, including the following specific steps S306, S307 and S308:

[0121] S306, for each unmanned device, the benefit cost data of a plurality of task areas is sorted in descending order to obtain the task area sequence for each unmanned device.

[0122] That is, the plurality of task areas are arranged in descending order according to the benefit cost data of each task area calculated for each unmanned device, each unmanned device corresponds to a task area sequence, and the task area sequences corresponding to different unmanned devices are different.

[0123] S307, constructing a task uniqueness constraint that any of the task areas is only assigned to one of the unmanned devices.

[0124] That is, a task area can only be assigned to one unmanned device, and the same unmanned device can perform the work task of multiple task areas.

[0125] S308, based on the task area sequence and the task uniqueness constraint, the task exclusive area corresponding to each unmanned device is obtained.

[0126] From the task area sequence corresponding to each unmanned device, the task area with high benefit cost data is selected first, and it is also ensured that the same task area can only be assigned to one unmanned device. Based on this, a plurality of task areas are allocated to the corresponding unmanned device, and after the task area is allocated to the unmanned device, the task area becomes the task exclusive area of the unmanned device, that is, only the unmanned device can perform the work task of the task exclusive area.

[0127] The embodiment can also construct an unmanned device task allocation optimization model based on the energy consumption upper limit of each unmanned device, the total task duration and other constraint conditions, and allocate a plurality of task areas to the corresponding unmanned device based on the task allocation optimization model.

[0128] The embodiment can also allocate the task area to the corresponding unmanned device based on the ant colony optimization algorithm as shown in Figure 2 .

[0129] As shown in Figure 2The real-time state and task sequence of the job equipment (job equipment refers to unmanned equipment) are initialized, and a graph is initialized (FIG . A three-dimensional topological graph of the job layer), the allocation solving process of the task area is converted into an ant colony routing process, and the movement of the ant colony from the starting point to the ending point in the graph achieves multi-objective model solving.

[0130] Let represent the probability of the ant colony selecting the task area with the starting point and the ending point at the th moment:

[0131] ;

[0132] wherein represents the visibility of the task area at the th moment, the starting point of the task area is , and the ending point is , represents the pheromone concentration of the task area at the th moment, represents a set composed of task areas that have not been accessed by the th unmanned equipment, and are both constants.

[0133] In the iteration process of the algorithm, the pheromone concentration of the ant colony needs to be calculated and updated, and the calculation method is as follows:

[0134] ;

[0135] wherein represents the pheromone evaporation rate of the path, represents the pheromone change amount, represents the pheromone concentration at the th moment.

[0136] Let represent the pheromone change amount corresponding to the th unmanned equipment:

[0137] ;

[0138] When the th unmanned equipment finds a complete feasible distance from the starting point to the ending point, then , otherwise .

[0139] wherein represents the pheromone change amount corresponding to the The model total efficiency obtained by the unmanned device can quickly obtain an optimal solution of the cooperative operation task and improve the efficiency of the cooperative operation task through multiple iterations of the model. Meanwhile, in the movement process of the multi-modal operation device (the multi-modal operation device is an unmanned device on the ground operation layer and an unmanned device on the air operation layer), the position of the device is tracked, when the position of the device abnormally changes or there is an emergency dynamic task scheduling demand, the cooperative task sequence and the optimization target set are updated, the multi-objective model optimization algorithm is used to solve and update the cooperative task again, and finally the task scheduling of the multi-intelligent operation device is realized.

[0140] The planning of the global path of the unmanned device between the task-specific areas in step S400 includes the following specific steps S401 and S402:

[0141] S401, constructing the cumulative energy consumption constraint of the unmanned device;

[0142] S402, determining the path corresponding to the minimum total travel time used by the unmanned device to patrol the task-specific areas based on the cumulative energy consumption constraint, and taking the path corresponding to the minimum total travel time as the global path.

[0143] That is, after obtaining the reachability topology model, the travel path of the unmanned device between the task-specific areas to which it belongs is planned based on the reachability topology model. The specific process is as follows:

[0144] In order to consider the overall efficiency of the multiple unmanned devices, the multiple task-specific area access problems corresponding to the multiple unmanned devices are modeled as VRP-TW problems, and VRP-TW represents Vehicle Routing Problem with Time Windows, which represents the vehicle routing problem with time windows. Let For the unmanned device The total travel time for completing all the task-specific areas of the unmanned device, the overall goal is:

[0145] ;

[0146] In the formula, represents a set composed of the task-specific areas of each unmanned device, represents the total number of unmanned devices.

[0147] In terms of constraints, it is necessary to meet the condition that each task-specific area is accessed only by one robot, and the working time (total time including working time) of each task-specific area falls within a preset time window, which is a time region composed of a set start working time and a set end working time; at the same time, the cumulative energy consumption of each unmanned device executing a complete route does not exceed its maximum available energy. In combination with the distance and time cost given by the edge weight in the task graph, the embodiment solves the order in which each unmanned device accesses its task-specific area and the global route skeleton through an improved heuristic search or genetic algorithm.

[0148] After the unmanned device obtains its task-specific area, the embodiment uses the node-edge topology structure in the reachability topology model to generate a feasible path according to constraints such as floor switching, passage direction, and no-entry area. A hybrid algorithm ( algorithm, that is, a heuristic graph search algorithm combining discrete graph search and continuous motion planning, is used to search for a path with the minimum cost from the current node to the target task area representative point between key topological nodes, and during path generation, narrow passages, stairwells, elevator shafts, and other spatial bottlenecks are considered. For possible unmanned device conflict areas, passing time is allocated in the time domain or a buffer node is introduced to reduce congestion and collision risks, so as to obtain a global path skeleton (global path) that can be executed in space and time.

[0149] The step S500 of planning a local path of the unmanned device inside the task-specific area based on the local shared map includes the following specific steps S501, S502, S503, and S504:

[0150] S501, acquiring a local map collected by a sensor of the unmanned device itself.

[0151] S502, registering the local shared map based on the local map.

[0152] The sensor is a laser radar installed on the unmanned device. The unmanned device on the ground operation layer collects a local map of its task-specific area through the laser radar. At the same time, since the range detected by the laser radar of the unmanned device on the ground operation layer is limited, the laser radar on the unmanned device on the aerial operation layer can detect a local map containing geographical information of the ground operation layer, which is shared to the unmanned device on the ground operation layer as a local shared map.

[0153] The local map and the local shared map are registered in the coordinate system to realize the registration of the local map and the local shared map.

[0154] S503, the local map and the registered local shared map are merged to obtain a merged map.

[0155] By merging the two maps, the missing geographical information in each other can be compensated for.

[0156] S504, Based on the fused map, plan the local path of the unmanned equipment within the mission-specific area.

[0157] This means that based on the fused map, the travel path of the unmanned equipment within the mission-specific area is planned to ensure that the unmanned equipment can avoid obstacles.

[0158] This embodiment can also plan local paths in the following way:

[0159] The unmanned equipment in the aerial operation layer uses lidar to perform overhead scans of a large mission-specific area, obtaining sparse but wide-coverage point clouds for real-time map updates and structural constraints; the unmanned equipment on the ground operation layer supplements local detail information through close-range high-density scanning for precise obstacle avoidance and local environment modeling.

[0160] The raw point clouds acquired by the two types of lidar are subjected to unified preprocessing, including time synchronization, distortion correction, noise filtering, outlier removal, and voxel downsampling. The point clouds are then unified to their respective unmanned equipment coordinate systems. Combined with odometry and attitude estimation results, continuous scan data is accumulated into a local map. This local map can be expressed as an occupancy grid, range field, or local voxel map, while also marking obstacles, free space, and safety buffer zones, providing a high-precision environmental description for local path planning, obstacle avoidance control, and local information gain assessment.

[0161] This embodiment uses, as follows: Figure 3 The stable feature map cross-attention fusion network shown addresses the issues of viewpoint differences and uneven point cloud density between heterogeneous sensors in the air and on the ground. Figure 3 In this context, GISCA-Net represents an end-to-end trainable fully convolutional network for text detection in natural scenes.

[0162] This involves selecting keyframe point clouds from local maps of various platforms (including ground and aerial operation layers) and performing uniform preprocessing to ensure comparability in scale and noise levels. Subsequently, robust features are extracted based on geometric structure and reflection intensity characteristics to construct a stable feature map. Specifically, for each keyframe point cloud, a region growing method based on normal consistency is used to extract major planar clusters. The geometric center, normal, and area of ​​each planar cluster are calculated, and these are defined as planar nodes. :

[0163] ;

[0164] in, Represents planar clusters The geometric center, Represents the plane normal. Represents the area of ​​a plane.

[0165] Meanwhile, a density-based clustering algorithm is used to cluster high-reflectivity regions. For each cluster, the centroid, mean intensity, and number of points are calculated, and these are defined as intensity nodes. :

[0166] ;

[0167] in, For the centroid coordinates of the cluster, The average reflection intensity This represents the number of cluster points.

[0168] All planar nodes and strength nodes together constitute a node set. :

[0169] ;

[0170] Based on the node set, an edge set is constructed using the geometric distance, normal angle, and height difference between nodes as weights. To form a stable feature map :

[0171] ;

[0172] in, This represents the structural similarity between two nodes. For two stable feature maps from different platforms, the system can use graph similarity-based neural networks or traditional graph matching methods to estimate the correspondence between nodes and generate a set of matching pairs. .

[0173] Based on the set of matching pairs First, the initial rigid body transformation is estimated using RANSAC (RANSAC stands for Random Sampling Consensus), eliminating outliers. Based on this, the generalized iterative nearest-point algorithm is preferentially used for fine registration to solve for the optimal rigid body transformation. To minimize the residual from the weighted point to the plane:

[0174] ;

[0175] in, and These are two target points, and and Forming a matching point pair for The direction of the law, for Rigid body transformation, Obtained through rigid body transformation , represent and The relative positions between them.

[0176] After obtaining the cross-platform relative pose, this embodiment constructs a joint factor graph or pose graph, unifying the odometry constraints, loop closure constraints, and cross-platform constraints of each platform into a factor set, and solving for the optimal pose of all keyframes through sparse nonlinear optimization. The objective function is as follows:

[0177] ;

[0178] In the formula, Representing the The predicted values ​​of each constraint factor for the state. Represents the corresponding observed value. This represents the covariance matrix.

[0179] Ultimately, a globally consistent 3D map is obtained in a 3D coordinate system, which is then used as a real-time local point cloud model shared by all unmanned devices. This local point cloud model represents a local map. During mission execution, each unmanned device can periodically obtain updates from the real-time point cloud map, enabling cross-platform map sharing and collaborative decision-making.

[0180] After obtaining the shared local map, the unmanned equipment begins to navigate according to the local map. In this embodiment, a model predictive control framework (MPC) is used to navigate the unmanned equipment so that it can avoid obstacles.

[0181] The cost function of MPC is defined as the sum of the squares of the state and velocity errors.

[0182] To achieve autonomous navigation of unmanned equipment in complex dynamic environments, this embodiment employs a Model Predictive Control (MPC) framework to minimize path tracking error and control costs while satisfying obstacle avoidance constraints. During the planning process, state and control constraints are particularly important; the cost function is defined as the sum of the squares of state and velocity errors. The sum of squares representing the state and velocity errors:

[0183] ;

[0184] in and These are the reference state and reference speed of the unmanned equipment. The weight coefficients corresponding to the states. The weighting coefficients corresponding to speed, The state of the unmanned equipment to be calculated. The speed of the unmanned equipment to be calculated. Represents the total duration. and This is used to control the degree of penalty for unmanned equipment deviating from the reference state and reference velocity. Through the definition of this penalty function, MPC can predict the unmanned equipment path and perform optimization solutions over multiple future time steps.

[0185] To achieve obstacle avoidance, unmanned equipment must maintain a safe distance from obstacles in the environment. The unmanned equipment must operate within a certain time frame. The state is represented by the center point, denoted as .in The value depends on the dimensions of the workspace. When the unmanned device moves in a plane, the value is taken as... ,at this time ,in Location of unmanned equipment. This indicates the azimuth angle of the unmanned equipment. To ensure safety, the minimum distance between the unmanned equipment and obstacles must meet the following requirements:

[0186] ;

[0187] in, The above formula describes the outline of an unmanned device (which can be a robot) at a predetermined safe distance. With obstacles The minimum distance between them Since the distance function is non-convex, directly using it for optimization would significantly increase computational complexity. Therefore, this embodiment linearizes the distance function and introduces a dynamic safety distance strategy. Indicates time The corresponding safe distance. To increase the flexibility of obstacle avoidance planning, this embodiment adopts... Regularization ,use Represents the result after regularization Regularization is:

[0188] ;

[0189] in This is a weighting factor used to adjust obstacle avoidance capability. By adding a regularization term, the objective of MPC becomes minimizing the total cost function under obstacle avoidance, state evolution, and boundary constraints:

[0190] ;

[0191] The constraints are:

[0192] ;

[0193] ;

[0194] ;

[0195] ;

[0196] ;

[0197] In the formula, Representing unmanned equipment The state at any given moment, Represents the minimum speed. Represents the maximum speed. Represents the minimum acceleration. Represents the maximum acceleration. Representing unmanned equipment The speed at that time Representing unmanned equipment The speed at that time This represents the minimum distance between the unmanned device and the obstacle. This represents the maximum distance between the unmanned device and the obstacle. Represents obstacles The set constituted Indicates that the system is in Known bias term or external disturbance compensation term at time, Used to characterize states that are not included in the state matrix during discretization or linearization. With control matrix Explicitly described influencing factors. That is... Compensation is provided for state shifts caused by model linearization errors, environmental slope changes, uneven terrain, wheel-to-ground or foot-to-ground contact disturbances, or external interference, enabling the prediction model to more accurately approximate the actual motion evolution of unmanned equipment within local time windows. This is achieved by introducing... It can improve the accuracy of state prediction and the robustness of trajectory planning while maintaining the convexity and solvability of the model predictive control problem.

[0198] Through dynamic adjustment This allows for better handling of complex obstacle distributions during path planning.

[0199] The embodiment adopts an alternating direction multiplier method to solve the above-mentioned total cost function minimization. The alternating direction multiplier method decomposes the total cost function minimization into multiple groups of sub-problems, and improves the calculation efficiency by solving the dual variables in parallel. The alternating direction multiplier method includes the following steps one, step two and step three:

[0200] Step one, the obstacle avoidance constraint is converted into a dual form:

[0201] ;

[0202] ;

[0203] wherein, and are dual variables, is a matrix transpose of , is a matrix transpose of , is a dual cone corresponding to the cone, represents a standard symbol in convex optimization and dual theory, represents a geometric position mapping function corresponding to of the unmanned device at , is a constant vector or a support function parameter, represents a geometric boundary parameter vector of . Through this dual transformation, the original non-convex problem can be converted into a biconvex form which is easier to solve.

[0204] Step two, in order to solve this biconvex optimization problem, first construct the augmented Lagrangian function :

[0205] ;

[0206] wherein, represents an indicator function of the state, represents an indicator function of the dual variable, is a penalty parameter, is used to balance the relationship between the objective function and the constraint condition, represents the total number of obstacles, , are both Lagrange multipliers, which are used to update the dual constraint residuals in the form of equality to promote the feasibility conditions of the original variables and the dual variables in the iteration process. To assist the relaxation variables, the inequality barrier constraints are split and introduced into the alternating direction multiplier method framework, so as to decouple the constraints and improve the parallel solving efficiency. represents the residual function corresponding to the barrier constraint, which is used to measure whether the dual feasibility constraint composed of the state variable, the dual variable and the relaxation variable is satisfied. represents the cone constraint consistency residual function, which is used to measure whether the dual variable satisfies the geometric constraint condition of the dual cone space to which it belongs.

[0207] Step three, in each iteration, the alternating direction multiplier method updates each variable by the following steps:

[0208] Original variable update: minimizing the part of the augmented Lagrangian function about , the next step state and control vector are obtained:

[0209] ;

[0210] In the formula, the represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration, represents the value of produced in the iteration.

[0211] Dual variable update: minimizing the part of the augmented Lagrangian function about , the new value of the dual variable is obtained, which is represented by :

[0212] ;

[0213] Dual gradient update: according to the constraint condition, the Lagrange multiplier is updated to obtain the new Lagrange multiplier :

[0214] ;

[0215] ;

[0216] wherein, denotes the structure matrix corresponding to the cone constraint or geometric constraint, used to map to the corresponding constraint space; denotes the geometric description matrix of , used to characterize the linear inequality representation of the obstacle convex polyhedron or convex set; denotes the linear mapping matrix from state to geometry space, used to map to its position or contour representation in the workspace, denotes the value of generated by the th iteration, denotes the geometric position mapping function corresponding to when the unmanned device is at denotes the value of generated by the th iteration.

[0217] The embodiment splits the original problem into multiple sub-problems that can be processed in parallel to improve computational efficiency and scalability. The embodiment also further accelerates convergence through a warm-start strategy, which uses the solution from the previous iteration as the initial value, reducing the overhead of repeated calculations. The premise for using the warm-start strategy is that during autonomous navigation, the changes in the state of the unmanned device and obstacles are usually small between consecutive time steps. Therefore, the current solution can be well used as the initial value for subsequent problems to reduce the number of iterations and accelerate algorithm convergence. The computational complexity of the entire RDA (Distributed Planner) process is evaluated in the following aspects.

[0218] First, the original problem solving involved in each iteration (i.e., state and control vector updates) is usually completed by a convex optimization solver, and its computational complexity is , is the computational complexity symbol, denotes the dimension of the state variable in the original optimization problem, i.e., the number of degrees of freedom contained in the state vector of the unmanned device within a single time step; denotes the dimension of the control variable, i.e., the dimension of the control input vector within a single time step; therefore, corresponds to the total dimension of the decision variable involved in one iteration of the original convex optimization problem. Second, there are independent sub-problems regarding the optimization of dual variables, and the computational complexity of each sub-problem is , a serial number representing a sub-problem, a serial number representing a sub-problem, a serial number representing a sub-problem, a serial number representing a sub-problem, the number of dual variable dimensions or linear inequality constraints related to the

[0219] ;

[0220] wherein, represents the number of iterations required for RDA to converge.

[0221] The embodiment also provides an air-ground multi-unmanned system cooperative scheduling and task planning device, as shown in Figure 4 The device comprises the following components:

[0222] A topological model construction module 01 is configured to construct a reachability topological model of a work layer, wherein the reachability topological model is configured to represent a space region reachable by an unmanned device in a three-dimensional mode, and the work layer comprises a ground work layer and an air work layer.

[0223] A region division module 02 is configured to divide a space region allowed for the unmanned device to pass through on the work layer into a plurality of task regions based on the reachability topological model.

[0224] A task allocation module 03 is configured to determine a benefit and cost data of the unmanned device performing each of the task regions, and allocate a plurality of the task regions to corresponding unmanned devices based on the benefit and cost data to obtain a task exclusive region corresponding to each of the unmanned devices.

[0225] A global path planning module 04 is configured to plan a global path of the unmanned device between the task exclusive regions.

[0226] A local path planning module 05 is configured to obtain a local shared map shared by the unmanned devices on each of the work layers when the unmanned device reaches the task region along the global path, and plan a local path of the unmanned device in the task exclusive region based on the local shared map.

[0227] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments of the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0228] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An air-ground multi-unmanned system cooperative scheduling and task planning method, characterized in that, The method comprises the following steps: determining the node space features between the key nodes affecting the unmanned device passing in the space region on the operation layer, determining the reachable edges between the key nodes allowing the unmanned device to pass based on the unmanned device constraint data and the node space features, constructing the reachability topology model of the operation layer based on the reachable edges, and the reachability topology model is used to represent the space region that the unmanned device can reach in a three-dimensional mode, and the operation layer comprises a ground operation layer and an air operation layer; based on the reachability topology model, the space region allowing the unmanned device to pass on the operation layer is divided into a plurality of task regions; determining the benefit and cost data of each task region executed by the unmanned device, and assigning a plurality of task regions to the corresponding unmanned device based on the benefit and cost data, so as to obtain the task exclusive region corresponding to each unmanned device; planning the global path of the unmanned device between the task exclusive regions; when the unmanned device reaches the task region along the global path, obtaining the local shared map shared by the unmanned devices on each operation layer, and planning the local path of the unmanned device in the task exclusive region based on the local shared map; determining the benefit and cost data of each task region executed by the unmanned device comprises: obtaining the distance between the unmanned device and the task region, obtaining the energy consumption consumed by the unmanned device to reach the task region, and obtaining the visibility of the task region, the visibility representing the degree of line of sight of the unmanned device reaching the task region; obtaining the cost data of the unmanned device executing the task region according to the distance, the energy consumption and the visibility; obtaining the task completion quality evaluation of the task region and the information gain corresponding to the task region, the information gain being the number of effective information that the unmanned device can collect by inspecting the corresponding task region, and the task completion quality evaluation representing the influence of the task region on the completeness of the task executed by the unmanned device; obtaining the benefit data corresponding to the task region according to the task completion quality evaluation corresponding to the task region and the information gain and the energy consumption; determining the benefit and cost data of each task region executed by the unmanned device according to the ratio of the benefit data to the cost data; planning the global path of the unmanned device between the task exclusive regions comprises: constructing the cumulative energy consumption constraint of the unmanned device; determining the path corresponding to the minimum total passing time used by the unmanned device to inspect the task exclusive region based on the cumulative energy consumption constraint, and taking the path corresponding to the minimum total passing time as the global path.

2. The air-ground multi-UAV system collaborative scheduling and task planning method of claim 1, wherein, constructing the reachability topology model of the operation layer comprises: obtaining the three-dimensional space data of the operation layer; obtaining the key nodes in the space region on the operation layer based on the three-dimensional space data; constructing the device constraint data of the unmanned device; determining the node space features between the key nodes; determining the reachable edges between the key nodes based on the device constraint data and the node space features; Based on the reachable edge, a reachability topology model of the work layer is constructed.

3. The air-ground multi-UAV system collaborative scheduling and task planning method of claim 2, wherein, Device constraint data of the unmanned device is constructed, including: The maximum safe passing speed of the unmanned device is obtained, the maximum attitude inclination angle allowed for the unmanned device is obtained, and the minimum safe distance allowed between the unmanned device and the obstacle is obtained; Based on the maximum safe passing speed, the maximum attitude inclination angle, and the minimum safe distance, the device constraint data of the unmanned device is constructed.

4. The air-ground multi-UAV system collaborative scheduling and mission planning method of claim 1, wherein, Based on the reachability topology model, a space region on the work layer that allows the unmanned device to pass through is divided into a plurality of task regions, including: The forbidden space on the work layer is obtained, and based on the forbidden space and the reachability topology model, the free space on the work layer that allows the unmanned device to pass through is obtained; A space unit scale is set, and the free space is divided into a plurality of task regions according to the space unit scale.

5. The air-ground multi-UAV system collaborative scheduling and mission planning method of claim 1, wherein, Based on the benefit cost data, a plurality of the task regions are assigned to the corresponding unmanned device, and the task exclusive region corresponding to each unmanned device is obtained, including: The benefit cost data of a plurality of the task regions corresponding to each unmanned device is sorted in descending order to obtain a task region sequence corresponding to each unmanned device; Based on the task region sequence, the task exclusive region corresponding to each unmanned device is obtained.

6. The air-ground multi-UAV system collaborative scheduling and mission planning method of claim 5, wherein, Based on the task region sequence, the task exclusive region corresponding to each unmanned device is obtained, including: A task uniqueness constraint that any one of the task regions is only assigned to one of the unmanned devices is constructed; Based on the task region sequence and the task uniqueness constraint, the task exclusive region corresponding to each unmanned device is obtained.

7. The air-ground multi-UAV system collaborative scheduling and mission planning method of claim 1, wherein, Based on the local shared map, a local path of the unmanned device within the task exclusive region is planned, including: A local map collected by a sensor of the unmanned device itself is obtained; The local shared map is registered based on the local map; The local map and the registered local shared map are fused to obtain a fused map; Based on the fused map, a local path of the unmanned device within the task exclusive region is planned.

8. An air-ground multi-unmanned system cooperative scheduling and task planning device, characterized in that, The device includes the following components: A topology model construction module is configured to determine node space features between key nodes affecting the passing of the unmanned device in a space region on a work layer, determine reachable edges between the key nodes allowing the unmanned device to pass based on device constraint data of the unmanned device and the node space features, and construct a reachability topology model of the work layer based on the reachable edges, the reachability topology model being used to represent a space region reachable by the unmanned device in a three-dimensional mode, the work layer including a ground work layer and an air work layer; A region division module is configured to divide a space region on the work layer allowing the unmanned device to pass into a plurality of task regions based on the reachability topology model; and a task allocation module, configured to determine benefit-cost data of each of the task areas performed by the unmanned device, and allocate a plurality of the task areas to corresponding unmanned devices based on the benefit-cost data, to obtain a task-specific area corresponding to each of the unmanned devices; a global path planning module, configured to plan a global path of the unmanned device between the task-specific areas; a local path planning module, configured to obtain a local shared map shared by the unmanned devices on each of the work layers when the unmanned device reaches the task area along the global path, and plan a local path of the unmanned device within the task-specific area based on the local shared map; determining the benefit-cost data of each of the task areas performed by the unmanned device comprises: obtaining a distance between the unmanned device and the task area, obtaining energy consumption consumed by the unmanned device to reach the task area, and obtaining a visibility degree of the task area, the visibility degree being used to represent a degree of a line of sight of the unmanned device reaching the task area; obtaining cost data of the unmanned device performing the task area according to the distance, the energy consumption and the visibility degree; obtaining a task completion quality evaluation of the task area and information gain corresponding to the task area, the information gain being a quantity of effective information collected by the unmanned device in the corresponding task area, and the task completion quality evaluation being used to represent an influence of the task area on a completeness degree of the task performed by the unmanned device; obtaining benefit data corresponding to the task area according to the task completion quality evaluation and the information gain corresponding to the task area and the energy consumption; determining the benefit-cost data of each of the task areas performed by the unmanned device according to a ratio of the benefit data to the cost data; planning the global path of the unmanned device between the task-specific areas comprises: constructing a cumulative energy consumption constraint of the unmanned device; determining a path corresponding to a minimum total passing time used by the unmanned device to inspect the task-specific areas based on the cumulative energy consumption constraint, and taking the path corresponding to the minimum total passing time as the global path.

Images