Vehicle and unmanned aerial vehicle cooperative path planning method in depopulated area road network-free environment

Abstract

The invention relates to the technical field of cooperative path planning, and particularly discloses a vehicle and unmanned aerial vehicle cooperative path planning method in a depopulated area road-network-free environment, and the method comprises the steps: obtaining the terrain data of the depopulated area road-network-free environment, obtaining a grid terrain map, carrying out the region division, outputting a divided sub-region map, and carrying out the vehicle path planning. Carrying out unmanned aerial vehicle path planning; outputting an initial path planning result of an unmanned aerial vehicle; carrying out cooperative task allocation; outputting an optimized vehicle and unmanned aerial vehicle cooperative path planning result; and generating intelligent decision suggestions. According to the method, the problems that the coordination path is invalid, real-time resource change is not considered, the staticizing defect of task allocation exists, a single target is emphasized, multi-target coordination is neglected and the environmental adaptability is insufficient due to the fact that the map is difficult to update in real time in the existing algorithm are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of collaborative path planning, and specifically to a method for collaborative path planning between vehicles and unmanned aerial vehicles in an uninhabited area without a road network. Background Art

[0002] Traditional transportation and exploration methods that rely on road networks are ineffective in uninhabited areas (such as plateaus, deserts, and polar regions) due to their complex terrain, harsh climate, and lack of infrastructure. Collaboration between vehicles and drones can overcome the limitations of a single platform: vehicles act as mobile bases to provide endurance, while drones, with their aerial perspective, enable wide-area reconnaissance and precise positioning. The combination of these two can accomplish tasks such as material transportation, disaster relief, and ecological monitoring. Drones are limited by battery capacity, typically with a single flight time of less than two hours and limited payload capacity. By using vehicles as mobile charging stations, drones can return for resupply between missions, significantly extending their operational time. Ground vehicles are easily obstructed by obstacles and limited field of view in complex terrain. Drones equipped with equipment such as lidar and multispectral sensors can generate real-time three-dimensional maps, assisting vehicles in avoiding hazards such as potholes and steep slopes.

[0003] At present, there are still some deficiencies in the research on collaborative path planning between vehicles and UAVs in uninhabited areas without road networks. Specifically, sudden disasters such as mudslides and sandstorms can change the terrain instantly, but existing algorithms find it difficult to update maps in real time, resulting in the failure of collaborative paths. Existing studies mostly use preset task priorities, but do not consider real-time resource changes, and have static task allocation defects. Existing algorithms often focus on a single target, ignore multi-target collaboration, and lack environmental adaptability. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a method for collaborative path planning between vehicles and drones in an uninhabited area without a road network, which can effectively solve the problems involved in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for collaborative path planning of vehicles and drones in an unmanned area without a road network, comprising the following steps: obtaining terrain data of the unmanned area without a road network, preprocessing the terrain data of the unmanned area without a road network to obtain a gridded terrain map, performing regional division, and outputting a sub-area map after the division; collecting vehicle parameters, starting point and target point coordinates, combining the divided sub-area maps, performing vehicle path planning, and outputting a preliminary path planning result for the vehicle; collecting drone performance parameters, starting point and target point coordinates, combining the divided sub-area maps, performing drone path planning, and outputting a preliminary path planning result for the drone; obtaining vehicle and drone task types, combining the preliminary path planning results of the vehicle and the drone, performing collaborative task allocation, and outputting an optimized collaborative path planning result for the vehicle and drone; monitoring real-time environmental information of the unmanned area without a road network, dynamically adjusting the optimized collaborative path planning result for the vehicle and drone, and obtaining a dynamically adjusted path planning result for the vehicle and drone; performing risk prediction based on the dynamically adjusted path planning results for the vehicle and drone, and generating intelligent decision recommendations.

[0006] As a further method, the terrain data of the no-man's land without road network environment is obtained, and the terrain data of the no-man's land without road network environment is preprocessed to obtain a grid terrain map. The specific analysis process is: obtaining the terrain data of the no-man's land without road network environment, specifically including the no-man's land without road network terrain map, the no-man's land without road network elevation data map, and the no-man's land without road network obstacle distribution map; dividing the no-man's land without road network terrain map into regular grid units, each grid unit is 10m×10m; assigning a terrain type to each grid unit, and recording the elevation value of each grid unit based on the no-man's land without road network elevation data map, the terrain type is specific Including flat land, high slopes, gentle slopes, rivers, forests, mountains, canyons, and areas of unknown terrain types; based on the distribution map of obstacles in uninhabited areas without road networks, the impassable areas are marked as impassable in the grid terrain map. The impassable areas specifically include cliffs, deep water areas, and dense vegetation areas; for areas that are passable but have special restrictions, specifically including shallow water areas and gravel areas, they are marked as restricted access areas, and the access restrictions of the restricted access areas are recorded, including maximum speed limit, minimum speed limit, and load limit; the grid terrain map marked with terrain type, elevation value, and obstacles is output.

[0007] As a further method, regional division is performed and a sub-region map after division is output. The specific analysis process is as follows: regional division is performed on the gridded terrain map marked with terrain type, elevation value, and obstacles; grid cells with elevation changes less than 10% and terrain types of flat land or gentle slopes are selected as vehicle-passable areas; grid cells with terrain types of mountainous areas, canyons, and unknown terrain types are selected as drone-reconnaissance areas; material handover points are used as areas where vehicles and drones need to collaborate to complete tasks, and are recorded as collaborative operation areas; and a sub-region map after division is output, with vehicle-passable areas, drone-reconnaissance areas, and collaborative operation areas marked.

[0008] As a further method, vehicle parameters, starting point and target point coordinates are collected, and the vehicle path planning is carried out in combination with the divided sub-area map, and the preliminary path planning result of the vehicle is output. The specific analysis process is as follows: vehicle parameters, starting point and target point coordinates are collected, and the vehicle parameters specifically include maximum speed, drivable terrain type, and vehicle weight; the vehicle passable area is analyzed. If there is an impassable area, the impassable area is removed from the vehicle passable area. If there is a restricted pass area, the pass restriction conditions of the restricted pass area are obtained. If the vehicle can meet the pass restriction conditions of the restricted pass area, The restricted area in the vehicle passable area is retained. If the vehicle cannot meet the pass restriction conditions of the restricted area, the restricted area is removed from the vehicle passable area to obtain an updated vehicle passable area; the updated vehicle passable area is used as the search space, and the starting point and target point are used as the starting point and end point of the search; a heuristic search algorithm is used to search for a path in combination with vehicle parameters, which must meet the vehicle parameters; during the search process, a cost function is calculated for each grid cell: the terrain type-terrain cost mapping set stored in the database is obtained, and the matching terrain cost is determined based on the terrain type; Obtaining the precipitation amount and wind speed of the uninhabited area without a road network environment; marking the precipitation amount and wind speed of the uninhabited area without a road network environment as vehicle cost impact features; obtaining a cost impact feature-vehicle cost weight mapping table stored in a database; determining a matching vehicle cost weight based on the current cost impact feature, wherein the vehicle cost weight specifically includes a weight coefficient of the terrain cost and a weight coefficient of the vehicle energy consumption; The vehicle cost function is: vehicle cost = vehicle distance + α × terrain cost + β × vehicle energy consumption, where α is the weight coefficient of the terrain cost, β is the weight coefficient of the vehicle energy consumption, vehicle energy consumption is the energy consumed by the vehicle passing through the grid unit, which is dimensionless, and vehicle distance is the distance the vehicle passes through the grid unit, which is dimensionless; the path with the minimum vehicle cost is selected as the vehicle preliminary path planning result; the vehicle preliminary path planning result is smoothed and the vehicle preliminary path planning result is output.

[0009] As a further method, the performance parameters, starting point and target point coordinates of the UAV are collected, and the UAV path planning is carried out in combination with the divided sub-area map, and the preliminary path planning result of the UAV is output. The specific analysis process is as follows: the performance parameters, starting point and target point coordinates of the UAV are collected, and the UAV performance parameters specifically include the maximum flight speed and the flight altitude range; the UAV reconnaissance area is used as the search space, and the starting point and target point are used as the starting point and end point of the search; the Dijkstra path planning algorithm is used to combine the UAV performance parameters for path search, which must meet the UAV performance parameters; during the search process, the cost function is calculated for each grid cell: the current UAV flight speed and UAV flight wind direction are obtained, the UAV flight speed and UAV flight wind direction are stored as specified labels, the specified label-wind speed cost mapping set stored in the database is obtained, and the matching wind speed cost is determined based on the specified label; Obtain the precipitation amount in the roadless environment of the uninhabited area, the wind direction in the roadless environment of the uninhabited area, and the drone payload; mark the precipitation amount in the roadless environment of the uninhabited area, the wind direction in the roadless environment of the uninhabited area, and the drone payload as drone cost impact features; obtain a cost impact feature-drone cost weight mapping table stored in the database; determine the matching drone cost weight based on the current cost impact feature, where the drone cost weight specifically includes a weight coefficient of the wind speed cost and a weight coefficient of the drone energy consumption; The drone cost function is: drone cost = drone distance + γ × wind speed cost + δ × drone energy consumption, where γ is the weight coefficient of wind speed cost, δ is the weight coefficient of drone energy consumption, drone energy consumption is the energy consumed by the drone passing through the grid unit, dimensionless, and distance is the distance the drone passes through the grid unit, dimensionless; the path with the lowest drone cost is selected as the drone's preliminary path planning result; the drone's preliminary path planning result is highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collisions; the drone's preliminary path planning result is output.

[0010] As a further method, the vehicle and UAV task types are obtained, and the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV are combined to perform collaborative task allocation, and the optimized vehicle and UAV collaborative path planning results are output. The specific analysis process is: obtain the vehicle and UAV task types, combine the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV, and output a collaborative task allocation table; according to the collaborative task allocation table, the UAV flies in front of the vehicle, scouts the road conditions in advance and transmits information to the vehicle in real time, and the vehicle adjusts the driving path according to the UAV feedback. If the UAV detects that the path is blocked, it obtains the alternative detour path of the blocked path stored in the database. The path obstruction types specifically include landslides and the presence of other vehicles; recalculate the available The vehicle cost function and the UAV cost function are calculated for each grid cell of the alternative detour path, and the path with the minimum vehicle cost is selected as the vehicle collaborative path planning result of the alternative detour path, and the path with the minimum UAV cost is selected as the UAV collaborative path planning result of the alternative detour path; the vehicle collaborative path planning result of the alternative detour path is smoothed, and the UAV collaborative path planning result of the alternative detour path is highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collisions; the optimized vehicle and UAV collaborative path planning results are output, and based on the optimized vehicle and UAV collaborative path planning results, the optimized vehicle and UAV path coordinate point sequence, the estimated driving time of the planned path, and the total energy consumption of the path planning are obtained.

[0011] As a further method, the vehicle and UAV task types are obtained, and the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV are combined to output a collaborative task allocation table. The specific analysis process is: obtain the vehicle and UAV task types, and decompose the task types into multiple subtasks based on the predefined task splitting logic stored in the database; based on the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV, project the vertical reconnaissance range of the UAV path to the ground, perform geometric intersection calculation with the vehicle path, calculate the estimated time for the vehicle and UAV to arrive at the overlapping area, analyze the overlapping area of the vehicle and UAV in time and space, assign subtasks to the most suitable equipment, and obtain the corresponding requirements of the subtasks stored in the database: if the subtask requires a high-altitude perspective and is time-sensitive If the subtask requires a payload or long-term endurance, it is assigned to a vehicle. If the subtask is in a spatiotemporal overlap zone and requires dual-device collaboration, the drone and vehicle will execute it jointly. The execution subject and time node of each subtask are determined. Based on ROS+Gazebo, spatiotemporal overlap and task allocation are simulated, and paths are visualized in real time. Based on the Python library, Shapely is used for geometric analysis, and Pandas is used for spatiotemporal table processing. The subtask completion rate is verified to determine whether all subtasks are assigned and there are no conflicts. If there are conflicts, the corresponding priorities of the subtasks stored in the database are compared, and the subtask with the higher priority is completed first. If the conflicting subtasks have equal priorities, the total energy consumption of the completed subtasks is compared, and the subtask with the lower total energy consumption is completed first. The drone's built-in sensor collects the drone's remaining power in real time. When the drone's remaining power is lower than the preset power threshold stored in the database, it triggers a charging request to the drone's nest. The preset power threshold = the basic power threshold stored in the database + the power consumption of the drone's real-time corresponding position when returning. When the drone's remaining power is lower than the preset threshold stored in the database, the vehicle automatically drives to the nearest safe stop, which is a flat ground with an elevation change of less than 5%, and calls the vehicle's position coordinates. With the drone's current position as the starting point and the vehicle's position as the end point, the Dijkstra algorithm is used to plan the shortest path, avoiding areas with wind speeds above 10m / s. The drone returns to the drone's nest on the vehicle along the shortest path for charging. The vehicle pauses during charging and marks the vehicle speed as 0. After charging is completed, the drone continues to execute the subtask from the breakpoint. If there is no conflict, the verification is passed. The collaborative task allocation table is output, which contains the execution subject and time node of each subtask.

[0012] As a further method, the real-time environmental information of the uninhabited area without road network is monitored, and the specific analysis process is: the real-time environmental information of the uninhabited area without road network is monitored to obtain the real-time environmental information data of the uninhabited area without road network, specifically including the real-time weather data of the uninhabited area without road network and the real-time terrain data of the uninhabited area without road network, wherein: the real-time weather data of the uninhabited area without road network specifically includes the real-time wind speed of the uninhabited area without road network, the real-time wind direction of the uninhabited area without road network and the real-time precipitation of the uninhabited area without road network; the real-time terrain data of the uninhabited area without road network specifically includes the real-time terrain landslide situation of the uninhabited area without road network and the real-time mud-rock flow situation of the uninhabited area without road network. If there is a real-time terrain landslide in the uninhabited area without road network, the impact range of the real-time terrain landslide in the uninhabited area without road network and the real-time terrain landslide position coordinates of the uninhabited area without road network are obtained. If there is a real-time mud-rock flow in the uninhabited area without road network, the impact range of the real-time mud-rock flow in the uninhabited area without road network and the real-time mud-rock flow coordinates of the real-time mud-rock flow in the uninhabited area without road network are obtained.

[0013] As a further method, the optimized vehicle and UAV collaborative path planning results are dynamically adjusted to obtain dynamically adjusted vehicle and UAV path planning results. The specific analysis process is: based on the real-time environmental information data of the unmanned area without road network, the paths of the vehicle and UAV are dynamically adjusted: if the real-time wind speed of the unmanned area without road network ahead exceeds the wind speed threshold stored in the database or the real-time wind direction of the unmanned area without road network exceeds the wind direction threshold stored in the database or the real-time precipitation in the unmanned area without road network exceeds the precipitation threshold stored in the database or there is a real-time terrain collapse in the unmanned area without road network or there is a real-time mudslide in the unmanned area without road network, the vehicle re-plans the detour path and the UAV scouts the new feasible path in advance; selects the detour path with the shortest distance while ensuring that the path is still within the passable area of the vehicle and UAV; re-plans the detour path The vehicle cost function and the UAV cost function are calculated for each grid cell of the detour path with the shortest distance. The path with the minimum vehicle cost is selected as the vehicle detour collaborative path planning result, and the path with the minimum UAV cost is selected as the UAV detour collaborative path planning result; the vehicle detour collaborative path planning results are smoothed, and the UAV detour collaborative path planning results are highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collisions; based on the terrain type corresponding to the vehicle path coordinate point sequence, the matching vehicle speed stored in the database is determined; and the dynamically adjusted vehicle and UAV path planning results are obtained, including the path coordinate point sequence of the vehicle and UAV, the UAV flight altitude, the terrain type corresponding to the path coordinate point sequence, and the vehicle speed corresponding to the vehicle path coordinate point sequence.

[0014] As a further method, risk prediction is performed based on the dynamically adjusted vehicle and UAV path planning results to generate intelligent decision-making recommendations. The specific analysis process is as follows: the dynamically adjusted vehicle and UAV path planning results are used as a trained fault tree analysis model, combined with Monte Carlo simulation, to output the risk probability of the dynamically adjusted vehicle and UAV path planning results and the corresponding main risk sources; the risk probability of the dynamically adjusted vehicle and UAV path planning results and the corresponding main risk sources are stored as specified labels, and the specified label-intelligent decision recommendation mapping table stored in the database is obtained, and the corresponding intelligent decision recommendation is matched based on the current specified label.

[0015] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: This invention provides a method for collaborative path planning between vehicles and drones in uninhabited areas without a road network. Preprocessing converts the terrain of the uninhabited area into a gridded map, which can quantitatively express details such as terrain undulations and obstacle distribution, avoiding the coarse modeling flaws of traditional raster maps. Vehicle path planning takes into account both terrain constraints and hardware performance, while drone path planning leverages the advantages of three-dimensional space, enhancing reconnaissance efficiency. Dynamic adjustments to the division of labor based on path planning results avoid energy waste caused by "drones waiting for vehicles." A game theory model is used to determine the optimal collaborative strategy, taking into account objectives such as path length, energy consumption, and mission completion time. Dynamic changes in the uninhabited area are monitored in real time using sensors (such as lidar and weather stations). Once the original path is detected as unfeasible, local replanning is immediately triggered to avoid mission interruption. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described with reference to the accompanying drawings: Figure 1 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0017] Reference Figure 1 As shown, the present invention provides a method for collaborative path planning between vehicles and UAVs in an uninhabited area without a road network, including: obtaining terrain data of the uninhabited area without a road network environment, preprocessing the terrain data of the uninhabited area without a road network environment, obtaining a gridded terrain map, performing regional division, and outputting the divided sub-area map.

[0018] The specific analysis process is as follows: obtain terrain data of the uninhabited area without road network environment, including the uninhabited area without road network terrain map, the uninhabited area without road network elevation data map, and the uninhabited area without road network obstacle distribution map; divide the uninhabited area without road network terrain map into regular grid cells, each grid cell is 10m×10m; assign a terrain type to each grid cell, and record the elevation value of each grid cell based on the uninhabited area without road network elevation data map. The terrain types specifically include flat land, high slope, gentle slope, river, forest, mountainous area, canyon, and unknown terrain type area; based on the uninhabited area without road network obstacle distribution map, mark the impassable areas as impassable in the gridded terrain map. The impassable areas specifically include cliffs, deep water areas, and dense vegetation areas; areas that are passable but have special restrictions, including shallow water areas and gravel areas, are marked as restricted access areas, and the access restrictions of the restricted access areas are recorded, including maximum speed limit, minimum speed limit, and load limit; output a gridded terrain map marked with terrain type, elevation value, and obstacles.

[0019] Perform regional division and output the sub-region map after division. The specific analysis process is as follows: divide the grid terrain map marked with terrain type, elevation value, and obstacles into regions; select grid cells with elevation change less than 10% and terrain type of flat land or gentle slope as vehicle passable areas; select grid cells with terrain type of mountainous areas, canyons, and unknown terrain type areas as drone reconnaissance areas; use material handover points as areas where vehicles and drones need to collaborate to complete tasks, and record them as collaborative operation areas; output the sub-region map after division, and mark the vehicle passable areas, drone reconnaissance areas, and collaborative operation areas.

[0020] Through multi-source data fusion and refined annotation, the terrain characteristics of the uninhabited area are accurately portrayed, laying a solid foundation for subsequent path planning. By integrating topographic maps, elevation data maps, and obstacle distribution maps, a high-precision terrain model is constructed with 10m×10m grid cells. Various terrain types, such as flat land and high slopes, as well as inaccessible areas such as cliffs and deep water, are clearly marked. At the same time, access restrictions are recorded for restricted areas such as shallow water, ensuring comprehensive and accurate terrain information. On this basis, regions are divided based on terrain characteristics and mission requirements. Areas with small elevation changes and flat terrain are designated as vehicle-accessible areas, complex terrain is designated as drone reconnaissance areas, and material handover points are designated as collaborative operation areas. This allows vehicles and drones to perform their respective functions, effectively avoiding terrain risks, reducing planning complexity, improving the feasibility and efficiency of path planning, and ensuring the efficient and safe execution of missions in uninhabited areas.

[0021] Collect vehicle parameters, starting point and target point coordinates, combine them with the divided sub-area map, perform vehicle path planning, and output the preliminary path planning results of the vehicle.

[0022] The specific analysis process is as follows: collect vehicle parameters, starting point and target point coordinates, vehicle parameters specifically include maximum speed, drivable terrain type, vehicle weight; analyze the vehicle passable area. If there is an impassable area, the impassable area will be removed from the vehicle passable area. If there is a restricted passable area, obtain the restricted passable area access restriction conditions. If the vehicle can meet the restricted passable area access restriction conditions, the restricted passable area in the vehicle passable area will be retained. If the vehicle cannot meet the restricted passable area access restriction conditions, the restricted passable area will be removed from the vehicle passable area to obtain an updated vehicle passable area; use the updated vehicle passable area as the search space, and the starting point and target point as the starting and end points of the search; use a heuristic search algorithm, combined with vehicle parameters to perform path search, which must meet the vehicle parameters; during the search process, calculate the cost function for each grid cell: obtain the terrain type-terrain cost mapping set stored in the database, and determine the matching terrain cost based on the terrain type; Obtaining the precipitation amount and wind speed of the uninhabited area without a road network environment; marking the precipitation amount and wind speed of the uninhabited area without a road network environment as vehicle cost impact features; obtaining a cost impact feature-vehicle cost weight mapping table stored in a database; determining a matching vehicle cost weight based on the current cost impact feature, wherein the vehicle cost weight specifically includes a weight coefficient of the terrain cost and a weight coefficient of the vehicle energy consumption; The vehicle cost function is: vehicle cost = vehicle distance + α × terrain cost + β × vehicle energy consumption, where α is the weight coefficient of the terrain cost, β is the weight coefficient of the vehicle energy consumption, vehicle energy consumption is the energy consumed by the vehicle passing through the grid unit, which is dimensionless, and vehicle distance is the distance the vehicle passes through the grid unit, which is dimensionless; the path with the minimum vehicle cost is selected as the vehicle preliminary path planning result; the vehicle preliminary path planning result is smoothed and the vehicle preliminary path planning result is output.

[0023] Specifically, the following steps are included: Initialization: Add the starting point to the open list (the set of nodes to be searched), record the actual cost (g value) and heuristic estimated cost (h value) to the starting point, and the h value is usually based on the Manhattan distance or Euclidean distance to estimate the shortest distance from the current node to the target point.

[0024] Node extension: Select the node with the smallest total cost (f value = g value + h value) from the open list as the current node.

[0025] Traverse the adjacent grids of the current node (such as the four-neighborhood or eight-neighborhood of the top, bottom, left, and right) to check whether the adjacent nodes are in the passable area and whether the terrain type meets the vehicle's drivable conditions.

[0026] For adjacent nodes that meet the conditions, calculate their g value (current node g value + distance cost from adjacent node to current node) and update the h value (heuristic function estimates the distance to the target point).

[0027] Cost function integration: When calculating the g value of adjacent nodes, the vehicle cost function (vehicle distance + α × terrain cost + β × vehicle energy consumption) is combined, where: Vehicle distance: The grid distance between the adjacent node and the current node (dimensionless, such as 1 grid distance is 1).

[0028] Terrain cost: Get the corresponding cost value from the database mapping set according to the terrain type of the adjacent node.

[0029] Vehicle Energy Consumption: Calculates the energy consumption (dimensionless) required to pass through the grid based on vehicle parameters (such as weight, speed) and terrain type.

[0030] Loop and terminate: Repeat the node expansion process until the target point is added to the open list or the open list is empty (no solution). When the target point is selected, the path is backtracked to obtain the optimal path from the starting point to the target point.

[0031] Each time a node is expanded, it is necessary to verify that the terrain type of the node is within the range of vehicle-travelable terrain types (for example, if the vehicle cannot pass through forest terrain, the node will be skipped). If the node is in a restricted access area (such as shallow water), it is necessary to verify that the vehicle's maximum speed, weight, and other parameters meet the access restrictions of the area (for example, speed ≤ 10 km / h, load ≤ 500 kg). If not, the node will be skipped.

[0032] After the search is complete, the path with the lowest total cost (calculated based on the cost function) is selected as the initial vehicle path planning result. The coordinate point sequence in the path is smoothed (the original paper does not specify the algorithm, but it can be understood as reducing sudden changes at path turning points through curve interpolation) to output a path coordinate point sequence that conforms to the vehicle's driving dynamics.

[0033] Through parameter adaptation, area screening, and cost optimization, accurate path planning is achieved in the complex environment of no-man's land. Its advantages are: combining parameters such as the maximum vehicle speed and the type of drivable terrain, dynamically screening the vehicle's traversable areas, eliminating impassable areas, and further optimizing the feasible space based on whether the vehicle can meet the conditions of the restricted traversable area, ensuring that the planned path meets the vehicle's actual traversability; using the updated area as the search space, using a heuristic search algorithm, and integrating the terrain type-terrain cost mapping set to construct a cost function, comprehensively considering factors such as vehicle distance, terrain cost, and energy consumption, balancing path length and driving cost, and selecting the optimal path; finally, smoothing the path to reduce unreasonable planning such as frequent turns and sudden braking, improving path safety and vehicle driving stability, thereby providing an efficient, safe, and economical path solution for vehicle driving in no-man's land.

[0034] Collect the UAV performance parameters, starting point and target point coordinates, combine them with the divided sub-area map, perform UAV path planning, and output the UAV's preliminary path planning results.

[0035] The specific analysis process is as follows: collecting UAV performance parameters, starting point and target point coordinates, including maximum flight speed and flight altitude range; using the UAV's reconnaissance area as the search space, and the starting point and target point as the starting and ending points of the search; using the Dijkstra path planning algorithm, combined with the UAV performance parameters, to search for paths that meet the UAV performance parameters; during the search process, calculating the cost function for each grid cell: obtaining the current UAV flight speed and UAV flight wind direction, storing the UAV flight speed and UAV flight wind direction as specified labels, obtaining the specified label-wind speed cost mapping set stored in the database, and determining the matching wind speed cost based on the specified label; Obtain the precipitation amount in the roadless environment of the uninhabited area, the wind direction in the roadless environment of the uninhabited area, and the drone payload; mark the precipitation amount in the roadless environment of the uninhabited area, the wind direction in the roadless environment of the uninhabited area, and the drone payload as drone cost impact features; obtain a cost impact feature-drone cost weight mapping table stored in the database; determine the matching drone cost weight based on the current cost impact feature, where the drone cost weight specifically includes a weight coefficient of the wind speed cost and a weight coefficient of the drone energy consumption; The drone cost function is: drone cost = drone distance + γ × wind speed cost + δ × drone energy consumption, where γ is the weight coefficient of wind speed cost, δ is the weight coefficient of drone energy consumption, drone energy consumption is the energy consumed by the drone passing through the grid unit, dimensionless, and distance is the distance the drone passes through the grid unit, dimensionless; the path with the lowest drone cost is selected as the drone's preliminary path planning result; the drone's preliminary path planning result is highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collisions; the drone's preliminary path planning result is output.

[0036] Specifically, the following steps are included: Determine the search space and drone performance parameter constraints: Search space: The search space is based on the drone's reconnaissance area (terrain types such as mountains, canyons, and grid cells of unknown terrain), and the starting point and target point are the starting and ending coordinates of the path planning.

[0037] Maximum flight speed: The flight speed of any grid cell in the path must not exceed the maximum flight speed of the drone (for example, if the maximum speed of the drone is 20m / s, the speed of each section in the path must be ≤20m / s). Flight altitude range: The flight altitude must be within the range allowed by the drone (for example, the minimum flight altitude is 10m and the maximum flight altitude is 200m) to avoid collisions with terrain or obstacles.

[0038] initialization: Create an open list (nodes to be searched) and a closed list (nodes already searched). Add the starting point to the open list and set its initial cost to 0. Set the initial costs of all other nodes to infinity. Maintain a predecessor pointer for each node for path backtracking.

[0039] Node extension: Select the node with the smallest cost from the open list as the current node and move it to the closed list.

[0040] Traverse the adjacent grids of the current node (such as the four-neighborhood or eight-neighborhood) to check whether the adjacent nodes are within the drone's reconnaissance area and whether the flight altitude meets the terrain and obstacle avoidance requirements (processed in the subsequent altitude optimization step).

[0041] For adjacent nodes that meet the conditions, calculate the new cost from the starting point through the current node to the adjacent node.

[0042] Cost function calculation: Drone distance: The grid distance between the adjacent node and the current node (dimensionless, such as 1 grid distance is 1).

[0043] Wind Speed Cost: Obtain the current UAV flight speed and wind direction, combine them into a "specified tag" (e.g., "15m / s_northeast wind"). Match the corresponding wind speed cost value from the "specified tag-wind speed cost mapping set" in the database (e.g., the wind speed cost corresponding to the tag is 3).

[0044] Drone energy consumption: Based on the real-time precipitation, wind direction, and drone payload in the uninhabited area, the wind speed cost weight coefficient γ and energy consumption weight coefficient δ are obtained from the "cost impact characteristics-drone cost weight mapping table", and the energy consumption (dimensionless) passing through the grid is calculated in combination with the drone dynamics model.

[0045] Total cost formula: Drone cost = Drone distance + γ × wind speed cost + δ × Drone energy consumption.

[0046] Looping and Termination: Repeat the node expansion process until the target point is added to the closed list or the open list is empty (no solution). When the target point is processed, backtrack through the predecessor node pointer to find the minimum cost path from the starting point to the target point.

[0047] Maximum flight speed check: When calculating the cost of adjacent nodes, if the current flight speed exceeds the maximum speed of the drone, the node will be skipped (for example, if the maximum speed of the drone is 20m / s and the current speed is calculated to be 25m / s, the path is not feasible). Flight altitude pre-check: Although altitude optimization is handled in subsequent steps, during path search, it is necessary to ensure that the flight altitude of the preliminary path is not lower than the highest point of the terrain + the safety distance (for example, if the terrain elevation is 100m and the safety distance is 10m, the flight altitude should be ≥110m) to avoid obvious collision risks on the preliminary path.

[0048] After the Dijkstra algorithm traverses all feasible paths, it selects the path with the minimum total cost as the preliminary path planning result of the UAV, and generates the preliminary path planning result of the UAV including the path coordinate point sequence, the flight altitude of each point, the expected flight time, and the total energy consumption.

[0049] By combining drone performance with environmental factors, safe and efficient flight planning is achieved in complex, uninhabited areas. Using performance parameters such as the drone's maximum flight speed and altitude range as constraints, the drone's reconnaissance area is precisely targeted as the search space, ensuring that the planned path meets the drone's actual flight capabilities. Using the Dijkstra path planning algorithm, a cost function is constructed that integrates dynamic factors such as flight speed and wind direction to quantify the flight cost under different environmental conditions, balancing flight distance, wind speed impact, and energy consumption to select the optimal path. Based on the preliminary planning results, height optimization is performed based on terrain and obstacle distribution, and flight altitude is dynamically adjusted to effectively avoid collision risks. This fully utilizes the drone's aerial reconnaissance advantages while ensuring its flight safety in complex environments, providing a precise and reliable path plan for drone missions.

[0050] Obtain the vehicle and drone mission types, combine the preliminary path planning results of the vehicle and the drone, perform collaborative task allocation, and output the optimized vehicle and drone collaborative path planning results.

[0051] The specific analysis process is as follows: obtain the vehicle and UAV task types, combine the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV, and output the collaborative task allocation table; according to the collaborative task allocation table, the UAV flies in front of the vehicle, scouts the road conditions in advance and transmits information to the vehicle in real time. The vehicle adjusts the driving path according to the feedback from the UAV. If the UAV detects that the path is blocked, it obtains the alternative detour path stored in the database. The path obstruction types include landslides and the presence of other vehicles; recalculate the vehicle cost function and the UAV cost function for each grid cell of the alternative detour path, and selects the path with the minimum vehicle cost. The path is taken as the vehicle collaborative path planning result of the alternative detour path, and the path with the minimum cost of the drone is selected as the drone collaborative path planning result of the alternative detour path; the vehicle collaborative path planning result of the alternative detour path is smoothed, and the drone collaborative path planning result of the alternative detour path is highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collision; the optimized vehicle and drone collaborative path planning result is output, and the optimized vehicle and drone path coordinate point sequence, the estimated driving time of the planned path, and the total energy consumption of the path planning are obtained based on the optimized vehicle and drone collaborative path planning result.

[0052] Obtain the vehicle and UAV task types, combine the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV, and output the collaborative task allocation table. The specific analysis process is as follows: obtain the vehicle and UAV task types, and decompose the task type into multiple subtasks based on the predefined task splitting logic stored in the database; based on the preliminary path planning results of the vehicle and the preliminary path planning results of the UAV, project the vertical reconnaissance range of the UAV path to the ground, perform geometric intersection calculation with the vehicle path, calculate the estimated time for the vehicle and UAV to reach the overlapping area, analyze the overlapping area of the vehicle and UAV in time and space, assign the subtasks to the most suitable equipment, and obtain the corresponding requirements of the subtasks stored in the database: if the subtask requires a high-altitude perspective and is time-sensitive, then assign To the drone; if the subtask requires load or long-term endurance, it is assigned to the vehicle; if the subtask is in the spatiotemporal overlap area and requires dual-device collaboration, the drone and vehicle will jointly execute it; determine the execution subject and time node of each subtask; based on ROS+Gazebo, simulate spatiotemporal overlap and task allocation, and visualize the path in real time; based on the Python library, use Shapely for geometric analysis and Pandas for spatiotemporal table processing; verify the subtask completion rate and determine whether all subtasks are assigned and there are no conflicts: if there is a conflict, compare the corresponding priorities of the subtasks stored in the database, and give priority to the subtask with higher priority; if the conflicting subtasks have equal priorities, compare the total energy consumption of the completed subtasks, and give priority to the subtask with lower total energy consumption; The drone's built-in sensor collects the drone's remaining power in real time. When the drone's remaining power is lower than the preset power threshold stored in the database, it triggers a charging request to the drone's nest. The preset power threshold = the basic power threshold stored in the database + the power consumption of the drone's real-time corresponding position when returning. When the drone's remaining power is lower than the preset threshold stored in the database, the vehicle automatically drives to the nearest safe stop, which is a flat ground with an elevation change of less than 5%, and calls the vehicle's position coordinates. With the drone's current position as the starting point and the vehicle's position as the end point, the Dijkstra algorithm is used to plan the shortest path, avoiding areas with wind speeds above 10m / s. The drone returns to the drone's nest on the vehicle along the shortest path for charging. The vehicle pauses during charging and marks the vehicle speed as 0. After charging is completed, the drone continues to execute the subtask from the breakpoint. If there is no conflict, the verification is passed. The collaborative task allocation table is output, which contains the execution subject and time node of each subtask.

[0053] Through task decomposition, intelligent allocation, and dynamic collaboration, vehicles and drones achieve efficient collaboration in unmanned areas. Based on predefined task splitting logic, tasks are broken down into subtasks. Combined with the preliminary path planning results for the vehicle and drone, geometric intersection and spatiotemporal analysis accurately identify overlapping areas, ensuring that task allocation aligns with equipment capabilities. Execution entities are intelligently matched based on subtask characteristics (such as high-altitude perspective and payload requirements), prioritizing high-priority or low-energy tasks. Simultaneously, ROS+Gazebo simulation and Python library analysis ensure conflict-free and visual task allocation. During execution, the drone scouts road conditions in advance and provides real-time feedback. When encountering obstructions, it quickly calculates alternative detours, simultaneously optimizes the path costs of both the vehicle and drone, and performs smoothing and altitude adjustments. Furthermore, the vehicle's onboard pod provides charging support for the drone, addressing its battery life limitations. The resulting optimized output includes path coordinates, estimated duration, and total energy consumption, significantly improving the collaborative efficiency, reliability, and resource utilization of task execution in unmanned areas.

[0054] The real-time environmental information of the roadless area is monitored, and the optimized vehicle and UAV collaborative path planning results are dynamically adjusted to obtain the dynamically adjusted vehicle and UAV path planning results.

[0055] The specific analysis process is: monitoring the real-time environmental information of the uninhabited area without road network, obtaining the real-time environmental information data of the uninhabited area without road network, specifically including the real-time weather data of the uninhabited area without road network and the real-time terrain data of the uninhabited area without road network, among which: the real-time weather data of the uninhabited area without road network specifically includes the real-time wind speed of the uninhabited area without road network, the real-time wind direction of the uninhabited area without road network and the real-time precipitation of the uninhabited area without road network; the real-time terrain data of the uninhabited area without road network specifically includes the real-time terrain landslide situation of the uninhabited area without road network and the real-time mud-rock flow situation of the uninhabited area without road network. If there is a real-time terrain landslide in the uninhabited area without road network, obtain the real-time terrain landslide impact range and the real-time terrain landslide location coordinates of the uninhabited area without road network. If there is a real-time mud-rock flow in the uninhabited area without road network, obtain the real-time mud-rock flow impact range and the real-time mud-rock flow coordinates of the uninhabited area without road network.

[0056] The optimized vehicle and UAV collaborative path planning results are dynamically adjusted to obtain dynamically adjusted vehicle and UAV path planning results. The specific analysis process is: based on the real-time environmental information data of the unmanned area without road network, the paths of the vehicle and UAV are dynamically adjusted: if the real-time wind speed of the unmanned area without road network ahead exceeds the wind speed threshold stored in the database, or the real-time wind direction of the unmanned area without road network exceeds the wind direction threshold stored in the database, or the real-time precipitation of the unmanned area without road network exceeds the precipitation threshold stored in the database, or there is a real-time terrain collapse in the unmanned area without road network, or there is a real-time mudslide in the unmanned area without road network, the vehicle replans the detour path, and the UAV scouts the new feasible path in advance; selects the detour path with the shortest distance, and ensures that the path is still within the passable area of the vehicle and UAV; recalculates each distance The vehicle cost function and the UAV cost function are calculated for each grid cell of the shortest detour path, and the path with the minimum vehicle cost is selected as the vehicle detour collaborative path planning result, and the path with the minimum UAV cost is selected as the UAV detour collaborative path planning result; the vehicle detour collaborative path planning result is smoothed, and the UAV detour collaborative path planning result is highly optimized, and the flight altitude is adjusted according to the terrain and obstacle distribution to avoid collisions; based on the terrain type corresponding to the vehicle path coordinate point sequence, the matching vehicle speed stored in the database is determined; and the dynamically adjusted vehicle and UAV path planning results are obtained, including the path coordinate point sequence of the vehicle and UAV, the UAV flight altitude, the terrain type corresponding to the path coordinate point sequence, and the vehicle speed corresponding to the vehicle path coordinate point sequence.

[0057] Through real-time environmental monitoring and intelligent path correction, the adaptability and mission reliability of vehicles and drones in the complex environments of uninhabited areas are significantly improved. Comprehensive real-time weather (wind speed, wind direction, precipitation) and terrain (landslides, mudslides) data is collected to accurately capture environmental changes. When environmental parameters exceed thresholds or geological disasters occur, a coordinated response between the vehicle and drone is quickly triggered. The vehicle replans a detour, while the drone leverages its aerial reconnaissance capabilities to explore the route ahead, ensuring that the new route is both within the equipment's traversable area and minimizes distance. By recalculating the cost function, the optimal detour paths for the vehicle and drone are selected, and smoothing and height optimization are performed separately to effectively avoid collision risks. Furthermore, vehicle speeds are dynamically adjusted based on terrain type to further ensure driving safety. This mechanism enables the collaborative system to dynamically respond to emergencies, reduce the risk of mission interruption, and improve the stability and efficiency of operations in uninhabited areas.

[0058] Based on the dynamically adjusted vehicle and drone path planning results, risk prediction is performed and intelligent decision-making recommendations are generated.

[0059] The specific analysis process is as follows: the dynamically adjusted vehicle and UAV path planning results are used as a trained fault tree analysis model, combined with Monte Carlo simulation, to output the risk probability of the dynamically adjusted vehicle and UAV path planning results and the corresponding main risk sources; the risk probability of the dynamically adjusted vehicle and UAV path planning results and the corresponding main risk sources are stored as specified labels, and the specified label-intelligent decision suggestion mapping table stored in the database is obtained, and the corresponding intelligent decision suggestion is matched based on the current specified label.

[0060] The fault tree analysis and Monte Carlo simulation process is as follows: Input parameter definition: Dynamically adjusted path planning results: including the path coordinate point sequence of the vehicle and drone, terrain type at each point, flight altitude, estimated driving / flight time, etc. Fault tree model: trained fault tree structure and basic event probability parameters.

[0061] Monte Carlo simulation iterative process: Random sampling: Each random variable (such as the wind speed in the current path area) is sampled according to its probability distribution to generate a set of specific environmental and equipment status parameters.

[0062] Fault tree traversal: Based on the sampling parameters, determine whether the basic event has occurred (for example, if the wind speed sampling value exceeds the threshold, a "strong wind event" occurs), and calculate the occurrence status (yes / no) of the top event (path failure) through logic gate transfer.

[0063] Risk source record: If a top event occurs, trace back the fault tree and record the basic event chain that led to the failure (such as "strong wind → unstable drone flight → collision with obstacles") to determine the main risk source.

[0064] Usually 10 4 to 10 6 Independent simulations are performed to ensure the convergence of probability calculations.

[0065] Risk probability and risk source statistics: Risk probability calculation: The ratio of the number of top events (path failures) to the total number of simulations is calculated as the risk probability of the dynamically adjusted path.

[0066] Identification of major risk sources: The frequency of basic events leading to the top event across all simulations is calculated. The top N most frequent basic events (e.g., the top three) are identified as the major risk sources. For example, in 100,000 simulations, if "landscape" caused 15,000 failures, "strong wind" caused 12,000 failures, and "drone battery low" caused 8,000 failures, then the major risk sources are "landscape" and "strong wind." This risk prediction and decision-making mechanism, based on fault tree analysis and Monte Carlo simulation, provides a reliable risk management solution for operations in unmanned areas by quantifying risks and providing intelligent matching recommendations. With the help of a trained fault tree analysis model combined with Monte Carlo simulation, a systematic risk assessment is conducted on the dynamically adjusted path planning results. This can accurately calculate the risk probability and locate major risk sources such as sudden terrain changes and extreme weather. By mapping and matching the risk assessment results with the intelligent decision-making suggestions in the database, targeted response strategies can be quickly output, such as adjusting driving speed, changing flight altitude, or suspending the mission. This method transforms complex risk scenarios into quantifiable and actionable decision-making basis, helping vehicles and drones avoid potential threats in advance, reducing the probability of accidents, and at the same time reducing human decision-making errors, significantly improving the safety and reliability of mission execution in uninhabited areas, and providing scientific and efficient decision-making support for emergency response and resource scheduling.

Claims

1. A method for collaborative path planning between vehicles and drones in an uninhabited area without a road network, characterized in that: The following steps are involved: Obtaining terrain data of an uninhabited area without a road network, preprocessing the terrain data of the uninhabited area without a road network, obtaining a gridded terrain map, performing regional division, and outputting a sub-region map after division; Collect vehicle parameters, starting point and destination point coordinates, combine them with the divided sub-area map, perform vehicle path planning, and output the preliminary path planning results of the vehicle; Collect the drone's performance parameters, starting point and target point coordinates, combine them with the divided sub-area map, perform drone path planning, and output the drone's preliminary path planning results; Obtain the vehicle and drone mission types, combine the preliminary path planning results of the vehicle and the drone, perform collaborative task allocation, and output the optimized vehicle and drone collaborative path planning results; Monitor the real-time environmental information of the roadless area and the unmanned area, dynamically adjust the optimized vehicle and UAV collaborative path planning results, and obtain the dynamically adjusted vehicle and UAV path planning results; Based on the dynamically adjusted vehicle and drone path planning results, risk prediction is performed and intelligent decision-making recommendations are generated.

2. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 1, characterized in that: The terrain data of the uninhabited area without road network is obtained, and the terrain data of the uninhabited area without road network is preprocessed to obtain a gridded terrain map. The specific analysis process is as follows: Obtain terrain data of uninhabited areas without road networks, including terrain maps of uninhabited areas without road networks, elevation data maps of uninhabited areas without road networks, and obstacle distribution maps of uninhabited areas without road networks; The uninhabited area and road network topographic map is divided into regular grid cells, each grid cell is 10m×10m; Assign a terrain type to each grid cell and record the elevation value of each grid cell based on the uninhabited area without road network elevation data map. The terrain types include flat land, high slope, gentle slope, river, forest, mountain, canyon, and unknown terrain type area. Based on the distribution map of obstacles in uninhabited areas without road networks, impassable areas are marked as impassable on the grid terrain map. Impassable areas include cliffs, deep water areas, and dense vegetation areas. Areas that are accessible but have special restrictions, such as shallow water areas and gravel areas, are marked as restricted access areas, and the access restrictions in the restricted access areas are recorded, including maximum speed limits, minimum speed limits, and load limits. Output is a gridded terrain map with terrain type, elevation value, and obstacles marked.

3. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 2, characterized in that: The region division is performed and the sub-region map after division is output. The specific analysis process is as follows: Divide the gridded terrain map into regions with terrain types, elevation values, and obstacles marked; Grid cells with elevation changes less than 10% and flat or gently sloping terrain are selected as vehicle-accessible areas; Select grid cells with terrain types such as mountainous areas, canyons, and unknown terrain types as the drone-reconnaissance area; The material handover point is regarded as the area where vehicles and drones need to collaborate to complete the task, which is recorded as the collaborative operation area; Output the divided sub-area map, marking the vehicle-passable area, drone-reconnaissance area, and collaborative operation area.

4. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 1, characterized in that: The vehicle parameters, starting point and target point coordinates are collected, and the vehicle path planning is performed in combination with the divided sub-area map, and the preliminary path planning result of the vehicle is output. The specific analysis process is as follows: Collect vehicle parameters, starting point and destination point coordinates. Vehicle parameters include maximum speed, drivable terrain type, and vehicle weight. Analyze the vehicle passable area. If there is an impassable area, remove the impassable area from the vehicle passable area. If there is a restricted passable area, obtain the access restriction conditions of the restricted passable area. If the vehicle can meet the access restriction conditions of the restricted passable area, retain the restricted passable area in the vehicle passable area. If the vehicle cannot meet the access restriction conditions of the restricted passable area, remove the restricted passable area from the vehicle passable area to obtain an updated vehicle passable area. The updated vehicle-accessible area is used as the search space, and the starting point and target point are used as the starting point and end point of the search; Use heuristic search algorithm to search for paths in combination with vehicle parameters, which must be met; During the search, a cost function is calculated for each grid cell: Obtain the terrain type-terrain cost mapping set stored in the database, and determine the matching terrain cost based on the terrain type; Obtain the precipitation amount and wind speed in the uninhabited area without road network; The precipitation in the uninhabited area without road network and the wind speed in the uninhabited area without road network are marked as vehicle cost impact features; Obtaining a cost impact feature-vehicle cost weight mapping table stored in a database; Based on the current cost impact characteristics, determine the matching vehicle cost weight, which specifically includes the weight coefficient of terrain cost and the weight coefficient of vehicle energy consumption; The vehicle cost function is: vehicle cost = vehicle distance + α × terrain cost + β × vehicle energy consumption, where α is the weight coefficient of terrain cost, β is the weight coefficient of vehicle energy consumption, vehicle energy consumption is the energy consumed by the vehicle passing through the grid cell, dimensionless, and vehicle distance is the distance the vehicle passes through the grid cell, dimensionless; Select the path with the minimum vehicle cost as the vehicle's preliminary path planning result; The preliminary path planning results of the vehicle are smoothed and output.

5. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 1, characterized in that: The drone performance parameters, starting point and target point coordinates are collected, and the drone path planning is performed in combination with the divided sub-area map, and the preliminary path planning results of the drone are output. The specific analysis process is as follows: Collect UAV performance parameters, starting point and target point coordinates. UAV performance parameters include maximum flight speed and flight altitude range; The drone's reconnaissance area is used as the search space, and the starting point and target point are used as the starting and ending points of the search; Use Dijkstra path planning algorithm and combine it with the performance parameters of the UAV to search for paths, which must meet the performance parameters of the UAV; During the search, a cost function is calculated for each grid cell: Obtain the current UAV flight speed and UAV flight wind direction, store the UAV flight speed and UAV flight wind direction as specified labels, obtain the specified label-wind speed cost mapping set stored in the database, and determine the matching wind speed cost based on the specified label; Obtain precipitation in uninhabited areas without road networks, wind direction in uninhabited areas without road networks, and drone payload; The precipitation in the roadless environment, the wind direction in the roadless environment, and the drone payload are marked as the drone cost influencing features; Obtain the cost impact feature-drone cost weight mapping table stored in the database; Based on the current cost impact characteristics, the matching drone cost weight is determined. The drone cost weight specifically includes the weight coefficient of wind speed cost and the weight coefficient of drone energy consumption; The drone cost function is: drone cost = drone distance + γ × wind speed cost + δ × drone energy consumption, where γ is the weight coefficient of wind speed cost, δ is the weight coefficient of drone energy consumption, drone energy consumption is the energy consumed by the drone passing through the grid cell, dimensionless, and distance is the distance the drone passes through the grid cell, dimensionless. Select the path with the minimum cost for the UAV as the initial path planning result of the UAV; Highly optimize the initial path planning results of the drone, adjust the flight altitude according to the terrain and obstacle distribution, and avoid collisions; Output the preliminary path planning results of the UAV.

6. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 1, characterized in that: The vehicle and drone mission types are obtained, and the preliminary path planning results of the vehicle and the drone are combined to perform collaborative task allocation, and the optimized vehicle and drone collaborative path planning results are output. The specific analysis process is as follows: Obtain the vehicle and drone mission types, combine the vehicle's preliminary path planning results with the drone's preliminary path planning results, and output a collaborative task allocation table; According to the collaborative task allocation table, the drone flies ahead of the vehicle, scouting road conditions in advance and transmitting information to the vehicle in real time. The vehicle adjusts its driving path based on the drone's feedback. If the drone detects a blocked path, it obtains alternative detours stored in the database. Path obstruction types include landslides and the presence of other vehicles. Recalculate the vehicle cost function and the UAV cost function for each grid cell of the alternative detour path, select the path with the minimum vehicle cost as the vehicle collaborative path planning result of the alternative detour path, and select the path with the minimum UAV cost as the UAV collaborative path planning result of the alternative detour path; Smoothing the collaborative path planning results of vehicles with alternative detours and highly optimizing the collaborative path planning results of drones with alternative detours, adjusting flight altitude based on terrain and obstacle distribution to avoid collisions; Output the optimized vehicle and UAV collaborative path planning results, and based on the optimized vehicle and UAV collaborative path planning results, obtain the optimized vehicle and UAV path coordinate point sequence, the estimated driving time of the planned path, and the total energy consumption of the path planning.

7. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 6, characterized in that: The vehicle and drone mission types are obtained, and the preliminary path planning results of the vehicle and the drone are combined to output a collaborative task allocation table. The specific analysis process is as follows: Obtain vehicle and drone mission types, and decompose the mission types into multiple subtasks based on the predefined task splitting logic stored in the database; Based on the preliminary path planning results for the vehicle and the drone, the vertical reconnaissance range of the drone's path is projected onto the ground, and a geometric intersection calculation is performed with the vehicle's path. The estimated time for the vehicle and drone to reach the overlapping area is calculated. The overlapping area of the vehicle and drone in time and space is analyzed, and the subtasks are assigned to the most suitable devices. The requirements corresponding to the subtasks stored in the database are obtained: If the subtask requires a high-altitude perspective and is time-sensitive, it is assigned to the drone; If the subtask requires heavy load or long endurance, it is assigned to the vehicle; If the subtask is in a spatiotemporal overlapping area and requires dual-device collaboration, the UAV and vehicle will perform it jointly; Determine the execution entity and time node of each subtask; Based on ROS+Gazebo, simulate spatiotemporal overlap and task allocation, and visualize paths in real time; Based on Python libraries, Shapely is used for geometric analysis and Pandas is used for spatiotemporal table processing; Verify the subtask completion rate and determine whether all subtasks are assigned and there are no conflicts: If there is a conflict, the corresponding priorities of the subtasks stored in the database are compared, and the subtask with higher priority is completed first. If the conflicting subtasks have equal priorities, the total energy consumption of the subtasks is compared, and the subtask with lower total energy consumption is completed first. The drone's built-in sensor collects the drone's remaining power in real time. When the drone's remaining power falls below the preset power threshold stored in the database, it triggers a request for charging from the drone's nest. The preset power threshold is the basic power threshold stored in the database plus the power consumption of the drone's real-time corresponding position when returning to the destination. When the remaining battery power of the drone falls below a preset threshold stored in the database, the vehicle automatically drives to the nearest safe stopping point, which is flat ground with an elevation change of less than 5%, and calls the vehicle's position coordinates; Using the current position of the drone as the starting point and the vehicle's position as the end point, the Dijkstra algorithm is used to plan the shortest path, avoiding areas with wind speeds exceeding 10m / s. The drone returns to the drone nest on the vehicle via the shortest path for charging; The vehicle pauses during charging and the vehicle speed is marked as 0. After charging is completed, the drone continues to execute the subtask from the breakpoint; If there is no conflict, the verification passes; Output collaborative task allocation table, including the execution entity and time node of each subtask.

8. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 1, characterized in that: The specific analysis process of monitoring the real-time environmental information of the roadless area is as follows: Monitor the real-time environmental information of the no-man's land without road network to obtain real-time environmental information data of the no-man's land without road network, specifically including real-time weather data of the no-man's land without road network and real-time terrain data of the no-man's land without road network, wherein: The real-time weather data for uninhabited areas without road networks specifically include real-time wind speed, real-time wind direction, and real-time precipitation in uninhabited areas without road networks; The real-time terrain data of the uninhabited area with no road network specifically includes the real-time terrain landslide situation and the real-time mud and rock flow situation in the uninhabited area with no road network. If there is a real-time terrain landslide in the uninhabited area with no road network, the influence range of the real-time terrain landslide in the uninhabited area with no road network and the real-time terrain landslide location coordinates of the uninhabited area with no road network are obtained. If there is a real-time mud and rock flow in the uninhabited area with no road network, the influence range of the real-time mud and rock flow in the uninhabited area with no road network and the real-time mud and rock flow coordinates of the real-time mud and rock flow in the uninhabited area with no road network are obtained.

9. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle in an uninhabited area without a road network according to claim 8, characterized in that: The optimized vehicle and UAV collaborative path planning results are dynamically adjusted to obtain dynamically adjusted vehicle and UAV path planning results. The specific analysis process is as follows: Based on real-time environmental information data of uninhabited areas without roads, the paths of vehicles and drones are dynamically adjusted: If the real-time wind speed in the uninhabited area with no road network ahead exceeds the wind speed threshold stored in the database, or the real-time wind direction in the uninhabited area with no road network exceeds the wind direction threshold stored in the database, or the real-time precipitation in the uninhabited area with no road network exceeds the precipitation threshold stored in the database, or there is a real-time terrain collapse in the uninhabited area with no road network, or there is a real-time debris flow in the uninhabited area with no road network, the vehicle will replan a detour route, and the drone will scout a new feasible route in advance; Choose the shortest detour route while ensuring it is still within the traversable area for vehicles and drones; Recalculate the vehicle cost function and the UAV cost function for each grid cell of each shortest detour path, select the path with the minimum vehicle cost as the vehicle detour collaborative path planning result, and select the path with the minimum UAV cost as the UAV detour collaborative path planning result; Smoothing the collaborative path planning results for vehicles and highly optimizing those for drones, adjusting flight altitude based on terrain and obstacle distribution to avoid collisions; Determining a matching vehicle speed stored in a database based on a terrain type corresponding to the sequence of vehicle path coordinate points; The dynamically adjusted vehicle and UAV path planning results are obtained, including the path coordinate point sequences of the vehicle and UAV, the flight altitude of the UAV, the terrain type corresponding to the path coordinate point sequence, and the vehicle speed corresponding to the vehicle path coordinate point sequence.

10. The method for collaborative path planning between vehicles and drones in an uninhabited area without a road network according to claim 1, characterized in that: Based on the dynamically adjusted vehicle and drone path planning results, risk prediction is performed to generate intelligent decision-making suggestions. The specific analysis process is as follows: The dynamically adjusted vehicle and UAV path planning results are used as the trained fault tree analysis model, combined with Monte Carlo simulation to output the risk probability and corresponding main risk sources of the dynamically adjusted vehicle and UAV path planning results; The risk probabilities of the dynamically adjusted vehicle and drone path planning results and the corresponding main risk sources are stored as specified tags. The specified tag-intelligent decision suggestion mapping table stored in the database is obtained, and the corresponding intelligent decision suggestion is matched based on the current specified tag.

Images