Cooperative distribution method for unmanned vehicle and unmanned aerial vehicle in complex field environment

Abstract

The invention discloses an unmanned vehicle and unmanned aerial vehicle cooperative distribution method in a complex field environment. The method comprises the following steps: acquiring data of ground terrain parameters; based on the initial obstacle density evaluation, high-density obstacle areas which are not suitable for vehicle passing are eliminated; obstacle recognition constraint is carried out by using a Transform model, obstacle features are confirmed, and constraint calculation is carried out; calculating a trafficability score of each region; calculating the trafficability score and the passing time of the path by using a bilevel programming model, and finding out an optimal path; when the path of the unmanned vehicle is not feasible, the unmanned aerial vehicle takes over the task and generates an optimal path according to the surrounding environment and the distribution distance; the map generation module generates a map for displaying the optimal path; and the unmanned aerial vehicle automatically takes goods and completes a distribution task according to the map. According to the invention, by combining the cooperative capability of the unmanned vehicle and the unmanned aerial vehicle, the limitation of terrains and obstacles on transportation path selection in a complex environment is effectively solved, and the transportation efficiency and the traffic safety are improved.

Description

Technical Field

[0001] The present invention relates to the fields of transportation and intelligent transportation, and particularly to a method for collaborative delivery of unmanned vehicles and unmanned aerial vehicles in complex wild environments. Background Art

[0002] Vehicle path planning in complex environments has always been an important research topic in the fields of transportation and intelligent transportation. During the driving process, vehicles often encounter various complex terrains, such as mountains, wetlands, deserts, etc. These complex terrains pose high requirements for path selection because improper path selection may cause the vehicle to be unable to pass or even get into danger, increasing transportation costs and time. In addition, traditional path planning methods often have many deficiencies when dealing with these complex terrains, unable to comprehensively consider the influence of various environmental factors, resulting in problems such as low traffic efficiency and insufficient safety.

[0003] Traditional path planning methods mainly rely on pre-set map data and manual experience, and it is difficult to respond to environmental changes in real time. For example, the steep slopes and unstable ground in mountainous terrains, the mud and wading obstacles in wetlands, and the soft sand and sandstorm interference in deserts will all have a huge impact on the vehicle's passing ability. In these cases, relying solely on pre-set map data cannot accurately reflect the real-time terrain conditions, easily leading to improper path selection and thus affecting the smooth completion of transportation tasks.

[0004] With the rapid development of unmanned aerial vehicle technology, its application in complex environments provides a new technical means for delivery planning. Unmanned aerial vehicles can not only quickly and accurately perform environmental perception, collect complex terrain data, including information such as elevation, slope, obstacle position and density, providing reliable basic support for path planning, but also cover a wide area, especially in scenarios where it is difficult for humans to reach, obtaining detailed environmental information, bringing greater flexibility and accuracy to path planning in complex environments. In addition, unmanned aerial vehicles can directly undertake the task of cargo delivery. Especially in situations where unmanned vehicles cannot pass, relying on their flexible maneuverability and optimal path planning to achieve fast and efficient transportation. By integrating environmental perception and delivery functions, unmanned aerial vehicles not only significantly improve the delivery efficiency, but also expand the applicable range of delivery scenarios, providing strong support for the development of intelligent delivery technology.

[0005] In summary, traditional path planning methods have many deficiencies in complex environments. By using unmanned aerial vehicle (UAV) technology for terrain data collection and analysis, combined with advanced data processing techniques and a two-layer planning model, these problems can be effectively solved. By obtaining real-time and accurate terrain data through UAVs, conducting comprehensive constraint calculations and path optimization, the traffic efficiency and safety of vehicles in complex environments can be significantly improved. This method has broad application prospects and is applicable to various scenarios that require efficient path planning. Through the deep integration of intelligent transportation technology and unmanned distribution means, it provides an important theoretical basis and practical reference for the development of future intelligent transportation systems. Summary of the Invention

[0006] Object of the Invention: The present invention provides a method for collaborative distribution of unmanned vehicles and UAVs in complex wild environments. By preferentially optimizing the path of the unmanned vehicle and combining the rapid response ability of the UAV, efficient and safe collaborative distribution in complex environments is achieved.

[0007] Technical Solution: The present invention proposes a method for collaborative distribution of unmanned vehicles and UAVs in complex wild environments, including the following steps:

[0008] (1) Within the set range of a complex wild environment area, based on the data of the ground terrain parameters scanned and obtained by the UAV, obtain the positions of obstacles;

[0009] (2) Based on the positions of the obstacles, conduct an initial assessment of the obstacle density, and exclude high-density obstacle areas where the vehicle cannot pass;

[0010] (3) Use the Transformer model for obstacle recognition constraints to confirm the obstacle characteristics, and then conduct constraint calculations from five aspects: terrain width, climbing ability, wading depth, load-bearing weight, and traction force according to the physical parameters of the vehicle, ground terrain parameters, and obstacle characteristics;

[0011] (4) Randomly generate a starting point and an ending point within the area, and calculate the passability score of the path connecting the starting and ending points in each area according to the constraint condition scoring criteria of different terrains. Each score is weighted and summed to obtain the overall passability score;

[0012] (5) Construct a two-layer planning model; the lower layer optimizes the path with the highest passability, considering the constraint condition scoring criteria of different terrains, and calculates the passability score of the path; the upper layer optimizes the path with the shortest time, calculates the passing time of the path according to the distance and the set speed, and finds the optimal path generated by the starting and ending points in each group of areas connecting the departure place and the destination;

[0013] (6) Generate the final 3D and 2D terrain maps, which display the starting point, destination, optimal path, as well as the elevation of the mountainous areas and the water depth of the wetland areas passed by the optimal path, to support the delivery by the unmanned vehicle;

[0014] (7) When the unmanned vehicle cannot pass, the system will re-plan the path of the unmanned vehicle according to the current environmental conditions to ensure that it can select a feasible path closest to the destination. After the unmanned vehicle arrives at the designated location, the remaining delivery tasks will be taken over and completed by the drone;

[0015] (8) The drone takes over the task, re-plans the path according to the environment, generates the optimal path for drone delivery and displays it on the map;

[0016] (9) The drone picks up the goods on the unmanned vehicle by itself and completes the delivery task according to the path.

[0017] Further, the implementation process of the step (1) is as follows:

[0018] Use the drone to perform lidar scanning on the ground, obtain the Digital Surface Model (DSM) and Digital Terrain Model (DTM), identify the terrain and divide the terrain areas, including plains, mountains, wetlands, and deserts;

[0019] By subtracting the DTM from the DSM, determine the positions of the obstacles; process the above data to obtain the width between adjacent obstacles, the elevation difference of the terrain, the horizontal distance, and the water depth parameters.

[0020] Further, the implementation process of the step (2) is as follows:

[0021] Detect the obstacle density in each area:

[0022]

[0023] Among them, is the area of the obstacles in area u i , is the area of area u i . Exclude the areas with a large obstacle density to ensure the passability and safety of path planning.

[0024] Further, the implementation process of the step (3) is as follows:

[0025] Determine the vehicle performance parameters, including the minimum turning radius, rated load, empty load, length, width, height, maximum climbing gradient, minimum traction force, maximum wading depth, tire height, tire width, and rim diameter;

[0026] Using the Transformer model for obstacle recognition constraints; extracting video frames from high-altitude shooting videos of different environments collected, then using Labelimg to label the obstacles in the extracted pictures, and finally training the Transformer model using the labeled dataset in xml format; classifying obstacles into moving obstacles and fixed obstacles;

[0027] The terrain width constraint reflects the minimum passing width of the vehicle on different terrain types, ensuring that the vehicle can pass safely without being restricted. The terrain width angle is used for constraint calculation, and the formula is as follows:

[0028]

[0029] Where, is the minimum terrain width between nodes x and y in region u i , r min is the minimum turning radius of the vehicle, and b vehicle is the width of the vehicle;

[0030] The climbing ability constraint evaluates the performance of the vehicle on different slope terrains, ensuring that the vehicle can pass safely through the slope without landsliding or losing control. The climbing ability angle is used for constraint calculation, and the formula is as follows:

[0031]

[0032]

[0033] Where, is the slope between nodes x and y in region u i , θ max is the maximum climbing slope of the vehicle; Δh xy is the elevation difference between nodes x and y, and d xy is the horizontal distance between nodes x and y;

[0034] The wading depth constraint evaluates the maximum water depth that the vehicle can pass safely through, ensuring that the vehicle does not enter waters that exceed its wading depth capacity. The wading depth angle is used for constraint calculation, and the formula is as follows:

[0035]

[0036] Where, is the water depth between nodes x and y in region u i , f max is the maximum wading depth of the vehicle;

[0037] The load-bearing weight constraint reflects the maximum load capacity of the vehicle on different terrains, ensuring that the vehicle does not overload and cause damage or inability to move. The load-bearing weight angle is used for constraint calculation, and the formula is as follows:

[0038]

[0039] Among them, is the load-bearing weight between nodes x and y in area u i w, and w max is the maximum loading capacity of the vehicle vehicle is the unladen weight of the vehicle; is the ground bearing capacity between nodes x and y in area u i S, and S bear is the load-bearing area of the vehicle; is the compressive strength of area u i SCF is the slope correction factor, and WTCF is the water level correction factor is the maximum water depth of area u i MCF is the humidity correction factor, and WCF is the wind force correction factor; is the deformation coefficient of the tire under load, b tyre is the tire width, h tyre is the tire height, D rim is the rim diameter;

[0040] The traction force constraint evaluates the traction ability of the vehicle on different terrains, ensuring that the vehicle can overcome the ground friction without skidding or stalling. The traction force angle is used for constraint calculation, and the formula is as follows:

[0041]

[0042] Among them, is the traction force between nodes x and y in area u i e, and e min is the minimum traction force requirement of the vehicle; is the ground friction coefficient between nodes x and y in area u i .

[0043] Furthermore, the implementation process of step (4) is as follows:

[0044] The scoring criteria for the constraint conditions of different terrains are as follows: The plain area is scored according to the constraint conditions of width, load-bearing weight, and traction force; the mountainous terrain is scored according to the constraint conditions of width, climbing ability, load-bearing weight, and traction force; the wetland terrain is scored according to the constraint conditions of width, wading depth, load-bearing weight, and traction force; the desert terrain is scored according to the constraint conditions of width, load-bearing weight, and traction force;

[0045] According to the obtained terrain data and the scoring criteria for the constraint conditions of different terrains, calculate the passability score of the road section p connecting the starting point and the end point in area u i , and the formula is as follows:

[0046]

[0047] Among them, is the passability score of road segment p in area u i and w j is the weight of constraint condition c j ; is the value of area u i under constraint condition c j ; V j is the performance parameter of the vehicle under constraint condition c j .

[0048] The value of area u i under constraint condition c j is:

[0049]

[0050] Furthermore, the implementation process of step (5) is as follows:

[0051] Input the departure place and the destination, and set the starting point of the next area as the ending point of the previous area; the lower layer calculates the path with the highest passability connecting the departure place and the destination under different combinations of starting and ending points, and the formula is as follows:

[0052]

[0053] The upper layer calculates the path with the shortest time in the path, and the formula is as follows:

[0054]

[0055] Among them, is the passing time of road segment p in area u i ;

[0056] The passing time of road segment p in area u i is:

[0057]

[0058] Among them, is the length of road segment p in area u i ; is the recommended speed of road segment p in area u i .

[0059] Furthermore, the implementation process of step (7) is as follows:

[0060] Input the current position and find the passable point closest to the destination, and the formula is as follows:

[0061] minL(O feasible ,O target )

[0062] Among them, O feasible is the passable point coordinate, and O target is the destination coordinate. L(O feasible ,O target ) is the distance from the passable point to the destination;

[0063] Constraint calculations are carried out from two perspectives: passability and obstacle area. The passability constraint reflects the high passability of the path, ensuring that the unmanned vehicle can pass on the path. The constraint calculation from the passability perspective is as follows:

[0064]

[0065] Among them, is the passability score of the path connecting the current position and the nearest passable point, and A threshold is the threshold at which the vehicle can pass safely;

[0066] The obstacle area constraint reflects that the nearest passable point must not be located in a high-density obstacle area, ensuring that the unmanned vehicle can reach the designated position smoothly. The constraint calculation from the obstacle area perspective is as follows:

[0067]

[0068] Among them, U hd is the high-density obstacle area.

[0069] Furthermore, the implementation process of step (8) is as follows:

[0070] Input the current position of the UAV and the destination position; constraint calculations are carried out from two perspectives: distance and flight altitude. The distance constraint reflects the distance that the UAV needs to deliver, ensuring that the UAV can return to the unmanned vehicle within its endurance range. The constraint calculation from the distance perspective is as follows:

[0071]

[0072] Among them, v drone is the recommended speed of UAV delivery, and D endurancd is the endurance of the UAV. If the distance constraint is satisfied, the UAV can deliver; if not, the UAV is charged;

[0073] The flight altitude constraint reflects the flight altitude of the UAV, ensuring that the UAV can fly safely in the air. The constraint calculation from the flight altitude perspective is as follows:

[0074] F flight ≤Fthreshold

[0075] Among them, F flight is the actual flight altitude of the UAV, and F threshold is the flight altitude threshold of the UAV;

[0076] With the goal of minimizing the delivery time, the optimal path is selected by combining the constraint conditions. The formula is as follows:

[0077]

[0078] Among them, is the UAV delivery time between nodes x and y within area u i ; the UAV delivery time between nodes x and y within area u i is:

[0079]

[0080] Beneficial effects: Compared with the prior art, the present invention has the following advantages: The present invention can comprehensively consider uncertain factors in a complex environment (such as changes in obstacles, terrain changes, etc.), effectively cope with high-density obstacle areas, avoid path planning failures, improve adaptability and robustness, and thus increase the success rate of path planning; The present invention constructs a two-layer planning model that focuses on optimizing the path selection of the unmanned vehicle in a complex environment. The lower-layer model evaluates the feasibility of the path to ensure its passability; the upper-layer model aims at the shortest passing time and optimizes the driving path of the unmanned vehicle. When the unmanned vehicle encounters an impassable area, the UAV will quickly take over the delivery task and generate the optimal path based on the principle of the shortest time; by preferentially optimizing the path of the unmanned vehicle and combining the flexibility of the UAV, it is possible to achieve efficient and safe delivery in complex terrains, maximizing the adaptability and robustness of path planning; Compared with traditional methods, the present invention comprehensively considers multi-dimensional constraint conditions, significantly optimizes the real-time performance and reliability of path planning, and is applicable to various scenarios that require efficient path optimization, such as field rescue, logistics distribution, and emergency medical transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 is the flow chart of the present invention;

[0082] Figure 2 is the passability score flow chart of the present invention;

[0083] Figure 3 is the 3D topographic map of the embodiment of the present invention;

[0084] Figure 4 is the 2D topographic map of the embodiment of the present invention;

[0085] Figure 5This is the replanned 2D combined path diagram of the embodiment of the present invention. Detailed implementation manners

[0086] The present invention will be further described in detail below with reference to the accompanying drawings.

[0087] As Figure 1 shown, a collaborative delivery method for unmanned vehicles and unmanned aerial vehicles in a complex wild environment according to the present invention includes: First, using unmanned aerial vehicle technology and data processing technology to obtain data on ground terrain parameters; Second, based on the initial obstacle density assessment, excluding high-density obstacle areas that are not suitable for vehicle passage; Third, according to the physical parameters of the vehicle and the ground terrain parameters, performing constraint calculations from five aspects: terrain width, climbing ability, wading depth, load-bearing weight, and traction force; Fourth, calculating the passability score of each area according to the constraint condition scoring criteria for different terrains, and summing up the scores weighted by weights to obtain the overall passability score; Using a bilevel programming model, the lower layer optimizes the path with the highest passability, considering the constraint condition scoring criteria for different terrains, calculating the passability score of the path, and the upper layer optimizes the path with the shortest time, calculating the passing time of the path according to the distance and the set speed, and finding the optimal path; Fifth, when the path of the unmanned vehicle is infeasible, the unmanned aerial vehicle takes over the task, generates the optimal path according to the surrounding environment and the delivery distance, and displays it on the map; Finally, the unmanned aerial vehicle picks up the goods by itself and completes the delivery task according to the map. The specific implementation process is as follows:

[0088] Step 1: Scan and obtain data on ground terrain parameters within the set area range.

[0089] Using an unmanned aerial vehicle to perform lidar scanning on the ground within the set area range to obtain a digital surface model (DSM) and a digital terrain model (DTM), processing the above data, identifying the terrain and classifying the terrain area into one of plain, mountain, wetland, and desert, subtracting the DTM from the DSM to determine the position and size of the obstacles, and further obtaining the width between adjacent obstacles, the elevation difference of the terrain, the horizontal distance, and the water depth parameters.

[0090] In the data acquisition part of the present invention, the drone conducts terrain scanning by carrying a Light Detection and Ranging (LiDAR) sensor to obtain surface data. The LiDAR measures the distance from the surface to the sensor by emitting pulsed lasers and recording their reflection times, thereby obtaining terrain height information. The raw data obtained by the LiDAR is registered and corrected to correct the deviations and errors during the scanning process. The Digital Surface Model (DSM) and Digital Terrain Model (DTM) are extracted from the raw data. Generally, the DSM represents all objects above the surface (such as buildings, trees, etc.), while the DTM is the true height model of the surface. By subtracting the DTM from the raw height data, the DSM is obtained, which reflects the height of the objects above the surface. Using the DSM and DTM data, obstacle detection and extraction can be carried out. Obstacles are generally defined as areas or objects with a height significantly greater than the surrounding terrain. According to the height information and the surrounding terrain features, the position, height, and size features of the obstacles are extracted.

[0091] Step 2: Based on the obstacle positions, conduct an initial assessment of the obstacle density and exclude high-density obstacle areas that are not suitable for vehicle passage.

[0092] Detect the obstacle density within each area, and the formula is as follows:

[0093]

[0094] where, is the area of the obstacles within region u i and is the area of region u i . Exclude the areas where the obstacle density is greater than 70% to ensure the passability and safety of path planning.

[0095] Four different environments were set up to conduct verification experiments of the algorithm in different environments, obtaining the success rate of path planning and the number of target points reached before collision or exhaustion of the time steps in each round. No obstacles were set in Environment 1, static obstacles with different shapes were placed in Environment 2, two different types of dynamic obstacles, one that can move along the x-axis and y-axis and one that can rotate along the z-axis, were set in Environment 3, and the most complex scenario was set in Environment 4, including narrow channels, static obstacles with different shapes, and dynamic obstacles with multiple motion modes. The research shows that as the obstacle density in the environment increases, the difficulty of path planning increases and the success rate of path planning decreases significantly. Usually, the obstacle density in these areas will exceed 50%. Therefore, in practical path planning applications, setting a reasonable obstacle density is crucial. To ensure the smooth progress of path planning and improve the success rate, the obstacle density that meets the passage conditions is set to be less than 70% to avoid the failure of path planning caused by overly dense obstacles.

[0096] Step 3: Use the Transformer model to identify obstacles and identify their characteristics. Then, calculate the constraints from different angles based on the vehicle’s physical parameters, ground terrain parameters, and obstacle characteristics.

[0097] Determine vehicle performance parameters; including minimum turning radius, rated load, empty load, dimensions length, width, height, maximum gradeability, minimum traction, maximum wading depth, tire height, tire width and rim diameter.

[0098] The Transformer model is used to perform obstacle recognition constraints, including confirming and supplementing the obstacles detected in step (1), and identifying ground obstacles not covered by the drone scan to provide more detailed obstacle features.

[0099] Fixed obstacles (such as rocks, trees, and buildings) are usually fixed in the environment, while mobile obstacles (such as pedestrians, animals, and vehicles) may change over time or position. In complex outdoor environments, being able to effectively distinguish between these two types of obstacles is crucial for path planning and safety. Video frames are extracted based on the collected high-altitude videos of different environments, and obstacles are annotated on the extracted images using Labelimg. Finally, the Transformer model is trained using the annotated dataset in XML format to make the model more accurate; more detailed obstacle features include distinguishing between mobile obstacles and fixed obstacles.

[0100] According to vehicle performance parameters, ground terrain parameters and obstacle characteristics, constraint calculations are performed from five perspectives: terrain width, climbing ability, wading depth, load-bearing weight and traction.

[0101] 1) The terrain width constraint reflects the minimum passing width of the vehicle on different terrain types, ensuring that the vehicle can pass safely without restriction. The terrain width angle is used for constraint calculation, and the formula is as follows:

[0102]

[0103] in, is the region u i Minimum terrain width between inner nodes x and y, r min is the minimum turning radius of the vehicle, b vehicle is the width of the vehicle.

[0104] 2) The gradeability constraint evaluates the vehicle's performance on terrains with different slopes to ensure that the vehicle can safely pass the slope without sliding or losing control. The gradeability angle is used for constraint calculation, and the formula is as follows:

[0105]

[0106] Among them, is the slope between nodes x and y within region u, θ i where θ is the maximum climbing slope of the vehicle. max The slope between nodes x and y within region u is given by the following formula:

[0107] Region u i The slope between nodes x and y within it is as follows:

[0108]

[0109] where Δh xy is the elevation difference between nodes x and y, and d xy is the horizontal distance between nodes x and y.

[0110] 3) The wading depth constraint evaluates the maximum water depth that the vehicle can safely pass through, ensuring that the vehicle does not enter waters exceeding its wading depth capacity. The wading depth angle is calculated with the following constraint formula:

[0111]

[0112] where is the water depth between nodes x and y within region u, f i and f is the maximum wading depth of the vehicle. max The maximum wading depth of the vehicle is given by the following formula:

[0113] 4) The load-bearing weight constraint reflects the maximum load capacity of the vehicle on different terrains, ensuring that the vehicle does not overload and cause damage or immobility. The load-bearing weight angle is calculated with the following constraint formula:

[0114]

[0115] where is the load-bearing weight between nodes x and y within region u, w i and w is the maximum loading capacity of the vehicle, w max is the unloaded weight of the vehicle. vehicle The load-bearing weight between nodes x and y within region u is given by the following formula:

[0116] Region u i The load-bearing weight between nodes x and y within it is as follows:

[0117]

[0118] where is the ground bearing capacity between nodes x and y within region u, S i and S is the load-bearing area of the vehicle. bear The ground bearing capacity between nodes x and y within region u is given by the following formula:

[0119] Region u i The ground bearing capacity between nodes x and y within it is as follows:

[0120]

[0121] Among them, is the compressive strength of area u i For plain terrain, the compressive strength is taken as 300 kPa, for mountainous terrain it is 200 kPa, for wetland terrain it is 100 kPa, and for desert terrain it is 50 kPa. SCF is the slope correction factor, and WTCF is the water level correction factor. is the maximum water depth of area u i MCF is the humidity correction factor, which is taken as 0.2 for wetland terrain and 0.05 for desert terrain. WCF is the wind force correction factor, which is taken as 1.2 for desert terrain.

[0122] The slope correction factor, the formula is as follows:

[0123]

[0124] The water level correction factor, the formula is as follows:

[0125]

[0126] The load-bearing area of the vehicle, the formula is as follows:

[0127]

[0128] Among them, is the deformation coefficient of the tire under load, taken as 0.9, b tyre is the tire width, h tyre is the tire height, D rim is the rim diameter.

[0129] 5) The traction force constraint evaluates the traction ability of the vehicle on different terrains to ensure that the vehicle can overcome the ground friction without skidding or stalling. The traction force angle is calculated by constraint, and the formula is as follows:

[0130]

[0131] Among them, is the traction force between nodes x and y in area u i , and e min is the minimum traction force requirement of the vehicle.

[0132] The traction force between nodes x and y in area u i , the formula is as follows:

[0133]

[0134] Among them, is the area ui The ground friction coefficient between internal nodes x and y. For plain terrain, the friction coefficient is taken as 0.4; for mountainous terrain, it is taken as 0.6; for wetland terrain, it is taken as 0.5; and for desert terrain, it is taken as 0.3.

[0135] Step 4: As Figure 2 shown, randomly generate a starting point and an ending point within the area, and calculate the passability score of the path connecting the starting and ending points in each area according to the scoring criteria for the constraint conditions of different terrains. Each score is weighted and summed according to the weights to obtain the overall passability score.

[0136] Randomly generate a starting point and an ending point within the area; set the scoring criteria for the constraint conditions of different terrains. The scoring criteria for the constraint conditions of different terrains are as follows: Plain areas are usually flat, but there may be local narrow passages or require a high load-bearing capacity, so the score is calculated according to the constraint conditions of width, load-bearing weight, and traction force; Mountainous terrain usually has steep slopes and complex terrain features, and has high requirements for vehicle performance, so the score is calculated according to the constraint conditions of width, climbing ability, load-bearing weight, and traction force; Wetland terrain usually has deep water areas and slippery terrain, and requires the vehicle to have good wading ability and traction ability, so the score is calculated according to the constraint conditions of width, wading depth, load-bearing weight, and traction force; Desert terrain usually has mobile sand dunes and low traction ability, which poses a certain challenge to the adaptability of the vehicle, so the score is calculated according to the constraint conditions of width, load-bearing weight, and traction force.

[0137] According to the obtained terrain data and the scoring criteria for the constraint conditions of different terrains, calculate the passability score of the section p connecting the starting and ending points in area u i The passability score of the section p connecting the starting and ending points in area u is calculated as follows:

[0138]

[0139] Where is the passability score of the section p in area u i , w j is the weight of the constraint condition c j . The terrain width is taken as 0.35, the climbing ability is taken as 0.25, the wading depth is taken as 0.15, the load-bearing weight is taken as 0.2, and the traction force is taken as 0.05. is the value of area u i under the constraint condition c j , V j is the performance parameter of the vehicle under the constraint condition c j . The value of area u i under the constraint condition c j is calculated as follows:

[0140]

[0141] Step 5: Using a bilevel programming model, the lower layer optimizes the path with the highest passability, considering the constraint condition scoring criteria for different terrains, calculates the passability score of the path, and the upper layer optimizes the path with the shortest time. According to the distance and the set speed, calculates the passing time of the path, and finds the optimal path generated by the starting and ending points within each group of regions connecting the starting point and the destination.

[0142] Input the starting point and the destination; set the starting point of the next region as the ending point of the previous region; the lower layer calculates the path with the highest passability connecting the starting point and the destination under different combinations of starting and ending points, and the formula is as follows:

[0143]

[0144] The upper layer calculates the path with the shortest time among the paths obtained in (5.3), and the formula is as follows:

[0145]

[0146] Among them, is the passing time of section p in region u i ; the passing time of section p in region u i , and the formula is as follows:

[0147]

[0148] Among them, is the length of section p in region u i , the recommended speed of section p in region u i . Due to the complexity of the terrain, the recommended speed is taken as 30 km / h.

[0149] Step 6: According to the processed geographical data, use Python for model operation to generate a 3D topographic map and the corresponding 2D topographic map, as shown in Figure 3 and Figure 4 . In the 3D topographic map, the terrain of each region, the positions of obstacles, the starting point and the ending point are clearly marked, as shown in Figure 3 ; in the 2D topographic map, it shows the terrain of each region, the generated optimized path, and the elevation of the mountainous area and the water depth of the wet area passed by the optimized path, as shown in Figure 4 , thus providing intuitive decision-making support for the unmanned vehicle to help it select the best driving route in a complex environment.

[0150] Step 7: When the unmanned vehicle cannot pass, the system will re-plan the path of the unmanned vehicle according to the current environmental conditions to ensure that it can select a feasible path closest to the destination. When the unmanned vehicle reaches the designated location, the remaining delivery tasks will be taken over and completed by the unmanned aerial vehicle.

[0151] There are two situations where the unmanned vehicle cannot pass. One is that the unmanned vehicle has no way to go initially, and the initial path planning cannot find a suitable path (for example, all paths are blocked by obstacles or ineligible areas), and the system will automatically determine that the unmanned vehicle cannot continue to execute the task; the other is that the unmanned vehicle cannot continue to drive on the way. During the execution of the task, the unmanned vehicle encounters unexpected situations (such as unforeseen obstacles or environmental changes, etc.), resulting in the inability to move forward, and the system will immediately detect a path blockage or a problem.

[0152] Input the current location and find the passable point closest to the destination. The formula is as follows:

[0153] minL(O feasible ,O target )

[0154] Where O feasible is the coordinate of the passable point, and O target is the coordinate of the destination. L(O feasible ,O target ) is the distance from the passable point to the destination.

[0155] Conduct constraint calculations from two perspectives: passability and obstacle area:

[0156] 1) The passability constraint reflects the high passability of the path, ensuring that the unmanned vehicle can pass on the path. Conduct constraint calculations from the perspective of passability. The formula is as follows:

[0157]

[0158] Where is the passability score of the path connecting the current location and the closest passable point, and A threshold is the threshold for the vehicle to pass safely, which is calculated according to the vehicle performance parameters.

[0159] 2) The obstacle area constraint reflects that the closest passable point should not be located in a high-density obstacle area, ensuring that the unmanned vehicle can reach the designated location smoothly. Conduct constraint calculations from the perspective of the obstacle area. The formula is as follows:

[0160]

[0161] Where U hd is the high-density obstacle area.

[0162] Step 8: The drone takes over the task, re-plans the path according to the environment, generates the optimal path for drone delivery and displays it on the map. The process is as follows:

[0163] Input the current location and the destination location; conduct constraint calculations from two perspectives: distance and flight altitude:

[0164] 1) The distance constraint reflects the distance that the drone needs to deliver, ensuring that the drone can return to the unmanned vehicle within its endurance range. The distance and angle are used for constraint calculation, and the formula is as follows:

[0165]

[0166] Among them, v drone is the recommended speed of the drone delivery. Due to the relatively complex terrain and the existence of obstacles affecting flight, the recommended speed is taken as 40 km / h, and D endurancd is the endurance of the drone. If the distance constraint is satisfied, the drone can deliver; if not, the drone is charged.

[0167] 2) The flight altitude constraint reflects the flight altitude of the drone, ensuring that the drone can fly safely in the air. The flight altitude and angle are used for constraint calculation, and the formula is as follows:

[0168] F flight ≤F threshold

[0169] Among them, F flight is the actual flight altitude of the drone, and F threshold is the flight altitude threshold of the drone. According to the Interim Regulations on the Administration of Unmanned Aerial Vehicle Flight in China, the flight altitude threshold of the drone is set at 120 meters.

[0170] With the goal of minimizing the delivery time, the optimal path is selected by combining the constraint conditions, and the formula is as follows:

[0171]

[0172] Among them, is the drone delivery time between nodes x and y in area u i ; the drone delivery time between nodes x and y in area u i , and the formula is as follows:

[0173]

[0174] The optimal path of the drone is shown on the map, as Figure 5 shown, and the drone delivers goods according to the path.

[0175] Step 9: The drone picks up the goods on the unmanned vehicle by itself and completes the delivery task according to the path.

[0176] After the driverless vehicle reaches the nearest passable point, the hatch on the top opens; the drone takes off from the charging platform and reaches directly above the cargo box 1; the drone locks with the cargo box 1 to complete the delivery task. The claw hook under the drone enters the hole in the cargo box. The ball on the claw hook collides with the ball in the hole, and the ball on the claw hook moves upward, driving the rack upward, driving the four claws of the claw hook to open, and getting stuck with the card slot in the hole, thus achieving the locked state. The cargo box 2 reaches the position of the cargo box 1. When the cargo box 1 is taken away, the pressure sensor at the bottom of the driverless vehicle detects that the pressure decreases to the weight of the cargo box 1, and the lifting platform at the bottom of the driverless vehicle raises the cargo box 2 located directly below the cargo box 1 to the original position of the cargo box 1; the drone locks with the cargo box 2 to complete the delivery task; the cargo box 3 reaches the position of the cargo box 1. When the cargo box 2 is taken away, the pressure sensor at the bottom of the drone detects that the pressure decreases to the weight of the cargo box 2, and the moving platform on the right side of the driverless vehicle pushes the cargo box 3 located on the right side of the cargo box 2 to the position of the cargo box 2, and then the lifting platform at the bottom raises the cargo box 3 to the original position of the cargo box 1; the drone locks with the cargo box 3, and so on, to complete the delivery task.

[0177] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements and retouches can still be made, and these improvements and retouches should also be regarded as the protection scope of the present invention.

Claims

1. A method for collaborative delivery between unmanned vehicles and unmanned aerial vehicles in a complex field environment, characterized in that: The following steps are involved: (1) Obtain the location of obstacles within the set complex outdoor environment area by scanning and acquiring data on ground terrain parameters using drones; (2) Based on the obstacle locations, an initial obstacle density assessment is performed to eliminate high-density obstacle areas where vehicles cannot pass; (3) Use the Transformer model to identify obstacles and identify their characteristics. Then, based on the vehicle's physical parameters, ground terrain parameters, and obstacle characteristics, calculate the constraints from five perspectives: terrain width, gradeability, wading depth, load-bearing weight, and traction. (4) Randomly generate the starting point and the end point in the area, and calculate the passability score of the path connecting the starting point and the end point in each area according to the scoring criteria of the constraints of different terrains. The scores of each item are weighted and summed to obtain the overall passability score; (5) Construct a two-level planning model; the lower level optimizes the path with the highest passability, considers the scoring criteria for the constraints of different terrains, and calculates the passability score of the path; the upper level optimizes the path with the shortest time, calculates the passing time of the path based on the distance and the set speed, and finds the optimal path generated by the starting and ending points in each group of areas connecting the departure and destination; (6) Generate the final 3D and 2D terrain maps, which show the departure point, destination, optimal route, and the elevation of the mountainous area and the water depth of the wetland area through which the optimal route passes, to support unmanned vehicle delivery; (7) When the unmanned vehicle is unable to pass, the system will re-plan the route of the unmanned vehicle based on the current environmental conditions to ensure that it can choose a feasible route closest to the destination. When the unmanned vehicle arrives at the designated location, the remaining delivery task will be taken over and completed by the drone; (8) The drone takes over the task, replans the route according to the environment, generates the optimal route for drone delivery and displays it on the map; (9) The drone picks up the goods from the unmanned vehicle and completes the delivery task according to the route.

2. According to claim 1, a method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment is characterized in that: The implementation process of step (1) is as follows: Use drones to perform lidar scanning on the ground to obtain digital surface models (DSM) and digital terrain models (DTM), identify terrain and divide terrain areas, including plains, mountains, wetlands, and deserts; The location of obstacles is determined by subtracting DTM from DSM; the above data are processed to obtain the width between adjacent obstacles, the elevation difference of the terrain, the horizontal distance and the water depth parameters.

3. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment according to claim 1 is characterized in that: The implementation process of step (2) is as follows: Detect obstacle density in each area: in, is the region u i The area of ​​the internal obstacle, is the region u i area; exclude areas with high obstacle density to ensure the passability and safety of path planning.

4. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment according to claim 1 is characterized in that: The implementation process of step (3) is as follows: Determine vehicle performance parameters, including minimum turning radius, rated load, empty load, dimensions length, width, height, maximum gradeability, minimum traction, maximum wading depth, tire height, tire width and rim diameter; Use the Transformer model to identify obstacles; extract video frames based on the collected high-altitude videos of different environments, then use Labelimg to annotate the extracted images with obstacles, and finally use the annotated data set in xml format to train the Transformer model; distinguish obstacles into mobile obstacles and fixed obstacles; The terrain width constraint reflects the minimum passing width of the vehicle on different terrain types, ensuring that the vehicle can pass safely without restrictions. The terrain width angle is used for constraint calculation, and the formula is as follows: in, is the region u i Minimum terrain width between inner nodes x and y, r min is the minimum turning radius of the vehicle, b vehicle is the width of the vehicle; The gradeability constraint evaluates the vehicle's performance on terrains of different slopes to ensure that the vehicle can safely pass the slope without sliding or losing control. The gradeability angle is used for constraint calculation, and the formula is as follows: in, is the region u i The slope between the internal nodes x and y, θ max is the maximum climbing grade of the vehicle; Δh xy is the elevation difference between nodes x and y, d xy is the horizontal distance between nodes x and y; The wading depth constraint evaluates the maximum water depth that the vehicle can safely pass through, ensuring that the vehicle does not enter waters that exceed its wading depth capability. The wading depth angle is constrained and calculated using the following formula: in, is the region u i The water depth between the internal nodes x and y, f max is the maximum wading depth of the vehicle; The load-bearing weight constraint reflects the maximum load capacity of the vehicle on different terrains, ensuring that the vehicle will not be overloaded and damaged or unable to move. The load-bearing weight angle is used for constraint calculation, and the formula is as follows: in, is the region u i The load between internal nodes x and y, w max is the maximum load of the vehicle, w vehicle is the empty load of the vehicle; is the region u i The ground bearing capacity between the internal nodes x and y, S bear is the load-bearing area of ​​the vehicle; is the region u i The compressive strength, SCF is the slope correction factor, WTCF is the water level correction factor, is the region u i The maximum water depth, MCF is the moisture correction factor, and WCF is the wind correction factor; is the deformation coefficient of the tire under load, b tyre is the tire width, h tyre is the tire height, D rim is the rim diameter; The traction constraint evaluates the vehicle's traction ability on different terrains, ensuring that the vehicle can overcome ground friction without sliding or stagnating. The traction angle is constrained and calculated using the following formula: in, is the region u i The traction force between internal nodes x and y, e min is the minimum traction requirement of the vehicle; is the region u i The ground friction coefficient between the internal nodes x and y.

5. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment according to claim 1 is characterized in that: The implementation process of step (4) is as follows: The scoring criteria for the constraints of different terrains are as follows: the plain area is scored based on the constraints of width, load-bearing weight and traction; the mountainous terrain is scored based on the constraints of width, climbing ability, load-bearing weight and traction; Wetland terrain is scored based on the constraints of width, wading depth, load weight and traction; desert terrain is scored based on the constraints of width, load weight and traction; According to the terrain data obtained and the scoring criteria of different terrain constraints, the area u is calculated. i The passability score of the road segment p that connects the start and end points is as follows: in, is the region u i The passability score of the inner section p, w j is the constraint c j The weight of is the region u i Under the constraint condition c j The value below, V j is the vehicle under the constraint c j The following performance parameters: Area u i Under the constraint condition c j The following values ​​are:

6. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment according to claim 1 is characterized in that: The implementation process of step (5) is as follows: Enter the departure and destination, and set the starting point of the next area to be the end point of the previous area; the lower layer calculates the highest passable path connecting the departure and destination under different combinations of starting and ending points. The formula is as follows: The upper layer calculates the shortest path in the path, the formula is as follows: in, is the region u i The travel time of the inner section p; Area u i The travel time of inner section p is: in, is the region u i The length of the inner section p, Area u i Recommended speed for inner section p.

7. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex field environment according to claim 1 is characterized in that: The implementation process of step (7) is as follows: Enter the current location and find the nearest accessible point to the destination. The formula is as follows: minL(O feasible ,THE target ) Among them, O feasible is the coordinate of the traversable point, O target is the destination coordinate, L(O feasible ,O target ) is the distance from the traversable point to the destination; Constraint calculation is performed from two perspectives: passability and obstacle area. The passability constraint reflects the high passability of the path, ensuring that the unmanned vehicle can pass on the path. The constraint calculation can be performed from the perspective of passability. The formula is as follows: in, is the passability score of the path connecting the current location and the nearest passable point, A threshold is the threshold at which vehicles can pass safely; The obstacle area constraint reflects that the nearest passable point must not be located in a high-density obstacle area, ensuring that the unmanned vehicle can reach the specified location smoothly. The obstacle area angle is used for constraint calculation, and the formula is as follows: Among them, U hd It is a high-density obstacle area.

8. The method for coordinated delivery of unmanned vehicles and unmanned aerial vehicles in a complex outdoor environment according to claim 1 is characterized in that: The implementation process of step (8) is as follows: Input the current location of the drone and the destination location; perform constraint calculations from the perspectives of distance and flight altitude; the distance constraint reflects the distance that the drone needs to deliver, ensuring that the drone returns to the unmanned vehicle within its endurance range. The distance constraint is calculated using the following formula: Among them, v drone is the recommended speed for drone delivery, D endurancd is the endurance of the drone. If the distance constraint is met, the drone can deliver. If not, the drone needs to be recharged. The flight altitude constraint reflects the flight altitude of the drone, ensuring that the drone can fly safely in the air. The flight altitude angle is constrained and calculated using the following formula: F flight ≤F threshold Among them, F flight is the actual flight altitude of the UAV, F threshold is the flight altitude threshold of the drone; With the goal of minimizing delivery time, the optimal path is selected in combination with constraints. The formula is as follows: in, is the region u i Drone delivery time between nodes x and y in area u i The drone delivery time between internal nodes x and y is:

Images