Two-layer path planning algorithm for UAV and unmanned vehicle in uncertain environment

Abstract

The technical field to which the present invention relates is the field of UAV / unmanned vehicle collaborative systems, and in particular, to a dual-layer path planning algorithm for UAVs and unmanned vehicles in an uncertain environment suitable for cooperation between UAVs and unmanned vehicles, including: (1) considering the uncertainty of post-earthquake urban scenes and the compatibility of algorithms, the algorithm extracts necessary elements from high-precision maps based on the opendrive map standard, uses cubic polynomials to process the extracted elements, and establishes a map model; (2) based on the map model, the unmanned vehicle layer algorithm uses an improved A* algorithm to make it better adapt to post-earthquake urban scenes; (3) at the UAV algorithm layer, three segmentation methods are proposed to design heuristic algorithms for quickly solving variables; the two-layer algorithms are integrated according to the established model to obtain a dual-layer path planning algorithm. The present invention can improve the efficiency of task completion and has important reference value for path planning of post-earthquake urban rescue.

Description

Technical field:

[0001] The technical field to which the present invention relates is the field of UAV / unmanned vehicle collaborative systems, and in particular, to a double-layer path planning algorithm for UAVs and unmanned vehicles in an uncertain environment suitable for collaboration between UAVs and unmanned vehicles. Background technology:

[0002] Natural disasters occur frequently in the world today, and their destructiveness threatens human life and seriously affects the development of social economy. In order to minimize casualties and property losses, it is necessary to implement an efficient and orderly disaster relief plan. Since emergency rescue operations are usually time-sensitive and sudden, it is necessary to scientifically plan the driving routes of rescue teams transporting emergency supplies in a short period of time. At the same time, the degree of damage to the road network nodes, the order of repairs and the demand for materials in the disaster-stricken areas must be taken into account. If the emergency supplies are not delivered to the designated location within the demand buffer period, people will suffer from physical discomfort due to lack of supplies, and even threaten their lives. Therefore, in order to effectively reduce the degree of suffering of people during the rescue period and meet the material needs of the disaster-stricken areas as soon as possible, it is necessary to comprehensively consider road maintenance and material transportation routes.

[0003] The core of emergency rescue path selection is the shortest path problem. Due to changes in road conditions after the disaster and the destruction of communication equipment and image acquisition equipment, it is impossible to judge the accessibility of the road based on pre-earthquake equipment and technology. In traditional post-disaster unmanned rescue, unmanned vehicles and unmanned systems are often used for path planning, and the path is replanned after encountering obstacles, but this will reduce rescue efficiency.

[0004] Therefore, in scenarios where the extent of post-disaster road damage is unknown, if drones and unmanned vehicles can be used to work together, they can quickly search for the shortest path to the destination, saving time for disaster relief and material transportation. Summary of the invention:

[0005] In order to reduce the time taken by UAVs and unmanned vehicles to reach their destinations in post-earthquake urban scenarios, the present invention proposes a double-layer path planning algorithm for UAVs and unmanned vehicles in uncertain environments to achieve synchronous collaboration in space and time.

[0006] The present invention can be achieved by the following measures:

[0007] A dual-layer path planning algorithm for unmanned aerial vehicles and unmanned vehicles in an uncertain environment, characterized by comprising the following steps:

[0008] Step 1: Model the undirected graph of the high-precision map, select the OpenDrive format as the map model format, extract the necessary elements representing a road section from the map data: id, starting intersection, end intersection, length, and road shape parameters; extract the necessary elements representing a junction: id, coordinates, and the road sections connected to it. Convert the road direction parameters into the five parameters of a cubic curve: a, b, c, d, e, and form a cubic polynomial to represent the direction of the road section. The specific formula is as follows: ax 3 +bx 2 +cx+d=ey

[0009] The parameter obstacle_pos is introduced and combined with the coordinates of the intersection at the starting point of the road section to express the position of the obstacle in the road section. In the element of the intersection, a set of data link_road_id is introduced to store the road section connected to the intersection;

[0010] Step 2: Based on the requirements of the improved A* algorithm for data structure, re-plan the global path of the unmanned vehicle;

[0011] Step 3: Use three heuristic algorithms to quickly solve the return position of the drone to quickly solve the optimization variables, namely: a heuristic algorithm based on the number of segmentation nodes, a heuristic algorithm based on fixed spacing segmentation, and a heuristic algorithm based on binary segmentation;

[0012] Step 4-1: After the starting point and the end point are selected, the algorithm starts;

[0013] Step 4-2: The unmanned vehicle first calculates its position p V That is, the starting point position uses the A* algorithm for global path planning;

[0014] Step 4-3: Determine the current position of the drone D Whether the end point has been reached, if so, the algorithm ends, otherwise, execute step 4-4;

[0015] Step 4-4: Use the selected heuristic algorithm to find the position B of the unmanned vehicle when returning in advance V 、The position of the drone when returning B D , the positions of the two at the time of confluence M , and the speed at each position;

[0016] Step 4-5: The drone starts inspection and determines whether it is in B D Will there be obstacles before the vehicle? If yes, execute step 4-8; otherwise, execute step 4-6.

[0017] Step 4-6: Update p V 、p D is the final point where the two meet.M ;

[0018] Step 4-7: According to the final p V 、p D Store the path and the speed at each key point and execute step 4-3;

[0019] Step 4-8: Update p D is the position p of the detected obstacle O , and calculate the position p that the unmanned vehicle reaches after using the same time V ;

[0020] Step 4-9: Initialize the main elements of the A* algorithm, and the unmanned vehicle replans the global path based on its own position;

[0021] Step 4-10: Update the re-inspection point p c The uninspected starting position of the new path is determined, and it is determined whether the drone can return to the unmanned vehicle if it inspects again. If it can, execute steps 4-7. Otherwise, the unmanned vehicle waits for the drone to return and executes steps 4-7.

[0022] In step 2 of the present invention, the diagonal distance is used as the cost estimation heuristic function of the A* algorithm, as shown in the following formula: h(n)=D×max(abs(n x -goal x ),abs(n y -goal y )), add the replanning point as a new node, which will be connected to the two existing junctions of the road section. When the UAV encounters an obstacle, the unmanned vehicle must re-plan the global path at the current location, where D is the cost of moving a grid. The specific implementation steps are as follows:

[0023] Step 2-1: Initialize Junction and Road according to the extracted map element file, the movement cost G from the starting point to the specified square on the grid, the estimated movement cost H from the specified square to the end point, and the total movement cost F;

[0024] Step 2-2: Determine whether the point is a junction node. If not, add the point as a new node.

[0025] Step 2-3: The nodes where the two junctions of the road section are located are taken as the nodes connected to it. After calculating G, F, and H, the nodes are added to the open_list table, and their parent nodes are set to the number of intersections.

[0026] Step 2-4: Update the open_list table and the G and F of the current intersection according to the calculated size of G, and find the node with the smallest F value and add it to the close_list table;

[0027] Step 2-5: When the point reaches the end point, output the entire path according to the close_list table.

[0028] In step 3 of the present invention, the heuristic algorithm for determining the number of segmentation nodes is optimized as a heuristic function. After the global path is determined, all nodes of the path are used as segmentation points of the entire path. The entire arc path is discretized by the segmentation points and then the algorithm is used to optimize the return position of the desired drone. In this algorithm, the original section nodes are P 0 To P n The specific implementation steps are as follows:

[0029] Step 3-1-1: From P 0 Start to judge whether the drone can return to the unmanned vehicle until it cannot return.

[0030] Step 3-1-2: From P i-1 To P i The path of the node is evenly divided into k+1 arcs (k is the number of divisions), and the next division node is updated to p 1 , until the judgment reaches p k ,

[0031] Step 3-1-3: Determine whether the original section node and the subsequently added split node can meet the return conditions. If so, patrol to the node; if not, return at the current node.

[0032] The heuristic algorithm based on fixed-spacing segmentation used in step 3 is divided into two segmentation processes. The first segmentation process divides the original arc path into small segments with larger spacing. When it is determined which segment the UAV leaves from, the arc path is segmented again to determine the specific location, d 1 is the spacing of the first segmentation process, d 2 is the spacing of the second segmentation process. The specific implementation steps are as follows:

[0033] Step 3-2-1: From P 0 Start to determine whether the drone can return to the unmanned vehicle until it cannot return;

[0034] Step 3-2-2: In the first segmentation process, the drone starts from P 0 Every d 1 Visit a point, when the visited split node position p next +d 1 Greater than the original node Pi When the position of the split node is reached, the position of the split node will cross P according to the planned path i , so that the point falls on P i With P i+1 between;

[0035] Step 3-2-3: In the second segmentation process, the drone will 2 Visit a point and update the next split node. The size of the distance between the two processes affects the efficiency and accuracy of the algorithm. The distance d of the first split process 1 The spacing d of the second segmentation process is determined according to the map scale and the UAV's range. 2 Determined according to the required accuracy;

[0036] In solving the return position of the drone, we use the binary division method to solve the problem and propose a heuristic algorithm based on the binary division method. After the mission starts, the unmanned vehicle plans a global arc path. Since the unmanned vehicle and the drone are at the same position at this time, the position of the unmanned vehicle is set to a and the end point value is set to b. The specific implementation steps are as follows:

[0037] Step 3-3-1: Determine whether the drone can return after the unmanned vehicle takes the same amount of time to travel from point b to point b. If it can, the drone's return point is point b or it does not need to return. Otherwise, use the binary search method to find the return point.

[0038] Step 3-3-2: Set c = (a + b) / 2, and determine whether the difference between the remaining energy f(c) of the drone at that location and the energy g(c) used to return to the unmanned vehicle at that location satisfies the given accuracy ξ, that is, whether 0≤f(c)-g(c)≤ξ holds.

[0039] Step 3-3-3: If true, the return point is c, otherwise if f(c)-g(c)>ξ, let a=c, if f(c)-g(c)<0, let b=c, and repeat until the return point is found;

[0040] Step 3-3-4: When the drone returns, the drone and the unmanned vehicle are at the same position. The current position is set to a, and the binary search for the return value is performed again until the entire path inspection is completed.

[0041] The present invention is a dual-layer path planning algorithm for a cooperative system of unmanned aerial vehicles and unmanned vehicles, which is suitable for post-earthquake urban disaster relief in an uncertain environment. The algorithm uses the advantages of the unmanned aerial vehicle, such as wide field of view, high mobility, and high inspection efficiency, to provide map information for the actions of the unmanned vehicle, and uses the advantages of the unmanned vehicle, such as high load and long endurance, to transport materials. At the same time, the unmanned vehicle provides endurance for the unmanned aerial vehicle. With the help of the algorithm, the cooperative system of unmanned vehicles and unmanned aerial vehicles can improve the efficiency of disaster relief and ensure that rescue materials and personnel can quickly reach the rescue site. The main contents of the algorithm of the invention include: (1) Considering the uncertainty of the post-earthquake urban scene and the compatibility of the algorithm, the algorithm extracts the necessary elements in the high-precision map based on the opendrive map standard, uses cubic polynomials to process the extracted elements, and establishes a map model. (2) Based on the map model, the unmanned vehicle layer algorithm uses an improved A* algorithm to make it better adapted to the post-earthquake urban scene. (3) At the unmanned aerial vehicle algorithm layer, three segmentation methods are proposed to design a heuristic algorithm for quickly solving variables. The two-layer algorithm is integrated according to the constructed model to obtain a dual-layer path planning algorithm. The present invention can improve the efficiency of task completion and has important reference value for path planning of post-earthquake urban rescue. Description of the drawings:

[0042] Attached Figure 1 It is a principle block diagram of the present invention.

[0043] Attached Figure 2 It is a flow chart of the present invention.

[0044] Attached Figure 3 It is a schematic diagram of the undirected graph element composition of the present invention. Specific implementation method:

[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0046] This example proposes a dual-layer path planning algorithm for UAVs and unmanned vehicles in an uncertain environment, including path planning algorithms for unmanned vehicle layers and UAV layers. In the path planning algorithm for the unmanned vehicle layer, a feature extraction method and expression method based on the OpenDrive format are proposed, a map model is established, and an improved A* algorithm is used to complete the search for the shortest path in the map. Specifically, the diagonal distance is used as the cost estimation heuristic function of the A* algorithm. Combined with the unmanned vehicle and the A* algorithm, the road direction parameters are converted into 5 parameters of a cubic curve to form a cubic polynomial to represent the direction of the road section. Its purpose is to provide navigation function for the UAV in the working environment;

[0047] The drone layer has designed three heuristic algorithms for quickly solving optimization variables. The feature is that the best heuristic algorithm can be selected according to the length of the path distance, earthquake level and the number of obstacles in different scenarios, so as to make the best planning decision, avoid the occurrence of drones unable to return and low efficiency in executing tasks, and realize basic collaboration needs; the specific implementation steps are as follows:

[0048] Step 1: After the starting point and the end point are selected, the algorithm starts;

[0049] Step 2: The unmanned vehicle first calculates its position p V That is, the starting point position uses the improved A* algorithm for global path planning;

[0050] Step 3: Determine the current position of the drone D Whether the end point has been reached, if so, the algorithm ends, otherwise, execute step 4;

[0051] Step 4: Use the selected heuristic algorithm to find the position B of the unmanned vehicle when returning in advance V 、The position of the drone when returning B D , the positions of the two at the time of confluence M , and the speed at each position;

[0052] Step 5: The drone starts inspection and determines whether it is in B D Whether there will be obstacles before the vehicle, if yes, go to step 8, otherwise go to step 6;

[0053] Step 6: Update p V 、p D is the final point where the two meet. M ;

[0054] Step 7: According to the final p V 、p D Store the path and the speed at each key point, and execute step 3;

[0055] Step 8: Update p D is the position p of the detected obstacle O , and calculate the position p that the unmanned vehicle reaches after using the same time V ;

[0056] Step 9: Initialize the main elements of the A* algorithm, and the unmanned vehicle replans the global path based on its own position;

[0057] Step 10: Update the re-inspection point p c The uninspected starting position of the new path is determined, and it is determined whether the UAV can return to the unmanned vehicle if it inspects again. If so, execute step 7. Otherwise, the unmanned vehicle waits for the UAV to return and executes step 7.

[0058] The above two-layer path planning algorithm in this example includes the following parts:

[0059] (1) Undirected graph modeling of high-precision maps;

[0060] (2) Combined with the requirements of the improved A* algorithm for data structure, the global path planning of the unmanned vehicle is re-performed;

[0061] (III) Three heuristic functions are proposed to quickly solve the return position of the UAV to quickly solve the optimization variables, namely: a heuristic algorithm based on the number of segmentation nodes, a heuristic algorithm based on fixed spacing segmentation, and a heuristic algorithm based on binary segmentation;

[0062] (IV) Based on the collaborative model between UAVs and unmanned vehicles, the path planning algorithms of the unmanned vehicle layer and the UAV layer are combined to design a basic double-layer path planning algorithm.

[0063] Specifically, this example uses the OpenDrive format as the map model format, extracts the necessary elements representing a road section from the map data: id, starting intersection, end intersection, length, and road shape parameters; extracts the necessary elements representing a junction: id, coordinates, and the road sections connected to it. The road direction parameters are converted into the five parameters of a cubic curve: a, b, c, d, e, forming a cubic polynomial to represent the direction of the road section. The specific formula is as follows:

[0064] ax 3 +bx 2 +cx+d=ey

[0065] The parameter obstacle_pos is introduced and combined with the coordinates of the intersection at the start of the road segment to express the position of the obstacle in the road segment. In the elements of the intersection, a set of data link_road_id is introduced to store the road segments connected to the intersection. All the elements used are as follows Figure 3 shown.

[0066] The present invention improves the original A* algorithm and adopts the diagonal distance as the cost estimation heuristic function of the A* algorithm, as shown in the following formula: h(n)=D×max(abs(n x -goal x ),abs(n y -goal y ));

[0067] Add the replanning point as a new node, which will be connected to the two existing junctions of the road section. When the UAV encounters an obstacle, the unmanned vehicle must re-plan the global path at the current location, where D is the cost of moving one grid. The specific implementation steps are as follows:

[0068] Step 1: Initialize Junction and Road according to the extracted map element file, the movement cost G from the starting point to the specified square on the grid, the estimated movement cost H from the specified square to the end point, and the total movement cost F;

[0069] Step 2: Determine whether the point is the node where the junction is located. If not, add the point as a new node;

[0070] Step 3: The nodes where the two junctions of the road section are located are taken as the nodes connected to it. After calculating G, F, and H, the nodes are added to the open_list table, and their parent nodes are set to the number of intersections.

[0071] Step 4: Update the open_list table and the G and F of the current intersection according to the calculated size of G, and find the node with the smallest F value and add it to the close_list table;

[0072] Step 5: When the point reaches the end point, output the entire path according to the close_list table.

[0073] This example further optimizes the heuristic algorithm for determining the number of split nodes as a heuristic function. When the global path is determined, all nodes of the path are used as split points of the entire path. The entire arc path is discretized by the split points and then the algorithm is used to optimize the return position of the desired drone. In this algorithm, the original section nodes are P 0 To P n The specific implementation steps are as follows:

[0074] Step 1: From P 0 Start judging whether the drone can return to the unmanned vehicle until it cannot return.

[0075] Step 2: From P i-1 To P i The path of the node is evenly divided into k+1 arcs (k is the number of divisions), and the next division node is updated to p 1 , until the judgment reaches p k .

[0076] Step 3: Determine whether the original section node and the subsequently added split node can meet the return conditions. If so, patrol to the node; if not, return at the current node.

[0077] The heuristic algorithm based on fixed-spacing segmentation adopted in the present invention is mainly divided into two segmentation processes. The first segmentation process divides the original arc path into small segments with larger spacing. When it is determined which segment of the path the drone leaves from, the arc path is further segmented to determine the specific location. 1 is the spacing of the first segmentation process, d 2 is the spacing of the second segmentation process. The specific implementation steps are as follows:

[0078] Step 1: From P 0 Start judging whether the drone can return to the unmanned vehicle until it cannot return.

[0079] Step 2: In the first segmentation process, the drone starts from P 0 Every d 1 Visit a point, when the visited split node position p next +d 1 Greater than the original node P i When the position of the split node is reached, the position of the split node will cross P according to the planned path i , so that the point falls on P i With P i+1 between.

[0080] Step 3: In the second segmentation process, the drone moves every d 2 Visit a point and update the next split node.

[0081] The size of the distance between the two processes affects the efficiency and accuracy of the algorithm. The distance d of the first segmentation process 1 The spacing d of the second segmentation process is determined according to the map scale and the UAV's range. 2 Determined according to the required accuracy.

[0082] In solving the return position of the drone, the present invention uses the binary division method to solve the problem and proposes a heuristic algorithm based on the binary division method. After the mission starts, the unmanned vehicle plans a global arc path. Since the unmanned vehicle and the drone are at the same position at this time, the position of the unmanned vehicle is set to a and the end point value is set to b. The specific implementation steps are as follows:

[0083] Step 1: Determine whether the drone can return after the unmanned vehicle takes the same amount of time to travel from point B to point B. If so, the drone's return point is point B or there is no need to return. Otherwise, use the binary search method to find the return point.

[0084] Step 2: Set c = (a + b) / 2, and determine whether the difference between the remaining energy f(c) of the drone at that location and the energy g(c) used to return to the unmanned vehicle at that location satisfies the given accuracy ξ, that is, whether 0≤f(c)-g(c)≤ξ holds.

[0085] Step 3: If it is true, the return point is c. Otherwise, if f(c)-g(c)>ξ, let a=c. If f(c)-g(c)<0, let b=c. Repeat this process until the return point is found.

[0086] Step 4: When the drone returns, the drone and the unmanned vehicle are at the same position. The current position is set to a, and the binary search for the return value is performed again until the entire path inspection is completed.

[0087] According to the cooperative system model of the UAV and the unmanned vehicle, and the path planning algorithms of the UAV and the unmanned vehicle involved above, the present invention proposes a double-layer path planning algorithm for the cooperative system of the vehicle.

Claims

1. A two-layer path planning algorithm for drones and unmanned vehicles in uncertain environments. It is characterized in that The following steps are involved: Step 1: Model the undirected graph of the high-precision map, select the OpenDrive format as the map model format, extract the necessary elements representing a road section from the map data: id, starting intersection, end intersection, length, and road shape parameters; extract the necessary elements representing a junction: id, coordinates, and the road sections connected to it, and convert the road direction parameters into the five parameters of the cubic curve: a, b, c, d, e, to form a cubic polynomial to represent the road section direction. The specific formula is as follows: ax 3 +bx 2 +cx+d=ey The parameter obstacle_pos is introduced and combined with the coordinates of the intersection at the starting point of the road section to express the position of the obstacle in the road section. In the element of the intersection, a set of data link_road_id is introduced to store the road section connected to the intersection; Step 2: Based on the requirements of the improved A* algorithm for data structure, re-plan the global path of the unmanned vehicle; Step 3: Use three heuristic algorithms to quickly solve the return position of the drone to quickly solve the optimization variables, namely: a heuristic algorithm based on the number of segmentation nodes, a heuristic algorithm based on fixed spacing segmentation, and a heuristic algorithm based on binary segmentation; Step 4-1: After the starting point and the end point are selected, the algorithm starts; Step 4-2: The unmanned vehicle first calculates its position p V That is, the starting point position uses the A* algorithm for global path planning; Step 4-3: Determine the current position of the drone D Whether the end point has been reached, if so, the algorithm ends, otherwise, execute step 4-4; Step 4-4: Use the selected heuristic algorithm to find the position B of the unmanned vehicle when returning in advance V 、The position of the drone when returning B D , the positions of the two at the time of merging M , and the speed at each position; Step 4-5: The drone starts inspection and determines whether it is in B D Will there be obstacles before the vehicle? If yes, execute step 4-8; otherwise, execute step 4-6. Step 4-6: Update p V 、p D is the final point where the two meet. M ; Step 4-7: According to the final p V 、p D Store the path and the speed at each key point and execute step 4-3; Step 4-8: Update p D is the position p of the detected obstacle O , and calculate the position p that the unmanned vehicle reaches after using the same time V ; Step 4-9: Initialize the main elements of the A* algorithm, and the unmanned vehicle replans the global path based on its own position; Step 4-10: Update the re-inspection point p c The uninspected starting position of the new path is determined, and it is determined whether the drone can return to the unmanned vehicle if it inspects again. If it can, execute steps 4-7. Otherwise, the unmanned vehicle waits for the drone to return and executes steps 4-7.

2. According to the double-layer path planning algorithm for UAV and unmanned vehicle in uncertain environment as described in claim 1, It is characterized in that In step 2, the diagonal distance is used as the cost estimation heuristic function of the A* algorithm, as shown in the following formula: h(n)=D×max(abs(n x -goal x ),abs(n y -goal y )), Add the replanning point as a new node, which will be connected to the two existing junctions of the road section. When the UAV encounters an obstacle, the unmanned vehicle must re-plan the global path at the current position, where D is the cost of moving a grid. The specific implementation steps are as follows: Step 2-1: Initialize the Junction and Road according to the extracted map element file, the movement cost G from the starting point to the specified square on the grid, the estimated movement cost H from the specified square to the end point, and the total movement cost F; Step 2-2: Determine whether the point is a junction node. If not, add the point as a new node. Step 2-3: The nodes where the two junctions of the road section are located are taken as the nodes connected to it. After calculating G, F, and H, the nodes are added to the open_list table, and their parent nodes are set to the number of intersections. Step 2-4: Update the open_list table and the G and F of the current intersection according to the calculated size of G, and find the node with the smallest F value and add it to the close_list table; Step 2-5: When the point reaches the end point, output the entire path according to the close_list table.

3. According to the double-layer path planning algorithm for UAV and unmanned vehicle in uncertain environment as described in claim 1, It is characterized in that In step 3, the heuristic algorithm for determining the number of segmentation nodes is optimized as a heuristic function. When the global path is determined, all nodes of the path are used as segmentation points of the entire path. The entire arc path is discretized by the segmentation points and then the algorithm is used to optimize the return position of the desired drone. In this algorithm, the original section nodes are P 0 To P n The specific implementation steps are as follows: Step 3-1-1: From P 0 Start to judge whether the drone can return to the unmanned vehicle until it cannot return. Step 3-1-2: From P i-1 To P i The path of the node is evenly divided into k+1 arcs, k is the number of divisions, and the next division node is updated to p 1 , until the judgment reaches p k , Step 3-1-3: Determine whether the original section node and the subsequently added split node can meet the return conditions. If so, patrol to the node; if not, return at the current node; The heuristic algorithm based on fixed-spacing segmentation adopted in step 3 is divided into two segmentation processes. The first segmentation process divides the original arc path into small segments with larger spacing. When it is determined which segment the UAV leaves from, the segment is segmented again to determine the specific location, d 1 is the spacing of the first segmentation process, d 2 is the spacing of the second segmentation process. The specific implementation steps are as follows: Step 3-2-1: From P 0 Start to determine whether the drone can return to the unmanned vehicle until it cannot return; Step 3-2-2: In the first segmentation process, the drone starts from P 0 Every d 1 Visit a point, when the visited split node position p next +d 1 Greater than the original node P i When the position of the split node is reached, the position of the split node will cross P according to the planned path i , so that the point falls on P i With P i+1 between; Step 3-2-3: In the second segmentation process, the drone moves every d 2 Visit a point and update the next split node. The size of the distance between the two processes affects the efficiency and accuracy of the algorithm. The distance d of the first split process 1 The spacing d of the second segmentation process is determined according to the map scale and the UAV's range. 2 Determined according to the required accuracy; In solving the return position of the drone, the binary method is used for reference to solve it, and a heuristic algorithm based on binary segmentation is proposed. After the mission starts, the unmanned vehicle plans a global arc path. Because the unmanned vehicle and the drone are in the same position at this time, the position of the unmanned vehicle is set to a and the end value is set to b. The specific implementation steps are as follows: Step 3-3-1: Determine whether the drone can return after the unmanned vehicle takes the same amount of time to travel from point b to point b. If it can, the drone's return point is point b or it does not need to return. Otherwise, use the binary search method to find the return point. Step 3-3-2: Set c = (a + b) / 2, and determine whether the difference between the remaining energy f(c) of the drone at that location and the energy g(c) used to return to the unmanned vehicle at that location satisfies a given accuracy ξ, that is, whether 0≤f(c)-g(c)≤ξ holds; Step 3-3-3: If it is true, the return point is c, otherwise if f(c)-g(c)>ξ, let a=c, if f(c)-g(c)<0, let b=c, and repeat until the return point is found; Step 3-3-4: When the drone returns, the drone and the unmanned vehicle are at the same position. The current position is set to a, and the binary search for the return value is performed again until the entire path inspection is completed.

Images