• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Improved lazy theta* algorithm based on octree map for path planning of UAV

    2023-05-31 01:33:18MengshunYuanTongleZhouMouChen
    Defence Technology 2023年5期

    Meng-shun Yuan,Tong-le Zhou,Mou Chen

    College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, 211106, China

    Keywords: Unmanned aerial vehicle Path planning Lazy theta* algorithm Octree map Line-of-sight algorithm

    ABSTRACT This paper investigates the path planning method of unmanned aerial vehicle (UAV) in threedimensional map.Firstly,in order to keep a safe distance between UAV and obstacles,the obstacle grid in the map is expanded.By using the data structure of octree,the octree map is constructed,and the search nodes is significantly reduced.Then,the lazy theta* algorithm,including neighbor node search,line-of-sight algorithm and heuristics weight adjustment is improved.In the process of node search,UAV constraint conditions are considered to ensure the planned path is actually flyable.The redundant nodes are reduced by the line-of-sight algorithm through judging whether visible between two nodes.Heuristic weight adjustment strategy is employed to control the precision and speed of search.Finally,the simulation results show that the improved lazy theta* algorithm is suitable for path planning of UAV in complex environment with multi-constraints.The effectiveness and flight ability of the algorithm are verified by comparing experiments and real flight.

    1.Introduction

    With the development of UAV technology,UAV has been applied in many fields.In the field of national defence,Military UAV can accomplish tasks such as intelligence reconnaissance,signal transmission,electronic countermeasure,target hitting and target aircraft [1].Path planning is an indispensable part of the UAV mission planning system,which can significantly improve the operational capability of military UAV,so it has been widely concerned in recent years and has become a research hotspot of military UAV technology [2].In path planning,the algorithm needs to plan a reasonable flight path for UAV from the starting point to the ending point in the given map[3].The path should meet the safety,rapidity and the corresponding constraints [4].Thus the path planning has always been the research hot topic of intelligent control and decision making for UAV.

    The first step of path planning is to build the map,and the classical method is grid map modeling.In Refs.[5,6],the regular grid map was used for its simplicity and convenience.In threedimensional map,the number of path nodes in regular grid maps is huge,and the search algorithm takes a long time to run.In Ref.[7],based on the octree data structure,the octree map significantly reduced the number of map path nodes,shorten the search time,and achieve better results.When the UAV is flying,it not only needs to consider the performance limitations of UAV,but also needs to consider the constraints of threat area,such as no fly zone.In Refs.[8-10],these methods translate avoiding threat zones and no-fly zones into geometric calculations that allow the UAV to eventually reach the target position.At the same time,there exists an obstacle edge collision problem in the grid map [11].Thus,it is necessary to expand the obstacles so that the UAV can keep a safe distance from the obstacles.

    After the completion of map modeling,path planning should be carried out [12].In three-dimensional path planning,some works can be found in the literature.In Ref.[13],an improved ant colony(ACO) algorithm was used in path planning.Focusing on the path planning problem of UAV,an improved particle swarm optimization (PSO) algorithm was proposed to plan path in Ref.[14].Improved Rapidly-Exploring Random Tree (RRT) algorithm has strong randomness and can quickly obtain the feasible path,but it can't obtain the optimized path [15,16].Heuristic algorithm is a series of algorithms derived from A*algorithm[17].In Ref.[18],the modified A*algorithm was based on the cost function and was used to solve the three dimensional path planning problem.In Ref.[19],the Theta*algorithm was proposed based on A*algorithm to solve the any angle path planning problem.In Ref.[20],the lazy theta*algorithm was applied to the grid map to plan the path.The planned path of lazy theta* algorithm not only has fewer nodes and shorter search time,but also is smooth,which is beneficial for the UAV to fly along the path.Based on the above research results,in order to quickly plan a safe,smooth and low-cost path [21],an optimization algorithm is proposed by combining octree map with lazy theta* algorithm.

    This paper is organized as follows.Firstly,problem description of the path planning problem is given,including four parts: UAV constraints,obstacle expansion,the octree map and neighbor node search.Then,the flow of the algorithm is described in detail.This algorithm first performs obstacle grid expansion on the original map information,then builds the octree map,and then uses the improved lazy theta* algorithm to plan path,which improves the neighbor node search,line-of-sight algorithm and heuristic weight adjustment.Finally,the simulation results are given to show the planned path which can meet the requirements.

    2.Problem statement

    The path planning problem of UAV can be transformed into an optimization problem with multiple constraints[22].The planning space is a three-dimensional grid map,and each grid is a search node.Therefore,the problem of path planning [23]can be described as: under the premise of satisfying the constraint conditions,the continuous path nodes are selected from the grid map nodes,so that the UAV has the minimum path cost when flying along the path nodes[24].

    2.1. UAV constraint and threat area constraint

    To guarantee that the planning path is actually flyable,it is necessary to consider UAV maneuverability,terrain,threat and other constraints [25].

    (1) Maximum yaw angle constraint:Due to the limitation of UAV maneuverability,the planned yaw angle should be less than the maximum yaw angle [26]as shown in Fig.1.The maximum yaw angle constraint can be expressed as

    Fig.1.Maximum yaw angle constraint.

    where,θirepresents the yaw angle at path point of sectioni,and θmaxrepresents the maximum yaw angle.

    (2) Maximum pitch angle constraint: Due to the influence of UAV thrust,gravity and other factors,there is a certain limit value of UAV's pitching ability [27].The maximum pitch angle constraint can be expressed as

    where,φirepresents the pitch angle at path point of sectioni,and φmaxrepresents the maximum pitch angle of the UAV.

    (3) Minimum path segment length constraint: Limited by the maneuverability of UAV,the UAV must guarantee a shortest straight-line flight distance before changing its course[28]as shown in Fig.2.The minimum path segment length constraint can be expressed as

    Fig.2.Path segment length.

    where,lirepresents the path length of sectioni,andlminrepresents the shortest direct flight distance of UAV.

    (4) Minimum relative height constraint: In order to ensure the flight safety,the UAV needs to keep a certain relative safe distance from the ground during the flight as shown in Fig.3.The minimum relative height constraint can be expressed as

    Fig.3.Minimum relative height constraint.

    where,zrepresents the altitude of the current path point,hminrepresents the minimum relative height,andhTindicates the terrain altitude directly below the UAV.

    (5) Threat area constraint:In order to ensure flight safety,UAV is required to be located outside the threat area at any time[29].The model of threat area needs to be built on the specific situation of threat objects [30],generally including sphere,cylinder,etc.Fig.4 shows the cylindrical threat zone,indicating the threat of the no fly zone.

    Fig.4.No fly zone constraint.

    2.2. Obstacle grid expansion

    The path searched directly in the grid map may collide with the edge of the obstacle grid[31].Fig.5 shows the collision problem of the grid edge in the two-dimensional map.This problem also exists in three-dimensional map.In order to ensure the safety of UAV flight path,it is necessary to deal with the edge of obstacles.

    Fig.5.Edge collision problem.

    The obstacles expansion method is used to expand the obstacles in the initial map.Based on the obstacles,a certain area around the obstacles is set as the obstacle area.Hence,the corresponding expansion range can be set according to the threat degree of the obstacles.After obtaining the processed map data,the octree map is constructed.The new map can keep a safe distance between UAV and obstacles in flight.

    2.3. Octree map model

    Octree map has been studied in the map model,which is established based on the data structure of octree.The octree map schematic diagram is shown in Fig.6.The process of building octree map is to divide the big cube into eight pieces until it becomes the minimum grid.Furthermore,the size of the minimum grid can be set.Generally,the smaller map grid indicates the higher search accuracy.In the data structure of octree,the whole big cube is the root node,and the smallest block is the leaf node.In the octree map,the volume of the upper node is eight times of that of the lower node.

    Fig.6.Octree map schematic diagram.

    In octree map,nodes store their own information,including coordinate,size and state.When there is no obstacle in the grid,it is marked as free grid,otherwise it is marked as obstacle grid.Octree map is different from the traditional grid.When all the child nodes of a node are in the same state,it is not necessary to expand this node.In the process of map construction,the grid will be automatically divided into smaller grids in the area with dense obstacles.Therefore,in octree map,the size of grid is different and the number of nodes is small.These advantages are conducive to the subsequent path planning.

    2.4. Neighbor node search

    This paper builds an octree map,so the neighbor node search method in the algorithm needs to be changed.Obtaining a neighbor node set is divided into four steps.

    Step 1.The coordinate ofsis(x,y,z).The length,width and height ofsare bothw.Build a new gridBoxwith coordinate(x,y,z).The length,width,and height ofBoxare bothw+2.The grid that intersects the new grid is added to the neighbor node set.Theneiis the neighbor node,andP(s)is the parent node ofs.

    Step 2.The grid in the neighbor node set is judged,and the grid whose state is the leaf node constitutes a new set of neighbor nodes.

    Step 3.Considering the limits of the yaw and pitch angles of the UAV,the maximum yaw and maximum pitch angles need to be reasonably preset.If the neighbor node meets the yaw angle constraint and pitch angle constraint,it will be retained,otherwise it will be discarded.The algorithm deletes nodes that do not meet the constraints,and then gets a new set of neighbors.

    Step 4.The grid state is determined and the free grids constitute the final set of neighbor nodes.

    Step 5.Determine if all nodes are searched ? If yes,the search is finished,otherwise return to Step 2.

    The flow chart of obtaining a neighbor node set is as follows(Fig.7):

    Fig.7.Flow chart of neighbor node search.

    3.Improved lazy theta* algorithm

    Before introducing the improved algorithm,some variables and functions of the algorithm are defined.

    Definition 1.Record the current node ass,the neighboring node of the current node iss′,the starting node issstart,the ending node issgoal,and the Euclidean distance is expressed byeuc(s,s′).

    Definition 2.The node coordinate is the grid center coordinate.For the given starting node and ending node coordinate,the algorithm takes the starting grid node coordinate as the starting node coordinate,and the ending grid node coordinate as the ending node coordinate.

    Definition 3.Define the Euclidean distance cost betweensands′asc(s,s′).The Manhattan distance cost betweensandsgoalis defined ash(s).

    Definition 4.The line-of-sight algorithm for judging whether there is an obstacle between two nodes is represented bylineofsight(s,s′).TheNeighb(s)is the set of neighbor nodes ofs.The argmin(s)can get the node with the lowest cost in the intersection set of theNeighb(s)and the close list.

    The algorithm first expands obstacle grids on the original map information,then builds the octree map,and then uses the improved lazy theta* algorithm to search path.Improved lazy theta* algorithm modifies the neighbor node search,line-of-sight algorithm and heuristic weight adjustment,etc.Through these improvements,the desired path is finally found.

    3.1. Line-of-sight algorithm

    Line-of-sight algorithm can judge whether there are obstacles between two points.For grid matrix environments,the line-ofsight algorithm and the line drawing algorithm are very similar.The line drawing algorithm needs to calculate the position of the grid that is filled when the line is drawn on the grid matrix.For lineof-sight algorithm,a line is drawn between two points to determine whether there are obstacles on the line.

    Let the coordinates of the two grids be(x0,y0,z0)and(x1,y1,z1),dx,dy,dzare the difference of the three-axis coordinates of the two points,idis a variable that controls whether the line-of-sight algorithm ends,anddis the maximum of the |dx|,|dy|,|dz|.Thestepis the step size,anddistis the distance between the two points on the two-dimensional plane,and the calculation equation is as follows:

    The flow chart of line-of-sight algorithm is shown in Fig.8.In this figure,(ix,iy,iz)is the coordinate of the current step,andmap[x][y][z]==1 indicates that there is obstacle in the grid at coordinate(x,y,z).

    Fig.8.Line-of-sight algorithm.

    In Fig.8,condition 1 is shown in Eq.(6),andlen1 is the maximum size of the map.

    When the return value is true,there is no obstacle on the line between two points.When the return value is false,there are obstacles on the line between two points.

    3.2. Heuristics weight adjustment

    In the process of neighbor node search,the form of cost function can be expressed as

    whereg(s)represents the true cost from the starting node to the current nodes,h(s)represents the estimated cost of the current nodesto the ending node,and ε is the heuristic weighting factor(ε ≥1).The update of the heuristic weight is related to the location of the current nodes.If the distance between the current node and the ending node is farther,the planning speed will be faster.If the current node is closer to the ending node,the planning accuracy is higher.The update method is as follows:

    where △ε can be adjusted according to requirements.By increasing the weight factor of the heuristic,the search direction of the algorithm tends to end the node,thus reducing the search time of the algorithm.However,the weight of the heuristics should not be large,otherwise the algorithm will no longer have acceptability and will not satisfy the consistency condition.

    3.3. Path planning based on improved lazy theta* algorithm

    The main program of the improved lazy theta* algorithm is based on the lazy theta*algorithm,and the octree map modeling is added.The cost calculation method in the algorithm is described in detail in section 3.2.The main steps of the algorithm are as follows:

    (1) Processing the planning map,expanding the obstacles,and building an octree map model based on the data structure of octree;

    (2) Constructing open list and close list.During the search,the nodes waiting to be searched are added to the open list,and the searched nodes are added to the closed list;

    (3) Inserting the start node into the open list;

    (4) If the open list is empty,the calculation is stopped and the path search fails,otherwise performs the next step.

    (5) Selecting the node with the minimum cost in the open list as the new current nodes,removingsfrom the open list and inserting it to the close list;

    (6) Determining whethersis visible to its parent node.If the two points are visible,the algorithm performs the next step.Otherwise,the algorithm needs to find the node with the minimum cost from the intersection of theNeighb(s)and the close list,and set it as the parent node ofs;

    (7) Determining whether the current node is the end node?If it is the end node then the path is found.Otherwise search for neighbor nodes and add them to the open list,and return to Step 4).

    If the path is found,the complete path is obtained according to the relationship between the nodes.The main program flow chart of the improved lazy theta*algorithm is shown as follows (Fig.9).

    Fig.9.Flow chart of algorithm.

    In order to explain the algorithm clearly,the pseudocode of the main part of the algorithm is given below.The algorithm main function is shown in the pseudocode in Fig.10.FunctionMapCreate(map)is the octree map building function.FunctionSetvertex(s)sets parent node of current node.UpdateVertex(s,s′)updates the cost of adjacent nodes′.

    Fig.10.Main function.

    The sub-function is divided into two parts,one of which is shown in the pseudocode in Fig.11.This part is the basic subfunction of lazy teta * algorithm.FunctionSetvertex(s)determines whether s and its parent node are visible,corresponding to line 9 in the main function.UpdateVertex(s,s′)updates the cost of adjacent nodes′,corresponding to line 18 in the main function.ComputeCost(s,s′)is the function thatUpdateVertex(s,s′)needs,which calculates the cost of nodes′and sets its parent node.

    Fig.11.Help function 1.

    Another part of the sub-function is shown inHelperFunction2 in Fig.12.This part is octree map modeling and neighbor node search functions.MapCreate(map)is a function which builds an octree map based on the map information after the obstacle grids are expanded,corresponding to line 2 in the main function.The function is a recursive function that constantly calls itself until the map is divided into the smallest grid of constraints and finally returns the root noderoot.Node()is a node,set as a class in the program,which stores some information,including coordinate,size,node state,and so on.Theroot.valuemarks whether the node is obstacle grid,Trueindicates free grid,andFalseindicates obstacle grid.AllValueSame(map)is used to determine whether the grid in the map cube is in the same state.If the states are the same,the blocking of map is no longer continued,and the node state is marked and set as a leaf node.Theroot.ftopleftis the eighth subgrid of the current grid,and the omitting part is the construction of the other seven sub-grids,and the method is the same asroot.ftopleft.

    Fig.12.Help function 2.

    Neighb(s)is a neighbor node search function,and the function finally returns a set of neighbor node that meets the constraints,corresponding to line 13 in the main function.The function first usesIntersect(Box,s)to get the leaf nodes adjacent tos,then calculates the angle in the path according toAngle(P(s),s,nei),takes the nodes within the maximum yaw angle θmaxand the maximum pitch angle φmax,and finally adds the free grid to the set of neighbor nodes.

    4.Simulation results and analysis

    To verify the effectiveness of the improved Lazy Theta* algorithm,Windows 10 is used as the platform,and Python 3.6 is used to simulate the programming environment.The simulation hardware platform is Intel Core i5 4210 M processor,which has a clock speed of 2.6 GHz and 8 GB of memory.Under the same hardware platform,the algorithm is verified by using different sizes of maps.Then,in order to further illustrate the role of the algorithm,the ACO algorithm,improved ACO algorithm in Ref.[13],Improved RRT algorithm in Ref.[15],the sparse A* algorithm and the lazy theta*algorithm are selected as the comparison objects.In the case of a fixed starting position,a fixed ending position,and the same map,comparative experiments are conducted to analyze the algorithm.Finally,the planned path is transmitted to the UAV for actual flight verification.

    4.1. Simulation results

    In the simulation study,the parameters are selected as follows:the maximum yaw angle is 60°,the maximum pitch angle is 45°;the minimum relative height is 0.1 km;the minimum path length is 0.1 km;the initial heuristic weight coefficient is 1.3;the minimum grid length,width and height both are 0.1 km.

    First,on the mountain map 1 with the size of 64×64×64,the algorithm is used to search the path.In map 1,the first peak of mountain threat is (2 km,1.8 km,1.1 km),the second peak is(4.2 km,2 km,1.5 km),and the third peak is(3.5 km,4 km,2 km).Set the starting position of the UAV to(0.4 km,0.4 km,0.4 km),the ending position is(6 km,6 km,1.2 km).The path planned without the line-of-sight algorithm is shown in Fig.13,and the planned path with the line-of-sight algorithm is shown in Fig.14,where the black box is the sparse grid where the path node is located.The black line segment represents the path of UAV.

    Fig.13.Path without the line-of-sight algorithm in map 1.

    Fig.14.Path in map 1.

    Add threat area to map 1 to form map 2.The first center of the sphere threat area is (1.2 km,0.9 km,0.4 km),and the radius is 0.4 km;the second center of the sphere threat area is(1 km,3 km,0.5 km),and the radius is 0.5 km.The cylinder threat area is weather threat,the plane center coordinates are (4.7 km,5.3 km),the radius is 0.5 km,and the height is 1.5 km.The start and end position of the search are the same as before.The path planned without the line-of-sight algorithm is shown in Fig.15,and the planned path with the line-of-sight algorithm is shown in Fig.16.

    Fig.15.Path without the line-of-sight algorithm in map 2.

    Fig.16.Path in map 2.

    On the threat map 3 with the size of 64×64×64,the algorithm is used to search the path.In the map 3,the cylinder,sphere and cube are the threat area.Set the starting position of the UAV to(0.4 km,0.4 km,0.4 km),the ending position is(6 km,6 km,1.2 km),The planned path is shown in Fig.17.

    Fig.17.Path in map 3.

    On the mountain map 4 with the size of 128 × 128 × 128,the algorithm is used to search the path.In map 4,the first peak of mountain threat is (4 km,3.6 km,2.2 km),the second peak is(8.4 km,4 km,3 km),and the third peak is(7 km,8 km,4 km).Set the starting position to (0.8 km,0.8 km,0.8 km),the ending position is (12 km,12 km,2.4 km),and the planned path is shown in Fig.18.

    Fig.18.Path in map 4.

    In different maps,the information of the path planned by the improved lazy theta*algorithm is shown in Table 1.

    Table 1 Path data of different map.

    Table 2 Path data of different algorithm in map 1.

    From the visual results of different maps,the improved lazy theta* algorithm can effectively solve the problem of threedimensional path planning for UAV.The planned path can keep a safe distance from the obstacles because the obstacles are expanded.Due to the angle constraint in the algorithm,the yaw angle and pitch angle of UAV at the path node are less than the preset extreme value.Because the line-of-sight algorithm can delete redundant nodes,the number of nodes in the path is low.

    Analysis of the data in Table 1 shows that the following conclusions.When the terrain is complex,the octree map modeling time and path searching time will increase.When the grid map in three-dimensional space increases,the map modeling time increases and the path search time becomes longer,but the number of path nodes does not change much.Moreover,the octree map modeling in the algorithm requires a certain time.In practical applications,the octree map data can be directly imported into the algorithm,which saves the time of mapping modeling.

    4.2. Comparative simulation

    On the mountain map 1 with the size of 64 × 64 × 64,set the starting position of the UAV to(0.4 km,0.4 km,0.4 km),the ending position is(6 km,6 km,1.2 km).The ACO algorithm,improved ACO algorithm in Ref.[13],Improved RRT algorithm in Ref.[15],sparse A* algorithm,lazy theta* algorithm and improved lazy theta*algorithm are used to search the path,and the search time,the number of nodes and the length are recorded.

    On the mountain map 2 with the size of 64 × 64 × 64,set the starting position of the UAV to(0.4 km,0.4 km,0.4 km),the ending position is (6 km,6 km,1.2 km).The information of the path planned by different algorithm is shown in Table 3.

    Table 3 Path data of different algorithm in map 2.

    From the analysis of the data in Table 2 and Table 3,it is clear that ACO algorithm,improved ACO algorithm and improved RRT algorithm have poor performance,with long algorithm search time,high number of nodes and long length.This is due to the excessive number of path nodes searched by the algorithm.Sparse A* Algorithm has a short search time and a low cost,because it is highly suggestive and has few search nodes.The lazy theta*algorithm has a long search time,a small number of nodes,and a slightly shorter length.The lazy theta* algorithm is judged by the line-of-sight algorithm,and the intermediate node can be omitted,hence the number of nodes is small.By reducing the number of nodes,the algorithm also reduces the path searching time and path cost.The improved lazy theta*algorithm has short search time,short length and few nodes.The algorithm in this paper keeps the advantages of lazy theta* algorithm and further reduces the path cost by constructing octree map,which makes the path search time significantly reduced.

    4.3. Actual flight verification

    In order to verify the actual flight feasibility of the planned path,the DJI Matrice M600 Pro UAV is used as the platform.At the same time,the manifold 2 airborne computer is installed on the M600 Pro to make it have the function of automatic cruise.On the mountain map 2 with the size of 64 × 64 × 64,set the starting position of the UAV to(0.4 km,0.4 km,0.4 km),the ending position is(6 km,6 km,1.2 km).The planned path is shown in Fig.16.Since the M600 Pro UAV can only fly short distances,the length,width and height of the minimum grid will be set to 1 m when the UAV flies.The length,width and height of the minimum grid will be set to 100 m when drawing.The UAV and remote control equipment are shown in Fig.19.The flight diagram of UAV is shown in Fig.20.

    Fig.19.UAV and remote control equipment.

    Fig.20.The flight diagram of UAV.

    The Manifold 2 airborne computer receives the planned flight path and then performs actual flight verification.The path planned by the algorithm consists of a series of path nodes,which record the three dimensional coordinates of the position.For the whole path,a series of coordinate deviation are obtained by subtracting the coordinates of the previous node from the coordinates of the latter node.The coordinate deviation is continuously sent to the manifold 2 airborne computer as a flight command,and the UAV can fly along the planned path.The actual flight path is shown in Fig.21.

    Fig.21.Path in map 2.

    The actual flight distance and position deviation of the UAV are shown in Table 4.

    Table 4 Path data of actual flight.

    According to the visualization results and flight data,the deviation between the actual flight path and the planned flight path is small.The UAV can track the planned path well,which verifies the effectiveness of the algorithm.

    5.Conclusions

    (1) In this paper,an improved algorithm based on lazy theta*algorithm is proposed.The obstacle grid expansion is used to keep the UAV and the obstacle at a safe distance,and the octree data structure is used to build a map with fewer path nodes.lazy theta*is used to improve the speed of the search and the smoothness of the path.The neighbor node search,line-of-sight algorithm and heuristic weight adjustment are improved.Thus,the improved algorithm is successfully applied to the octree map,and the paths are finally planned.

    (2) Compared with the path planned by ACO algorithm,improved ACO algorithm in Ref.[13],Improved RRT algorithm in Ref.[15],the improved lazy theta*algorithm tends to search in the target direction when it searches nodes in the octree map.Therefore,the search speed is fast and there are fewer search nodes.At the same time,the node path cost is calculated when searching,and the path with less path cost can be obtained.Finally,the redundant nodes are deleted by the line-of-sight algorithm,which can make the path smooth and less expensive.Therefore,the improved lazy theta* algorithm can plan fast,smooth,safe and low cost paths.

    (3) The algorithm is a fast path planning method in threedimensional static environment.The position error of the algorithm in the UAV flight verification meets the requirements,so it can be applied to the military UAV mission planning system,which is conducive to the military UAV to perform various complex tasks and improve the combat capability.At the same time,the military UAV is facing sudden threats when performing tasks.It is necessary to quickly reconstruct the map and plan a new path.Octree map has the characteristics of real-time construction.And the algorithm has fast search speed and excellent search results.Therefore,the next step will study the online path planning method of military UAV in dynamic environment,includes establishing more effective map and threat model,optimizing the calculation function and actual flight verification.

    Declaration of competing interest

    The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

    Acknowledgements

    This work was supported in part by the National Natural Science Foundation of China under Grant U2013201 and in part by the Key R&D projects (Social Development) in Jiangsu Province of China under Grant BE2020704.

    天堂影院成人在线观看| 亚洲国产高清在线一区二区三| 最近的中文字幕免费完整| 国产精品女同一区二区软件| 日日摸夜夜添夜夜爱| 男人爽女人下面视频在线观看| 韩国av在线不卡| 久久久久久久午夜电影| 国产一区二区三区av在线| 日韩人妻高清精品专区| 可以在线观看毛片的网站| 天堂网av新在线| 建设人人有责人人尽责人人享有的 | 91精品一卡2卡3卡4卡| 韩国高清视频一区二区三区| a级毛色黄片| 成年免费大片在线观看| 超碰97精品在线观看| 国产精品一区www在线观看| 99热全是精品| 99热这里只有是精品在线观看| 晚上一个人看的免费电影| 大又大粗又爽又黄少妇毛片口| 99视频精品全部免费 在线| 3wmmmm亚洲av在线观看| 日韩成人av中文字幕在线观看| 在线免费观看不下载黄p国产| 免费观看的影片在线观看| 亚洲成人中文字幕在线播放| 黄片wwwwww| 在线观看一区二区三区| 只有这里有精品99| 亚洲精品亚洲一区二区| av网站免费在线观看视频 | 亚洲激情五月婷婷啪啪| av免费在线看不卡| 一级毛片 在线播放| 少妇的逼好多水| 国内揄拍国产精品人妻在线| 国产午夜福利久久久久久| 男女边摸边吃奶| 人妻夜夜爽99麻豆av| 日本午夜av视频| 欧美日韩精品成人综合77777| 国产男人的电影天堂91| 国精品久久久久久国模美| 又爽又黄无遮挡网站| 国产亚洲一区二区精品| 亚洲人成网站在线观看播放| 网址你懂的国产日韩在线| 亚洲激情五月婷婷啪啪| or卡值多少钱| 韩国高清视频一区二区三区| 国产成人一区二区在线| av在线观看视频网站免费| 欧美另类一区| 国产麻豆成人av免费视频| 精品久久久噜噜| 国产男女超爽视频在线观看| 国产成人91sexporn| 69av精品久久久久久| 成人亚洲精品av一区二区| av在线亚洲专区| 国产极品天堂在线| 色综合站精品国产| 欧美xxxx性猛交bbbb| 国产淫语在线视频| 天天躁夜夜躁狠狠久久av| 色吧在线观看| 高清欧美精品videossex| 在线观看人妻少妇| 国产成人精品久久久久久| 国内揄拍国产精品人妻在线| 成人无遮挡网站| 亚洲精品国产av蜜桃| 国产成人freesex在线| 国产亚洲5aaaaa淫片| 午夜激情久久久久久久| av女优亚洲男人天堂| 观看免费一级毛片| 插逼视频在线观看| 国产成年人精品一区二区| 日本一二三区视频观看| 99热这里只有精品一区| 夫妻性生交免费视频一级片| 美女国产视频在线观看| 能在线免费看毛片的网站| 亚洲精品一区蜜桃| 大话2 男鬼变身卡| 一边亲一边摸免费视频| 午夜精品在线福利| 精品一区二区三卡| 午夜福利在线在线| 偷拍熟女少妇极品色| 又爽又黄无遮挡网站| 伊人久久精品亚洲午夜| 国产不卡一卡二| 精品久久久久久久末码| 久久久a久久爽久久v久久| 日本-黄色视频高清免费观看| 99久久中文字幕三级久久日本| 久久久色成人| 秋霞伦理黄片| 91久久精品电影网| 久久久久久久午夜电影| 国产乱来视频区| 91狼人影院| 欧美日韩在线观看h| 中国国产av一级| 水蜜桃什么品种好| 亚洲av中文字字幕乱码综合| 国产爱豆传媒在线观看| 日韩强制内射视频| 国产淫片久久久久久久久| 尤物成人国产欧美一区二区三区| 午夜免费观看性视频| 一个人看的www免费观看视频| 22中文网久久字幕| 91狼人影院| 亚洲精品一区蜜桃| 久久久久国产网址| 天堂中文最新版在线下载 | 免费少妇av软件| 中文乱码字字幕精品一区二区三区 | 老司机影院毛片| 人人妻人人澡欧美一区二区| 免费观看性生交大片5| 国产69精品久久久久777片| 国产免费又黄又爽又色| 午夜日本视频在线| 狂野欧美激情性xxxx在线观看| 久久精品国产亚洲网站| av在线亚洲专区| 精品久久久久久久久亚洲| 亚洲国产精品成人综合色| 少妇人妻精品综合一区二区| 国产成人一区二区在线| www.色视频.com| 午夜精品在线福利| 在线播放无遮挡| 亚洲美女视频黄频| 欧美日韩视频高清一区二区三区二| 久热久热在线精品观看| 亚洲经典国产精华液单| 免费av毛片视频| 午夜老司机福利剧场| 国产一区有黄有色的免费视频 | 91在线精品国自产拍蜜月| 亚洲精品乱码久久久v下载方式| .国产精品久久| av又黄又爽大尺度在线免费看| 亚洲欧美中文字幕日韩二区| 99re6热这里在线精品视频| 国产一区二区在线观看日韩| 午夜福利视频1000在线观看| 欧美一级a爱片免费观看看| 日韩欧美 国产精品| 国产精品爽爽va在线观看网站| 超碰97精品在线观看| 一个人看的www免费观看视频| 亚洲精品乱久久久久久| 青春草国产在线视频| 九草在线视频观看| 在线免费观看的www视频| 国产成人aa在线观看| 亚洲欧美成人精品一区二区| 国产三级在线视频| 成人无遮挡网站| 国产伦精品一区二区三区四那| 亚洲欧美精品专区久久| 一个人免费在线观看电影| 亚洲久久久久久中文字幕| 一个人看的www免费观看视频| 精品午夜福利在线看| 熟妇人妻不卡中文字幕| 高清午夜精品一区二区三区| 成人特级av手机在线观看| 午夜福利成人在线免费观看| 国产熟女欧美一区二区| 超碰av人人做人人爽久久| 特大巨黑吊av在线直播| 亚洲电影在线观看av| 舔av片在线| 麻豆精品久久久久久蜜桃| 看免费成人av毛片| 美女黄网站色视频| 麻豆精品久久久久久蜜桃| 亚洲经典国产精华液单| 噜噜噜噜噜久久久久久91| 青春草国产在线视频| 三级经典国产精品| 亚洲av成人精品一二三区| 91在线精品国自产拍蜜月| 精品亚洲乱码少妇综合久久| 内射极品少妇av片p| 国产精品无大码| 欧美成人a在线观看| 最近最新中文字幕大全电影3| 身体一侧抽搐| 狂野欧美激情性xxxx在线观看| 99热这里只有精品一区| 国产爱豆传媒在线观看| 天天躁日日操中文字幕| 天堂影院成人在线观看| av播播在线观看一区| 国产精品久久视频播放| 国产亚洲精品av在线| 最新中文字幕久久久久| 亚洲国产色片| 免费大片黄手机在线观看| 国产一区亚洲一区在线观看| 成年免费大片在线观看| 国产午夜福利久久久久久| 国产精品精品国产色婷婷| 蜜桃久久精品国产亚洲av| 国产高潮美女av| 久久综合国产亚洲精品| 老女人水多毛片| 久久久午夜欧美精品| 久久6这里有精品| 国产在视频线在精品| 男女啪啪激烈高潮av片| 性插视频无遮挡在线免费观看| 亚洲三级黄色毛片| 日韩不卡一区二区三区视频在线| 免费黄色在线免费观看| 看免费成人av毛片| 2018国产大陆天天弄谢| 中文字幕制服av| av网站免费在线观看视频 | 国产午夜精品一二区理论片| 国产精品一及| 99久国产av精品| 永久网站在线| 日日摸夜夜添夜夜爱| 可以在线观看毛片的网站| 在线观看一区二区三区| 性色avwww在线观看| 日韩成人伦理影院| 亚洲av福利一区| 极品教师在线视频| 男女啪啪激烈高潮av片| 亚洲欧美清纯卡通| 国产又色又爽无遮挡免| 蜜桃亚洲精品一区二区三区| 三级毛片av免费| 蜜臀久久99精品久久宅男| ponron亚洲| 国产午夜精品久久久久久一区二区三区| 久久久成人免费电影| 六月丁香七月| 亚洲av不卡在线观看| 免费少妇av软件| 一区二区三区高清视频在线| 男女国产视频网站| 在线免费观看不下载黄p国产| 欧美成人午夜免费资源| av在线蜜桃| 少妇的逼好多水| 久久97久久精品| 亚洲成人久久爱视频| 日韩成人伦理影院| 菩萨蛮人人尽说江南好唐韦庄| 综合色av麻豆| 亚洲av一区综合| 卡戴珊不雅视频在线播放| 午夜福利成人在线免费观看| 大香蕉久久网| 我的女老师完整版在线观看| 一区二区三区乱码不卡18| 99久久人妻综合| 午夜福利高清视频| 淫秽高清视频在线观看| 国产v大片淫在线免费观看| 精品一区二区三卡| 免费av不卡在线播放| 久久久久久伊人网av| 久久久欧美国产精品| 亚洲四区av| 最近最新中文字幕大全电影3| 国产乱人视频| 国产三级在线视频| 欧美日韩视频高清一区二区三区二| 国产精品熟女久久久久浪| 久久久欧美国产精品| 精品国产三级普通话版| 一区二区三区四区激情视频| 一级毛片我不卡| 可以在线观看毛片的网站| 国产精品1区2区在线观看.| 国产精品久久久久久精品电影| 毛片女人毛片| 搡老乐熟女国产| 亚洲高清免费不卡视频| 国产淫语在线视频| 亚洲精品视频女| 97在线视频观看| 成人av在线播放网站| 少妇的逼水好多| 久久久午夜欧美精品| 欧美一区二区亚洲| av黄色大香蕉| 熟女电影av网| 亚洲国产av新网站| 春色校园在线视频观看| 精品久久久久久久久久久久久| 成人毛片a级毛片在线播放| 国产精品女同一区二区软件| 毛片一级片免费看久久久久| 婷婷色麻豆天堂久久| 少妇的逼好多水| 毛片女人毛片| 国产激情偷乱视频一区二区| 亚洲乱码一区二区免费版| 搡老乐熟女国产| 亚洲av电影不卡..在线观看| 精品熟女少妇av免费看| 国产欧美另类精品又又久久亚洲欧美| 亚洲不卡免费看| 汤姆久久久久久久影院中文字幕 | 男人狂女人下面高潮的视频| 亚洲精品,欧美精品| 在现免费观看毛片| 大陆偷拍与自拍| 国产av国产精品国产| 尾随美女入室| 一级毛片久久久久久久久女| 中文资源天堂在线| 我要看日韩黄色一级片| 久久久欧美国产精品| 一级a做视频免费观看| videos熟女内射| 免费播放大片免费观看视频在线观看| 亚洲av电影不卡..在线观看| 少妇被粗大猛烈的视频| 亚洲国产成人一精品久久久| 91aial.com中文字幕在线观看| 97精品久久久久久久久久精品| 国产免费福利视频在线观看| 亚洲av免费高清在线观看| 国产高清不卡午夜福利| 精品久久久精品久久久| 伦理电影大哥的女人| 亚洲最大成人av| av专区在线播放| 欧美精品国产亚洲| 亚洲av成人精品一区久久| 久久精品夜夜夜夜夜久久蜜豆| 麻豆av噜噜一区二区三区| 亚洲精品亚洲一区二区| 丝袜美腿在线中文| 99热这里只有是精品50| 日本免费在线观看一区| 国产精品1区2区在线观看.| 久久久久精品久久久久真实原创| 欧美成人一区二区免费高清观看| 日日摸夜夜添夜夜添av毛片| 日韩伦理黄色片| 美女cb高潮喷水在线观看| 18+在线观看网站| 亚洲伊人久久精品综合| 免费黄色在线免费观看| 亚洲av二区三区四区| 久久久精品欧美日韩精品| 午夜精品国产一区二区电影 | 亚洲精品第二区| 精品一区二区三卡| 欧美日韩在线观看h| 一区二区三区四区激情视频| 中文字幕av在线有码专区| 欧美bdsm另类| 简卡轻食公司| 免费无遮挡裸体视频| 99久久精品热视频| 午夜激情福利司机影院| h日本视频在线播放| 午夜亚洲福利在线播放| 日韩视频在线欧美| 久久国产乱子免费精品| 人妻少妇偷人精品九色| 亚洲,欧美,日韩| 亚洲欧洲日产国产| 丝瓜视频免费看黄片| 国产精品国产三级专区第一集| 成人性生交大片免费视频hd| 三级经典国产精品| 九色成人免费人妻av| 色尼玛亚洲综合影院| 亚洲欧美精品专区久久| 一区二区三区乱码不卡18| 搞女人的毛片| 激情五月婷婷亚洲| 亚洲不卡免费看| 免费观看a级毛片全部| 最近手机中文字幕大全| 欧美变态另类bdsm刘玥| 亚洲国产精品专区欧美| 亚洲精品第二区| 欧美区成人在线视频| 欧美精品国产亚洲| 久久久久久久大尺度免费视频| av在线老鸭窝| 欧美+日韩+精品| 国产成人精品久久久久久| 国产不卡一卡二| 18禁裸乳无遮挡免费网站照片| 亚洲综合精品二区| 如何舔出高潮| 亚洲在久久综合| 免费播放大片免费观看视频在线观看| 午夜激情欧美在线| 欧美激情久久久久久爽电影| 亚洲精品国产成人久久av| av线在线观看网站| 免费av不卡在线播放| 狂野欧美激情性xxxx在线观看| 免费黄频网站在线观看国产| 99久久九九国产精品国产免费| 一区二区三区高清视频在线| 国产精品久久久久久精品电影| 久久久久精品性色| 亚洲av中文字字幕乱码综合| 亚洲精品日韩av片在线观看| 夫妻午夜视频| 性色avwww在线观看| 如何舔出高潮| 亚洲精华国产精华液的使用体验| 成人亚洲欧美一区二区av| 深夜a级毛片| 国产视频首页在线观看| 99久国产av精品国产电影| 午夜福利成人在线免费观看| 亚洲精品一二三| 欧美精品一区二区大全| 搡女人真爽免费视频火全软件| 精品一区在线观看国产| 男女视频在线观看网站免费| 亚洲一区高清亚洲精品| 天堂av国产一区二区熟女人妻| 你懂的网址亚洲精品在线观看| 日本与韩国留学比较| 色综合站精品国产| 久久久色成人| 99久久人妻综合| 欧美激情久久久久久爽电影| 国产黄频视频在线观看| 亚洲欧洲日产国产| 男人和女人高潮做爰伦理| 丝瓜视频免费看黄片| 熟妇人妻久久中文字幕3abv| 久久久亚洲精品成人影院| 在线免费观看的www视频| 国产av国产精品国产| 男人爽女人下面视频在线观看| 能在线免费观看的黄片| 特大巨黑吊av在线直播| 99久久人妻综合| 成年版毛片免费区| 亚洲精品成人久久久久久| av在线天堂中文字幕| 久久久久精品性色| 成人一区二区视频在线观看| 麻豆久久精品国产亚洲av| 国产精品一及| 最近手机中文字幕大全| 国产亚洲av片在线观看秒播厂 | 美女脱内裤让男人舔精品视频| 最新中文字幕久久久久| 免费观看无遮挡的男女| 国产精品久久久久久久久免| 伦精品一区二区三区| 国产一区二区三区av在线| 噜噜噜噜噜久久久久久91| 免费在线观看成人毛片| 80岁老熟妇乱子伦牲交| 男人舔奶头视频| 淫秽高清视频在线观看| 少妇人妻精品综合一区二区| 人人妻人人看人人澡| 国产极品天堂在线| 久久综合国产亚洲精品| 成人二区视频| 精品酒店卫生间| 精品久久久久久电影网| 美女内射精品一级片tv| 九九爱精品视频在线观看| 夫妻性生交免费视频一级片| 一级毛片 在线播放| 大香蕉97超碰在线| 观看免费一级毛片| av在线蜜桃| 一级毛片电影观看| 久99久视频精品免费| 久久久欧美国产精品| 久久精品人妻少妇| 青春草亚洲视频在线观看| 国产视频内射| 国产伦一二天堂av在线观看| 国产真实伦视频高清在线观看| 天美传媒精品一区二区| 久久草成人影院| 久久久久久久久久久免费av| av免费在线看不卡| 久久久久精品性色| 国产精品不卡视频一区二区| 亚洲自拍偷在线| 国产黄片视频在线免费观看| 丝瓜视频免费看黄片| 国产午夜福利久久久久久| 久久久久久国产a免费观看| 最近的中文字幕免费完整| 国产永久视频网站| 熟妇人妻久久中文字幕3abv| 亚洲精品色激情综合| 91精品国产九色| 亚洲真实伦在线观看| 男插女下体视频免费在线播放| 成人亚洲欧美一区二区av| 久久97久久精品| 欧美高清成人免费视频www| 色哟哟·www| 可以在线观看毛片的网站| 天堂√8在线中文| 国产色婷婷99| 美女国产视频在线观看| 干丝袜人妻中文字幕| 欧美日韩在线观看h| 国产精品一二三区在线看| 亚洲美女视频黄频| 日韩一区二区视频免费看| 一级毛片久久久久久久久女| av.在线天堂| 一级黄片播放器| 91av网一区二区| 国产精品麻豆人妻色哟哟久久 | 99久久精品一区二区三区| 69人妻影院| 国产免费福利视频在线观看| 99久国产av精品| 久久久国产一区二区| 国产成人freesex在线| 亚洲精品日本国产第一区| 精品久久久久久久久亚洲| 中文在线观看免费www的网站| 乱人视频在线观看| 久久久久久久久久久免费av| 97精品久久久久久久久久精品| 亚洲怡红院男人天堂| 国产淫语在线视频| 国产成人a∨麻豆精品| 亚洲精品自拍成人| 天堂影院成人在线观看| 欧美一区二区亚洲| 国内精品美女久久久久久| 18禁动态无遮挡网站| 精品午夜福利在线看| 看免费成人av毛片| 水蜜桃什么品种好| 亚洲国产精品专区欧美| 国产午夜精品论理片| 亚洲四区av| 日韩中字成人| 少妇的逼水好多| 一区二区三区乱码不卡18| 亚洲自拍偷在线| 欧美97在线视频| 亚洲欧美日韩卡通动漫| 久久精品人妻少妇| videos熟女内射| 日本免费a在线| 国产女主播在线喷水免费视频网站 | 免费av不卡在线播放| 日韩一区二区三区影片| 97超视频在线观看视频| 不卡视频在线观看欧美| 久久久色成人| 精品国产三级普通话版| 国产亚洲精品av在线| 亚洲最大成人手机在线| 国产不卡一卡二| 国产又色又爽无遮挡免| 免费观看的影片在线观看| 日韩欧美精品v在线| 午夜精品一区二区三区免费看| 91久久精品国产一区二区成人| 亚洲av国产av综合av卡| 国产一区有黄有色的免费视频 | 日韩在线高清观看一区二区三区| 美女cb高潮喷水在线观看| 欧美日韩精品成人综合77777| 精品久久久久久久人妻蜜臀av| 六月丁香七月| 黄片无遮挡物在线观看| 一级爰片在线观看| 色播亚洲综合网| 男人舔奶头视频| 久久久久久久亚洲中文字幕| 日韩av在线大香蕉| av在线观看视频网站免费| 99热这里只有是精品50| 在线观看免费高清a一片| 男人舔奶头视频| 国产成人精品一,二区| 一个人观看的视频www高清免费观看| 国产极品天堂在线| 夜夜爽夜夜爽视频| 日日摸夜夜添夜夜爱| 日产精品乱码卡一卡2卡三| 国产色爽女视频免费观看| 日日摸夜夜添夜夜爱| 日产精品乱码卡一卡2卡三| 夜夜爽夜夜爽视频| 欧美xxxx黑人xx丫x性爽| 亚洲精品日韩av片在线观看| 小蜜桃在线观看免费完整版高清|