• <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.

    在线a可以看的网站| 亚洲久久久久久中文字幕| 激情 狠狠 欧美| 亚洲精品日韩av片在线观看| 日本成人三级电影网站| 国产精品久久电影中文字幕| 色哟哟·www| 日韩精品青青久久久久久| 欧美潮喷喷水| 三级毛片av免费| 搡老妇女老女人老熟妇| 国产高清激情床上av| 国内精品美女久久久久久| 一级毛片aaaaaa免费看小| 在线免费观看不下载黄p国产| 免费看美女性在线毛片视频| 我的女老师完整版在线观看| 色在线成人网| 国产精品福利在线免费观看| 伦精品一区二区三区| 中文在线观看免费www的网站| 美女 人体艺术 gogo| av卡一久久| 波多野结衣高清作品| 精品一区二区三区人妻视频| 国产成人一区二区在线| 十八禁国产超污无遮挡网站| 麻豆乱淫一区二区| 一边摸一边抽搐一进一小说| 亚洲av.av天堂| eeuss影院久久| 老熟妇乱子伦视频在线观看| 国产在视频线在精品| 少妇被粗大猛烈的视频| 精品少妇黑人巨大在线播放 | 国内精品美女久久久久久| 国产免费男女视频| 日本在线视频免费播放| 国产极品精品免费视频能看的| 国产探花极品一区二区| 精品国产三级普通话版| 乱码一卡2卡4卡精品| 亚洲真实伦在线观看| 国产精品一二三区在线看| 久久精品国产清高在天天线| 亚洲精品一卡2卡三卡4卡5卡| 午夜精品国产一区二区电影 | 美女xxoo啪啪120秒动态图| 日本免费一区二区三区高清不卡| 国产亚洲精品av在线| 三级毛片av免费| 一区二区三区免费毛片| 美女内射精品一级片tv| 国产精品免费一区二区三区在线| 人妻丰满熟妇av一区二区三区| 悠悠久久av| 亚洲精品456在线播放app| 99久久无色码亚洲精品果冻| 国产一区二区激情短视频| 免费看a级黄色片| 亚洲经典国产精华液单| 国产成人a区在线观看| 91久久精品国产一区二区成人| 久久久久久久久大av| 婷婷精品国产亚洲av| 精品久久久噜噜| 国产淫片久久久久久久久| 久久精品影院6| 日本免费一区二区三区高清不卡| 亚洲性久久影院| 国产乱人偷精品视频| 99久国产av精品国产电影| 色吧在线观看| 国产精品人妻久久久久久| 黄色视频,在线免费观看| 国产av在哪里看| 国产精品99久久久久久久久| 在线播放国产精品三级| 少妇猛男粗大的猛烈进出视频 | 亚洲欧美精品综合久久99| 日本欧美国产在线视频| 乱人视频在线观看| 成人毛片a级毛片在线播放| 午夜日韩欧美国产| 亚洲图色成人| 夜夜夜夜夜久久久久| a级一级毛片免费在线观看| 中文字幕av成人在线电影| 国产精品一区二区三区四区久久| 亚洲国产色片| 亚洲欧美成人精品一区二区| 亚洲精品日韩av片在线观看| 国产精品久久久久久av不卡| 国产aⅴ精品一区二区三区波| av福利片在线观看| 一区福利在线观看| 精华霜和精华液先用哪个| 国产免费一级a男人的天堂| 国产精品久久久久久精品电影| 天天躁夜夜躁狠狠久久av| 国产伦一二天堂av在线观看| 日韩在线高清观看一区二区三区| 真人做人爱边吃奶动态| 真人做人爱边吃奶动态| 中文字幕免费在线视频6| 中文字幕精品亚洲无线码一区| 男女视频在线观看网站免费| 九色成人免费人妻av| 亚洲熟妇中文字幕五十中出| 色在线成人网| 老女人水多毛片| 国产淫片久久久久久久久| 国产欧美日韩一区二区精品| 免费搜索国产男女视频| 精品久久久久久久久av| 免费电影在线观看免费观看| 国产私拍福利视频在线观看| 国产精品免费一区二区三区在线| 亚洲av美国av| 1000部很黄的大片| 成人一区二区视频在线观看| 精品久久久久久久久久久久久| 欧洲精品卡2卡3卡4卡5卡区| 亚洲美女搞黄在线观看 | 亚洲一区二区三区色噜噜| 亚洲第一区二区三区不卡| 亚洲欧美精品自产自拍| 成人毛片a级毛片在线播放| 亚洲精品日韩av片在线观看| 亚洲精品粉嫩美女一区| 狠狠狠狠99中文字幕| 久久午夜亚洲精品久久| 亚洲人成网站在线播| 麻豆一二三区av精品| 亚洲专区国产一区二区| 天天躁夜夜躁狠狠久久av| 91麻豆精品激情在线观看国产| 国产成年人精品一区二区| 天天躁日日操中文字幕| 欧美3d第一页| 久久久久精品国产欧美久久久| 欧美+日韩+精品| av.在线天堂| 国产私拍福利视频在线观看| 精品一区二区三区av网在线观看| 国产av一区在线观看免费| 尾随美女入室| 99久久久亚洲精品蜜臀av| 国内揄拍国产精品人妻在线| 97超视频在线观看视频| 亚洲精品亚洲一区二区| 国产不卡一卡二| 色哟哟·www| 免费观看的影片在线观看| 日本三级黄在线观看| 色综合色国产| 亚洲成人中文字幕在线播放| 欧美性猛交黑人性爽| 给我免费播放毛片高清在线观看| 欧美极品一区二区三区四区| 亚洲最大成人中文| 两个人的视频大全免费| 亚洲18禁久久av| 亚洲国产色片| 看片在线看免费视频| 三级国产精品欧美在线观看| 亚洲美女黄片视频| 综合色丁香网| 午夜激情欧美在线| 国内精品久久久久精免费| 看十八女毛片水多多多| 亚洲美女黄片视频| 国产国拍精品亚洲av在线观看| 免费观看人在逋| 久久人人爽人人片av| 日韩欧美在线乱码| 国产av在哪里看| 少妇裸体淫交视频免费看高清| 国产亚洲欧美98| 联通29元200g的流量卡| 三级男女做爰猛烈吃奶摸视频| 3wmmmm亚洲av在线观看| 免费电影在线观看免费观看| 国产精品日韩av在线免费观看| 高清毛片免费观看视频网站| 午夜激情福利司机影院| 尤物成人国产欧美一区二区三区| 国产精品野战在线观看| 精品一区二区三区av网在线观看| 久久久久久久久久成人| 波多野结衣高清作品| a级毛色黄片| 成年版毛片免费区| 91久久精品国产一区二区三区| 午夜精品国产一区二区电影 | 日本成人三级电影网站| 听说在线观看完整版免费高清| 亚洲欧美精品综合久久99| 在线天堂最新版资源| 91久久精品国产一区二区三区| 最近在线观看免费完整版| 欧美色欧美亚洲另类二区| 波多野结衣高清作品| 最近中文字幕高清免费大全6| 国产精品嫩草影院av在线观看| 国产单亲对白刺激| 国产aⅴ精品一区二区三区波| 人妻丰满熟妇av一区二区三区| av福利片在线观看| 一a级毛片在线观看| 亚洲欧美清纯卡通| 熟妇人妻久久中文字幕3abv| 五月玫瑰六月丁香| 日本在线视频免费播放| 村上凉子中文字幕在线| 99久国产av精品国产电影| 内射极品少妇av片p| 国产精品人妻久久久影院| 免费看av在线观看网站| 午夜福利高清视频| 春色校园在线视频观看| 人妻制服诱惑在线中文字幕| 国产精品日韩av在线免费观看| 精品久久久久久成人av| 午夜福利在线在线| 亚洲欧美清纯卡通| 久久久久国产网址| 中文字幕熟女人妻在线| 国产色婷婷99| 婷婷六月久久综合丁香| 99久国产av精品| 99久久久亚洲精品蜜臀av| 两个人视频免费观看高清| 成人欧美大片| 高清日韩中文字幕在线| 国产精品久久电影中文字幕| 99国产极品粉嫩在线观看| 18禁裸乳无遮挡免费网站照片| 亚洲性久久影院| 国产精品久久久久久精品电影| 国产日本99.免费观看| 亚洲中文字幕日韩| 欧美在线一区亚洲| 国产午夜精品论理片| 天堂动漫精品| 成年av动漫网址| 天天躁夜夜躁狠狠久久av| 亚洲高清免费不卡视频| 久久鲁丝午夜福利片| 女生性感内裤真人,穿戴方法视频| 别揉我奶头 嗯啊视频| 日韩欧美 国产精品| 卡戴珊不雅视频在线播放| 免费一级毛片在线播放高清视频| 亚洲一级一片aⅴ在线观看| 色视频www国产| 日本黄色视频三级网站网址| 一区二区三区免费毛片| 成年版毛片免费区| 成人午夜高清在线视频| 日韩成人伦理影院| 有码 亚洲区| 天堂影院成人在线观看| 男人舔女人下体高潮全视频| 麻豆精品久久久久久蜜桃| 亚洲国产高清在线一区二区三| 一区二区三区四区激情视频 | 亚洲最大成人中文| 婷婷色综合大香蕉| 一本久久中文字幕| 女人被狂操c到高潮| 欧美xxxx黑人xx丫x性爽| 美女 人体艺术 gogo| 亚洲av中文字字幕乱码综合| 91久久精品电影网| 熟妇人妻久久中文字幕3abv| 女同久久另类99精品国产91| 搡女人真爽免费视频火全软件 | 永久网站在线| 国内少妇人妻偷人精品xxx网站| 亚洲内射少妇av| 久久久色成人| 亚洲中文日韩欧美视频| 少妇人妻一区二区三区视频| 搡老妇女老女人老熟妇| 中文字幕免费在线视频6| 两个人的视频大全免费| 99久久中文字幕三级久久日本| 91在线精品国自产拍蜜月| 一级黄片播放器| 亚洲中文日韩欧美视频| 蜜臀久久99精品久久宅男| 俄罗斯特黄特色一大片| 亚洲国产精品久久男人天堂| 美女免费视频网站| 国产69精品久久久久777片| 在线观看av片永久免费下载| 精品欧美国产一区二区三| 国产毛片a区久久久久| 18禁在线播放成人免费| 午夜免费男女啪啪视频观看 | 午夜日韩欧美国产| 亚洲自偷自拍三级| 最近最新中文字幕大全电影3| 久久精品久久久久久噜噜老黄 | 久99久视频精品免费| 欧美日韩一区二区视频在线观看视频在线 | 亚洲国产精品国产精品| 亚洲欧美清纯卡通| 在线播放国产精品三级| 小说图片视频综合网站| 国产精品人妻久久久影院| 亚洲七黄色美女视频| 国产美女午夜福利| 国产精品不卡视频一区二区| 人妻夜夜爽99麻豆av| 精品乱码久久久久久99久播| 久久久色成人| 精品久久久噜噜| 人妻丰满熟妇av一区二区三区| 寂寞人妻少妇视频99o| 精品免费久久久久久久清纯| 午夜久久久久精精品| 婷婷精品国产亚洲av在线| 国产成人影院久久av| 国产免费男女视频| 在线免费观看的www视频| 插阴视频在线观看视频| 赤兔流量卡办理| 无遮挡黄片免费观看| 久久久久国内视频| 日日摸夜夜添夜夜添av毛片| 亚洲自偷自拍三级| 一卡2卡三卡四卡精品乱码亚洲| 国产精品一区二区性色av| 国产高清有码在线观看视频| 亚洲av中文字字幕乱码综合| 高清日韩中文字幕在线| 97超级碰碰碰精品色视频在线观看| 日韩成人伦理影院| 亚洲av成人av| 色av中文字幕| 男女啪啪激烈高潮av片| 春色校园在线视频观看| 日韩精品中文字幕看吧| 久久久久久大精品| 免费观看人在逋| 亚洲欧美日韩高清专用| 国产精品亚洲美女久久久| 一级a爱片免费观看的视频| 男人狂女人下面高潮的视频| 91在线观看av| 日本色播在线视频| а√天堂www在线а√下载| 中文在线观看免费www的网站| ponron亚洲| 天美传媒精品一区二区| 欧美3d第一页| 一个人看视频在线观看www免费| 人人妻人人澡欧美一区二区| 狠狠狠狠99中文字幕| 精品人妻一区二区三区麻豆 | 一级黄色大片毛片| av黄色大香蕉| 深夜a级毛片| 毛片女人毛片| 日韩成人伦理影院| 成人国产麻豆网| 美女高潮的动态| 99视频精品全部免费 在线| 综合色av麻豆| 青春草视频在线免费观看| 久久久午夜欧美精品| 91在线精品国自产拍蜜月| 国产91av在线免费观看| 2021天堂中文幕一二区在线观| 一个人观看的视频www高清免费观看| 人妻丰满熟妇av一区二区三区| 五月伊人婷婷丁香| 国产欧美日韩一区二区精品| 亚洲精品一卡2卡三卡4卡5卡| 黄色视频,在线免费观看| 99精品在免费线老司机午夜| 国产一区二区在线观看日韩| 亚洲最大成人手机在线| 国内精品久久久久精免费| 午夜影院日韩av| 久久国内精品自在自线图片| 国产探花极品一区二区| 久久久久久伊人网av| 久久久久性生活片| 日韩在线高清观看一区二区三区| 变态另类丝袜制服| 99热全是精品| 亚洲,欧美,日韩| 亚洲一级一片aⅴ在线观看| 99热全是精品| 国产成人精品久久久久久| 日本欧美国产在线视频| 又粗又爽又猛毛片免费看| 久久精品国产亚洲网站| 村上凉子中文字幕在线| 成年av动漫网址| 国产精品久久久久久精品电影| 乱码一卡2卡4卡精品| 精品午夜福利在线看| 久久久精品大字幕| a级毛片a级免费在线| 18+在线观看网站| 日韩强制内射视频| 亚洲欧美日韩东京热| 国产高清有码在线观看视频| 亚洲成人av在线免费| 禁无遮挡网站| 久久精品人妻少妇| 麻豆av噜噜一区二区三区| 亚洲自拍偷在线| 色av中文字幕| 国产精品国产三级国产av玫瑰| 啦啦啦观看免费观看视频高清| 欧美日韩一区二区视频在线观看视频在线 | 一区二区三区高清视频在线| 在线免费十八禁| 波多野结衣高清无吗| 日本撒尿小便嘘嘘汇集6| 最好的美女福利视频网| 国产v大片淫在线免费观看| 搡女人真爽免费视频火全软件 | 日韩精品有码人妻一区| 国产在视频线在精品| 天堂动漫精品| 国产亚洲精品av在线| 国产高清有码在线观看视频| 在线天堂最新版资源| 国产欧美日韩一区二区精品| 亚洲精品一区av在线观看| 精品日产1卡2卡| 99热网站在线观看| 国产欧美日韩精品一区二区| 观看美女的网站| 亚洲av二区三区四区| 国产亚洲av嫩草精品影院| 亚洲一区二区三区色噜噜| 午夜激情福利司机影院| 亚洲精品日韩av片在线观看| 免费黄网站久久成人精品| 国产精品国产三级国产av玫瑰| 欧美潮喷喷水| 欧美日韩乱码在线| 女同久久另类99精品国产91| 可以在线观看的亚洲视频| 亚洲真实伦在线观看| 老司机午夜福利在线观看视频| 日韩制服骚丝袜av| 精品久久久久久久久亚洲| av天堂中文字幕网| 18禁裸乳无遮挡免费网站照片| 亚洲一区二区三区色噜噜| 热99re8久久精品国产| 国产男靠女视频免费网站| 午夜福利成人在线免费观看| 美女黄网站色视频| 中文字幕av在线有码专区| 欧美日韩乱码在线| 三级男女做爰猛烈吃奶摸视频| 一本一本综合久久| 国产三级在线视频| 蜜桃久久精品国产亚洲av| 免费不卡的大黄色大毛片视频在线观看 | 午夜视频国产福利| 亚洲av第一区精品v没综合| 亚洲成人av在线免费| 麻豆成人午夜福利视频| 国产一区二区三区av在线 | 99视频精品全部免费 在线| 亚洲欧美中文字幕日韩二区| 亚洲精品日韩av片在线观看| 久久久a久久爽久久v久久| 亚洲无线在线观看| 亚洲av五月六月丁香网| 久久午夜亚洲精品久久| 亚洲美女黄片视频| 老师上课跳d突然被开到最大视频| 亚洲av成人av| 一个人看视频在线观看www免费| 一边摸一边抽搐一进一小说| 国产高潮美女av| 成年免费大片在线观看| 久久婷婷人人爽人人干人人爱| 精品国内亚洲2022精品成人| 日本成人三级电影网站| 亚洲欧美中文字幕日韩二区| 亚洲成人中文字幕在线播放| 成人美女网站在线观看视频| 日本免费a在线| 中文亚洲av片在线观看爽| 国产淫片久久久久久久久| 自拍偷自拍亚洲精品老妇| 亚洲久久久久久中文字幕| 久久精品影院6| 久久精品国产亚洲av涩爱 | 久久精品国产亚洲av天美| 国产精品一区二区三区四区久久| 欧美人与善性xxx| 午夜福利在线观看吧| 中文字幕av成人在线电影| 亚洲av电影不卡..在线观看| 亚洲熟妇熟女久久| 亚洲无线观看免费| 久久6这里有精品| 身体一侧抽搐| 麻豆乱淫一区二区| 一区福利在线观看| 免费一级毛片在线播放高清视频| 日韩av在线大香蕉| 国产免费一级a男人的天堂| 国产单亲对白刺激| 国产精品一区二区三区四区久久| 欧美绝顶高潮抽搐喷水| 亚洲av免费高清在线观看| 国产午夜精品久久久久久一区二区三区 | 亚洲人成网站高清观看| 我要搜黄色片| 不卡一级毛片| 特大巨黑吊av在线直播| 亚洲熟妇熟女久久| 伦理电影大哥的女人| 老司机影院成人| 精华霜和精华液先用哪个| 久久久精品大字幕| а√天堂www在线а√下载| 精品一区二区三区人妻视频| 欧美中文日本在线观看视频| 波多野结衣高清作品| 免费看光身美女| 国产欧美日韩一区二区精品| 看非洲黑人一级黄片| 国产精品电影一区二区三区| 欧美在线一区亚洲| 又粗又爽又猛毛片免费看| 亚洲图色成人| 狂野欧美激情性xxxx在线观看| 精品福利观看| 久久精品夜色国产| 久久久久久久久大av| 亚洲国产日韩欧美精品在线观看| 久久天躁狠狠躁夜夜2o2o| 午夜福利高清视频| 国产精品永久免费网站| 嫩草影院精品99| 国产精品久久久久久久电影| 性欧美人与动物交配| 97超视频在线观看视频| 两性午夜刺激爽爽歪歪视频在线观看| 精品一区二区三区视频在线| 久久精品夜夜夜夜夜久久蜜豆| 看片在线看免费视频| 久久天躁狠狠躁夜夜2o2o| 日本五十路高清| 男女那种视频在线观看| 我要搜黄色片| 女同久久另类99精品国产91| av女优亚洲男人天堂| 国产爱豆传媒在线观看| 在线看三级毛片| 国产 一区精品| 免费黄网站久久成人精品| 亚洲av美国av| 国产精品99久久久久久久久| 给我免费播放毛片高清在线观看| 精品久久久久久久久亚洲| 国产伦精品一区二区三区四那| 卡戴珊不雅视频在线播放| 亚洲av中文av极速乱| 最近最新中文字幕大全电影3| 一个人免费在线观看电影| 寂寞人妻少妇视频99o| 欧美极品一区二区三区四区| 91久久精品国产一区二区三区| 亚洲中文字幕日韩| 99久国产av精品| 国产精品日韩av在线免费观看| 亚洲成人av在线免费| 亚洲一区高清亚洲精品| 亚洲av免费在线观看| 亚洲欧美日韩东京热| 亚洲人与动物交配视频| 国产精品,欧美在线| 欧美日韩国产亚洲二区| 国产片特级美女逼逼视频| 搡女人真爽免费视频火全软件 | 国产av不卡久久| 欧美最新免费一区二区三区| 欧美3d第一页| 精品久久久久久久久av| 在线观看免费视频日本深夜| 老司机午夜福利在线观看视频| 国产精品日韩av在线免费观看| 国产精品亚洲一级av第二区| 国产精品久久久久久久久免| 精品久久久久久久久av| 黄色日韩在线| 欧美色视频一区免费| 人人妻,人人澡人人爽秒播| 干丝袜人妻中文字幕| 黄色配什么色好看| 色哟哟·www| 91狼人影院| 欧美在线一区亚洲| 国产亚洲精品久久久久久毛片| 男人和女人高潮做爰伦理| 久久精品人妻少妇| 国产精品伦人一区二区| 亚洲综合色惰| 99热这里只有是精品50|