CN103528586A

CN103528586A - Sailing planning algorithm design based on grid failure

Info

Publication number: CN103528586A
Application number: CN201310478422.5A
Authority: CN
Inventors: 向永红; 于洋; 姜梁; 郭茜; 张国勇; 吴国强; 王静; 尹中义
Original assignee: China Academy of Aerospace Electronics Technology Co Ltd
Current assignee: China Aerospace Times Electronics Corp; China Academy of Aerospace Electronics Technology Co Ltd
Priority date: 2013-10-31
Filing date: 2013-10-31
Publication date: 2014-01-22
Anticipated expiration: 2033-10-31
Also published as: CN103528586B

Abstract

The invention provides a grid division sailing planning method based on an environment, and the method is used for unmanned aircraft sailing planning. According to the method, the flying coverage area is divided into two-dimensional grids and/or three-dimensional grids, topographical obstacles and various threats are modeled into failure nodes in grids, the planning method comprises the steps of initialization, environment modeling and node labeling, recognition of minimum communication failure blocks and construction of boundaries thereof, and computation of shortest route. The complexity of the sailing planning method and the number of grid nodes in the design are in linear relationship, and the complexity is low.

Description

Path Planning design based on fault grid

Technical field

The present invention relates to the path planning method of a kind of unmanned flight system (Unmanned Aerial System, UAS), particularly, based on flight range being carried out to the path planning method of two and three dimensions grid division, belong to unmanned aerial vehicle flight path planning technology field.

Technical background

The task of trajectory planning system is to produce optimal path according to various factorss such as aerial mission, threat, weather, air position and terrain datas, generally need to meet the index of two aspects: the one, and in planning process, require all kinds of threats as much as possible little; The 2nd, requirement path is short as much as possible.

Trajectory planning mainly solves the problem of following two aspects:

(a) battlefield surroundings modeling.Battlefield surroundings is comprised of two parts: the one, and physical environment model.Mainly contain the model of landform, various weathers, they are different for the impact of routeing.The 2nd, the model of full spectrum of threats, mainly the anti aircraft fire from enemy threatens, threat radar.

(b) shortest path requirement.Requiring optimum results is the shortest air route under full spectrum of threats environmental constraints.Often need to adopt optimized algorithm to calculate shortest path.Using the various factors that provided after the environmental modeling of (a) as constraint condition, more additional correlated conditions are as aircraft maneuverability (climbing ability and limiting altitude, minimal curve radius etc.), and the parameters such as aircraft oil mass are carried out optimization computation.

More representational in current Path Planning have: Artificial Potential Field Method, A* algorithm, dijkstra's algorithm, skeleton diagram method, dynamic programming, genetic algorithm, ant group algorithm, particle cluster algorithm etc.Wherein particle cluster algorithm, due to its less parameter, preferably convergence, stronger features such as robustness, is applied fully in many optimization problems, is generally subject to scholars' attention recently.In addition, A* algorithm is very effective for static trajectory planning, and its mutation D* and R* algorithm can adapt to dynamic route planning better.

Have in the prior art whole planning space is divided into territory, a plurality of rectangular sub-regions, the cell that has simultaneously comprised barrier is set to infeasible unit.By introducing distance metric and carrying out the grid division that range conversion realizes V figure.The value of each cell is this cell to the distance of nearest barrier, and range formula is provided by formula (1), wherein by connected side by side lattice spacing from being defined as 10, the lattice spacing that diagonal angle joins is from being defined as 14 (Fig. 1).

u(i,j)=Min(u(i–1,j–1)+b,u(i–1,j)+a,u(i–1,j+1)+b,u(i,j–1)+a,u(i,j+1)+a,u(i+1,j–1)+b,u(i+1,j)+a,u(i+1,j+1)+b) (1)

Wherein, a=10, b=14 is grid distance parameter, u (i, j) represents that the grid of the capable j row of i is to the distance of barrier.

By connecting the cell apart from each obstacle distance maximum, can obtain grid V figure.Fig. 2 (a) has provided the range image when having A, two barriers of B; Fig. 2 (b) has provided by border and has followed the tracks of and the grid V figure (black grid) that obtains.Using this grid V figure as initial path, can use various optimized algorithms to select suitable route segment, carry out obstacle-avoiding route planning.In fact this way can think cell is corresponded to a node in grid, and each node number of degrees is 8, and 8 nodes that each node is adjacent are connected, but the line on diagonal line is endowed larger distance value (seeing Fig. 2 (c)).

Planning space is divided into the method that discrete two-dimensional grid adopts various optimized algorithms to carry out optimal path computation later more all can be seen in many papers.Seldom having work is situation based on three-dimensional, and common way is through two dimension, planning obtains guaranteeing to meet after the flight path of unmanned plane horizontal maneuver constraint, carry out height and plan generation Three-dimensional Track.This method does not take into full account and utilizes three-dimensional information, only on the two-dimentional flight path of having planned, consider elevation, the optimality of greatly having limited to flight path, especially, this method has only used threat and the information of landform when a certain height to carry out design level flight path, has ignored on differing heights and to have threatened and the variation of landform.The optimized algorithm obtaining thus can only be the optimized algorithm based on a certain height.And directly based on three-dimensional trajectory planning, taken into full account the impact of the factors such as landform, threat, can carry out global optimization.But there is following difficulty in Three-dimensional Track planning: the first, and computation complexity is too large, is difficult to accomplish real-time trajectory planning and implements flight path weight-normality and draw; The second, common algorithm is easy to be absorbed in locally optimal solution.The present invention proposes environment to be divided into 3D grid, by evading obstacle, calculates fast the shortest route from starting point to destination, and the method proposing can improve the real-time of trajectory planning, avoids being absorbed in local optimum simultaneously.

Summary of the invention

The object of the invention is to plan that for solving unmanned aerial vehicle flight path computation complexity is too large, the problem that trajectory planning difficulty waits and is easily absorbed in Local Search in real time, has proposed the path planning method that a kind of grid based on environment is divided.

The object of trajectory planning is exactly to generate possible path by avoiding obstacle.First the present invention is divided into two-dimensional grid by flight overlay area, and by terrain obstruction, full spectrum of threats is modeled as the malfunctioning node in grid.As shown in Figure 3, in figure, gray area is barrier zone, and the region of having filled oblique line is the region being taken by obstacle.From origin-to-destination, set up a path connecting with straight-line segment.Further, flight overlay area is divided into 3D grid, landform and full spectrum of threats are modeled as various malfunctioning nodes, with Consideration of Three-dimensional trajectory planning problem.

If divided grid is corresponded to network topology grid (Mesh), barrier or topographic relief are corresponded to the malfunctioning node in Mesh, problem is converted into the routing algorithm based on fault grid.In there is the Mesh of fault, look for a shortest path from starting point s to terminal d.

Whole trajectory planning mainly comprises initialization, environmental modeling (two dimension or 3D grid generate) and node are marked and drawed, minimum connected trouble block (MCC) identification and its boundary formation, shortest route is calculated, then according to whether having new threat information and whether need refinement to do further planning (referring to Fig. 4).

Accompanying drawing explanation

Fig. 1 is grid distance metric schematic diagram in prior art;

Fig. 2 generates V figure according to line of equidistance in prior art: wherein Fig. 2 (a) has provided the range image when having A, two barriers of B; Fig. 2 (b) has provided by following the tracks of and the grid V that obtains schemes on border; Fig. 2 (c) is obstacle-avoiding route planning schematic diagram.

Fig. 3 is cell split plot design schematic diagram of the present invention;

Fig. 4 is the trajectory planning basic step schematic diagram in the present invention;

Fig. 5 is node mark and the MCC boundary formation in the present invention;

Fig. 6 is the simulation experiment result data of algorithm of the present invention.

Embodiment:

Following part elaborates to the specific embodiment of the present invention in connection with Figure of description:

Step 1. initialization

Be written into map and altitude figures, determine the body of a map or chart that flight range is corresponding and divide length and width and high unit length, thereby determining the three-dimensional size n that will set up ₁, n ₂and n ₃.Be written into various radars, artillery fire threat data.Given starting point and destination information.

Step 2. environmental modeling

The three dimensions of UAV flight environment of vehicle planning be expressed as geometric space region (x, y, z) | 0≤x≤n ₁, 0≤y≤n ₂, 0≤z≤n ₃.When carrying out two-dimentional modeling, by UAV flight environment of vehicle horizontal division, be two-dimentional n ₁* n ₂(Mesh is denoted as M to grid ₂).By n ₃individual M ₂according to unit distance, be stacked up, and connect a three-dimensional n of corresponding node formation ₁* n ₂* n ₃grid (is denoted as M ₃).

Provide below below by some basic terms of using.M ₂there is n ₁* n ₂individual node, inner each node number of degrees are 4, the given address (x of each node u _u, y _u), 0≤x wherein _u≤ n ₁, 0≤y _u≤ n ₂.Node (x _u+1, y _u) be called node u+X neighbours, correspondingly, node (x _u, y _u+1), (x _u-1, y _u), (x _u, y _u-1) be called node u+Y ， – X ， – Y neighbours.Manhattan (Manhattan) distance between any 2 u and v, represents with M (u, v), is defined as their geometric distance | x _u– x _v|+| y _u– y _v|.Suppose node s (x _s, y _s) be source node, u is present node, d (x _d, y _d) be destination node.Without loss of generality, suppose x _s=y _s=0, x _d, y _d>=0.Owing to there being malfunctioning node, may not have length is the path of M (s, d).Make D (s, d) for the shortest path length between s and d, and D (s, d) >=M (s, d).General using [x:x ', y:y '] represent that four nodes are respectively (x, y), (x, y '), (x ', y ') and (x ', rectangular area y).Especially, [x:x, y:y ']/[x:x ', y:y] represent that respectively one along the line segment of Y/X direction.3D grid M ₃there is n ₁* n ₂* n ₃individual node, the given address (x of each node u _u, y _u, z _u), similarly, whether two nodes of definable adjacent and node+X, and+Y ,+Z ， – X ， – Y ， – Z neighbours.

In two dimension (three-dimensional) grid, once useless node definition arrives after this node for path, next step must Yan – X/ – Y(– X/ – Y/ – Z) direction moves, and it is not the shortest making path; When a non-malfunctioning node+X and+Y(+X ,+Y and+Z) neighbor node of direction is while being malfunctioning node, is referred to as can ' t-reach node.

Given first two dimension path planning method, for two-dimentional Mesh M ₂situation provide the method that obtains shortest path.

Step 3. two-dimensional grid M ₂initial value given and mark

Once node in-scope belongs to that landform is blocked or the threatening area such as artillery fire, radar, by Node configuration for there being fault (unavailable).At first, all malfunctioning nodes are all noted as faulty, and all non-malfunctioning nodes are noted as safe.

If for a safe node+X neighbours and+Y neighbours be all faulty and useless the two one of, this node is just noted as useless so; As Qi – X neighbours of Guo He – Y neighbours be all noted as faulty and can ' t-reach the two one of, marking this node is can ' t-reach.So iteration mark is not until there is no new useless and the appearance of can ' t-reach node.All faulty, useless and can ' t-reach node are all called unsafe node, and other nodes are called safe node.Suppose that whole network is communicated with, and source node and destination node are the nodes that is labeled as safe.Supposing has a limit to be connected between two adjacent unsafe nodes, claims that the block that contains malfunctioning node obtaining is in this way minimum connected trouble block.All MCC are mutually disjoint; Hamming between any two MCC distance is at least 2, always exist one can be from the path of passing between two MCC.By using local algorithm, construct as seen the complexity of MCC and the diameter of maximum fault block is proportional.It is the edge node of corresponding MCC block that weighing-appliance has the safe node of a unsafe neighbor node.A safe node, if the neighbours of its – X and – Y-direction are the edge node of same MCC, is referred to as corner node (opposite corner); And initial angle node (initialization corner) is defined as a safe node, it+X and+neighbours in Y-direction are the edge nodes of same MCC.

Step 3.1. is initial, and marking all malfunctioning nodes is faulty, and all non-malfunctioning nodes are safe.

If step 3.2. node u is designated as safe, but its+X neighbours and+Y neighbours are faulty or useless, are designated as useless.

If step 3.3. node u is safe, but the neighbours of Qi – directions X are with the neighbours of – Y-direction are faulty or can ' t-reach, and mark u is can ' t-reach.

Step 3.4. recursively marks node until there is no new useless or the appearance of can ' t-reach node.

The faulty that step 3.5. is all, useless and can ' t-reach node are also referred to as unsafe node.

Step 4.MCC shape is determined

Determining of MCC shape be determining border, all fault zones.Unmanned aerial vehicle flight path can not pass through these regions, otherwise there will be in practical flight, hits mountain or enters the situations such as hazardous location.The present invention adopts the shape of determining MCC by the method that local boundary information is broadcasted in limited area.

Article two, identification information (being initialized as sky), carry separately subregion information, from initial angle node, by clockwise and counter clockwise direction, along the limit of MCC, broadcast away until arrive corner node, by collect they the positional information of node of process, the shape of MCC can be determined at diagonal angle Nodes.This broadcast is proceeded, until shape information F (c) is brought back to initial angle node.Region below F (c) is called forbidden zone (forbidden region), is designated as R _y(c); And the above region of next-door neighbour F (c) is called critical zone (critical region), be designated as R ' _y(c).For guiding routing procedure ， – X boundary information will carry F (c), R _yand R ' (c) _y(c) along x=x _cdirection is broadcasted until arrival net boundary, if met with another one MCC F (v) in way, is turned right and moved on along the border of F (v), from angle point v, and R _y(v) will and R _y(c) merge (R _y(c)=R _y(c) ∪ R _y(v)).， – Y boundary information will carry F (c), R similarly _xand R ' (c) _x(c) along y=y _cdirection broadcast is until arrive net boundary, left-hand rotation when being necessary.Fig. 5 is a diagram of said process, and its Oxford gray region is forbidden zone, and light gray areas is critical section.

From (0,0) to d (x _d, y _d) (x _d, y _d>=0) MCC mark process can be described as following steps: MCC-F (c) boundary formation

Step 4.1. is from initial angle node c, article two, identification information is (one by clockwise, another is by counterclockwise) along the limit of MCC, broadcast away until arrive corner node c ', the positional information of the corner node of all centres processes is all recorded, thereby locates to obtain shape F (c) at c '.Then can identify critical zone and forbidden zone (R _x(c), R _y(c), R ' _x(c), R ' _y(c)).

Step 4.2., after the triplet information of corner node c ' (and hereinafter, tlv triple all refers to the shape information that comprises a MCC, corresponding forbidden zone information, and corresponding critical zone information) is determined, will broadcast as (F (c), R herein _y+X(c), R ' _y(c)), from node c, tlv triple (F (c), R _x(c), R ' _x(c)) along straight line y=y _cbroadcast.If met as F (v) with another one MCC in way, along the border of F (v), turn left, after this, by R _x(v) merge to R _y(c).

Step 4.3. is from node c ', the broadcast of both direction by continuation until identification information returns to initial angle node c.

Step 4.4. is from node c, tlv triple (F (c), R _y(c), R ' _y(c)) along straight line x=x _cbroadcast.If met as F (v) with another one MCC in way, along the border of F (v), turn right, after this, by R _y(v) merge to R _x(c), Gou Jian – X border, forms tlv triple (F (c), R at each boundary node place _{y – X}(c), R ' _y(c)).

Step 4.5. simultaneously, will broadcast until arrive another one border along X dimension and then edge+Y dimension direction by the tlv triple of receiving at each boundary node place.To each node, can receive the tlv triple on two borders, obtain (F (c), R _y(c), R ' _y(c)).

Step 5. is calculated shortest route

Calculate the shortest route from starting point s to destination d.Find out a shortest safe flight path (avoiding all obstacles and threat) from starting point to destination.

If step 5.1. present node u=d, stops.

If step 5.2. u is not plugged, applies following routing decision and send forward routing iinformation:

If step 5.2.1. is x _d>0 and y _d>0, and+X/+Y neighbours do not have fault, and general+X/+Y direction is added into optional working direction collection P.

Step 5.2.2., to each tlv triple finding at a u place, if walk to cause route to enter forbidden zone R (c) from certain direction, removes this direction from P.

Step 5.2.3. application Disha fully adaptive route (any fully adaptive route all can) is selected working direction from set P.

Step 5.2.4. sends routing iinformation along selected direction.

Otherwise, the application method that detours below.

Step 5.3. is built from the border of the MCC of step 4, and the information in the MCC block sequence on known all obstructions Manhattan road, node s place, therefrom finds a sequence (F who approaches d most ₁, F ₂..., F _n).

Step 5.4. is used a shortest D (s, d) in formula (2) calculating path: (a) by node c ₁, be designated as P ₀, (b) by node c ' _n, be designated as P _n, and (c) through two continuous MCC (F _iand F _i+1, 1≤i≤n), be designated as P _i.

If step 5.5. is P0/ Pnbe chosen as shortest path, applying step 5.5.1～5.5.4 arrives intermediate objective d '=c1/ c 'n.Otherwise, for the shortest path P choosingi(1≤i<n), applying step 5.5.1～5.5.4 forms manhatton distance path M (s, c 'i) and M (c 'i, ci+1) arrive intermediate objective node d '=ci+1.

If step 5.5.1. is x _d>0/y _d>0 and+X/+Y neighbours are not malfunctioning nodes, increase+X/+Y direction is in optional working direction set P.

Step 5.5.2., for each tlv triple finding at present node u (comprising: F (c), R (c), R ' (c)), removes a direction from P, if when d ∈ R ' is (c) time, will make route enter R (c) from this direction.

Step 5.5.3. applies any fully adaptive route and select a working direction from set P.

Step 5.5.4. by routing iinformation along selected direction transmission.

Wherein

Above step is based on two-dimensional grid, to find the step of shortest path.

For Three-dimensional Track planning, the shortest-path method of design based on 3D grid.Owing to threatening and obstacle is mainly derived from ground, this means that fault appears at the bottom of network in the 3D grid of division.Therefore, any mid-side node does not have useless node in+Z direction.

Step 2 '. 3D grid initial value given and mark

In grid is divided, the level cross-sectionn of fault reduces along with the increase of height conventionally.So, if fruit dot u is safe, its+Z neighbours can not be just faulty or can ' t-reach.Therefore there is not useless node.

Step 2 ' .1. is initial, and all malfunctioning nodes are labeled as faulty, and all non-malfunctioning nodes are labeled as safe.

If step 2 ' .2. node u is safe, but its-X neighbours ,-Y neighbours, and-Z neighbours are faulty or can ' t-reach, mark u is can ' t-reach.

Step 2 ' .3. recursively marks node until do not have new can ' t-reach node to occur.

The faulty that step 2 ' .4. is all and can ' t-reach node are all called unsafe node.

Step 3 '. MCC identification and boundary formation in 3D grid

3D grid can be divided into two-dimensional grid and carry out MCC identification and boundary formation again.Due to threaten and landform all have lower go up greatly little, and the situation that can not occur depression in Z direction, so can be directly from the boundary node of Z=h layer, whether be boundary node and collect frontier point information if starting to identify this node layer downwards, find this MCC after all boundary nodes of this one deck, boundary information to be broadcasted along border.Owing to there will not be different MCC in Z direction, there is no the forbidden zone in Z direction yet, therefore always there is manhattan path.But when the flying height of unmanned plane is had to requirement, particularly requiring terrain following flight or aircraft flight height-limited when processed, unmanned plane can not unconfinedly soar, can be according to the restriction of actual conditions assigned altitute, and the space beyond maximum height limit will become forbidden zone.

If step 3 ' .1. is at Z=h place, there is a new malfunctioning node u (u in XY cross section _x, u _y, h), define a new MCC, determine point (u _x, u _yh+1) be boundary node, simultaneously centered by a u, adopt depth-first search strategy, in XY plane, successively to surrounding, whether be boundary node and collect frontier point information if expanding the neighbor node of judging point u, find this MCC after all boundary nodes of this one deck, boundary information is broadcasted along boundary node.The border having existed for the XY cross section at Z=h+1, in the XY cross section of Z=h along this border (from center, fault zone) successively outwards search definite border.

If two threatening areas of step 3 ' .2. (as antiaircraft gun threatening area and threat radar region) have lap, two MCC can start to intersect at some At The Heights, at this moment two MCC on this height are merged into a MCC at this XY section.

Step 3 ' .3. is according to actual conditions, artificially a given limitation in height region.As when flying along landform, can provide is overhead forbidden zone with threatening region more than certain altitude.

Step 3 ' .4. is the end of scan successively, and given after limitation in height region, boundary information is broadcasted at XZ and two planes of YZ, determine forbidden zone and critical zone, and definite border is (as in XY plane simultaneously, border determines that method is: at the M of mid-side node u (u), Q (u) and Q ' (u) all determined after, as along (+Y, – X)-Bian, information M (u), Q _yand Q ' (u) _y(u) being Aned in (+Y ， – the X)-border along it to – Y-direction propagates; In YZ plane, can determine similarly border).If run into another one MCC as M, it is by combining with the outstanding boundary member of M, simultaneously by forbidden zone Q _ymerge to Q _y(u) in.Thereby complete MCC identification and boundary formation.

Step 4 '. determine shortest route

Determine from s (0,0,0) to d (x _d, y _d, z _d) (x _d, y _d, z _d>=0) shortest route.Find the secure path of avoiding all threats and terrain obstruction.

Step 4 ' .1. feasibility checks: at source node s, along+X ,+Y-direction sends detect-message.When one as run into another MCC along the information of+directions X, it will add (– X ,+Y)-and (– X ,+Z)-border is as boundary formation.By inspection, whether these two detection information can arrive respectively surface [x to source node _d: x _d, 0:y _d, 0:z _d], [0:x _d, y _d: y _d, 0:z _d].If can not reach any one in these two surfaces, stop route, because there is not manhatton distance path.(illustrate: for unmanned plane path planning, in+Z direction, can not run into another MCC, therefore not right+Z direction checks.）

Step 4 ' .2. present node u routing decision and message transmit, and comprise source node s:

Step 4 ' .2.1. is increased to all optional directions in working direction set F, finds the MCC of all records.

Step 4 ' .2.2. is for each MCC finding in previous step, removes those and make impact point in critical zone and the neighbours of u will arrive the direction of forbidden zone along this direction from F.

Step 4 ' .2.3. application fully adaptive and Minimal routing are processed from set F and are selected a working direction.

Step 4 ' .2.4. sends to forward next node by routing iinformation along this selected direction.

Arrive this, provided the method for the shortest route between 2 based on two and three dimensions grid, correspondingly also just provided a method of the unmanned plane path planning of dividing based on unmanned plane during flying coverage 3D grid.Computing time, complexity and grid node number were linear.

Simulation and analysis: by Monte Carlo the Realization of Simulation, tested the MCC model routing algorithm based on 3D grid.By random arrangement malfunctioning node in 30 * 30 * 30 3D grid, set at random starting point s and impact point d.Experiment showed, that said method can find a shortest path from s to d as long as there is the manhattan path from s to d, thus method given here be optimum-can find shortest path; And, providing surpass 400 malfunctioning nodes in the situation that, can both Fast Convergent, even when providing while surpassing 500 malfunctioning nodes, still can be no more than 4 take turns calculate after convergence.The several ratios for total nodes that is designated as safe of node that participate in information broadcasting shared ratio when having 500 malfunctioning nodes is the highest, but does not surpass 50%.From each step, complexity and the grid node number of mark and identifying are linear.Therefore total computing time and node number is linear relationship, and experimental result has also confirmed this point (relevant experimental result data is referring to Fig. 6).

Beneficial effect: with the comparison of common unmanned aerial vehicle flight path planning algorithm, method advantage of the present invention is mainly reflected in:

(1) found from starting point to destination, avoided shortest path of all obstacles and threat, result is optimum.Method given here, except considering the influence of topography and full spectrum of threats factor, it is also conceivable that the factors such as aircraft aerodynamic performance, oil mass restriction, if these factors are converted to landform, threaten influence factor or flying height restriction etc. by certain mode.

(2) the path planning method complexity that the present invention proposes and the mesh node number of design are linear, and complexity is low.Can also first set up coarseness flight path (is smaller n ₁, n ₂and n ₃), at important area, do local mesh refinement, using node that original path enters this region as starting point, leave the node in this region as destination, adopting uses the same method cooks up a new flight path, and this refinement can successively advance.The flight path that can be optimized in this way, further reduces computation complexity, realizes real-time trajectory planning and flight path weight-normality is drawn.

(3) unmanned aerial vehicle flight path planning problem is mapped as to internet topology algorithmic issue, because the investigative technique of internet routing algorithm is relatively ripe, algorithm is abundant, by doing appropriate reconstruction and design about routing algorithm, can be more fast for unmanned plane be designed efficient path planning method.

(4) common unmanned plane path planning algorithm is not easy to realize parallel computation.Because each node in method proposed by the invention only need to know that part (neighbor node) information can be carried out border and domain of dependence is determined, can adopt easily distributed computing technology to realize and calculate acceleration.

Claims

1. the path planning method that grid based on environment is divided, for unmanned aerial vehicle flight path planning, it is characterized in that, described method is divided into two-dimensional grid and/or three-dimensional network by flight overlay area, terrain obstruction and full spectrum of threats are modeled as to the malfunctioning node in grid, and described planing method comprises the step that initialization, environmental modeling and node mark, minimum connected trouble block (MCC) identification and its boundary formation, shortest route are calculated.

2. method according to claim 1, it is characterized in that, described initialized step comprises: be written into map and altitude figures, determine the body of a map or chart that flight range is corresponding and divide length and width and high unit length, thereby determining the three-dimensional size that will set up n ₁, n ₂with n ₃; Be written into various radars, artillery fire threat data; Given starting point and destination information.

3. method according to claim 1, is characterized in that, described environmental modeling comprises two-dimentional modeling and/or three-dimensional modeling, when carrying out two-dimentional modeling, unmanned plane during flying ambient level is divided into two dimension n ₁' n ₂grid M ₂; During three-dimensional modeling by n ₃individual M ₂according to unit distance, be stacked up, and connect a three-dimensional of corresponding node formation n ₁' n ₂' n ₃grid M ₃.

4. method according to claim 3, is characterized in that, two-dimensional grid M ₂initial value is given specifically to be comprised the following steps with node mark:

1) marking all malfunctioning nodes is faulty, all non-malfunctioning nodes are safe;

2) if node ube designated as safebut, its+ xneighbours and+ yneighbours are faultyor useless, be designated as useless;

3) if node ufor safebut, Qi – xthe neighbours of direction are with – ythe neighbours of direction are faultyor can't-reach, mark ufor can't-reach;

4) recursively mark node until do not have new uselessor can't-reachnode occurs;

5) by all faulty, uselesswith can't-reachnode is called unsafenode.

5. method according to claim 4, is characterized in that, the MCC identification of two-dimensional grid and boundary formation are the shapes that the method by local boundary information is broadcasted in limited area is determined MCC, specifically comprise the following steps:

1) from initial angle node c, broadcast away until arrive corner node along the limit of MCC with counterclockwise two identification informations clockwise c', the positional information of the corner node of all middle processes is all recorded, thereby c' locate to obtain shape f( c), then can identify critical zone;

2) to corner node c'triplet information determine after, will broadcast as ( f( c), r _y+X( c), r' _y( c)), from node c, tlv triple ( f( c), r _x( c), r' _x( c)) along straight line y= y _cbroadcast; If way in another one MCC as f( v) meet, along f( v) border turn left, after this, will r _x( v) merge to r _y( c);

3) from node c', the broadcast of both direction by continuation until identification information returns to initial angle node c;

4) from node c, tlv triple ( f( c), r _y( c), r' _y( c)) along straight line x= x _cbroadcast, if way in another one MCC as f( v) meet, along f( v) border turn right, after this, will r _y( v) merge to r _x( c), Gou Jian – xborder, at each boundary node place, form tlv triple ( f( c), r _{y–
x}( c), r' _y( c));

5)meanwhile, the tlv triple of receiving at each boundary node place is by edge xdimension and then edge+ ythe broadcast of dimension direction is until arrive another one border; To each node, can receive the tlv triple on two borders, obtain ( f( c), r _y( c), r' _y( c)).

6. method according to claim 5, is characterized in that, the step that two-dimensional grid is found shortest path comprises:

1) if node u= d, stop, wherein drepresent destination;

2) if ube not plugged, application routing decision sends forward routing iinformation, otherwise, the application method that detours below:

2.1) from the step of the boundary formation of MCC at node sin the sequence information that place obtains, therefrom find the most approaching da sequence ( f ₁, f ₂..., f _n);

2.2) the shortest one in calculating path d( s, d): wherein pass through node c ₁, be designated as p ₀; Pass through node c' _n, be designated as p _n; Through two continuous MCC, be designated as p _i;

2.3) ifp0/pnbe chosen as shortest path, the intermediate objective node arriving isd'=c1/c'n; Otherwise, for the shortest path of choosingpi, 1 ￡ whereini<n, the intermediate objective node arriving isd'=ci+ 1.

7. method according to claim 3, is characterized in that, 3D grid M ₃initial value is given specifically to be comprised the following steps with node mark:

1) all malfunctioning nodes are labeled as faulty, all non-malfunctioning nodes are labeled as safe;

2) if node ufor safebut, its- xneighbours ,- yneighbours, and- zneighbours are faultyor can't-reach, mark ufor can't-reach;

3) recursively mark node until do not have new can't-reachnode occurs;

4) all faultywith can't-reachnode is all called unsafenode.

8. method according to claim 7, is characterized in that, in 3D grid, MCC identification and boundary formation comprise the following steps:

1) if z= hplace, xYthere is a new malfunctioning node in cross section u( u _x, u _y, h), define a new MCC, determine point ( u _x, u _y, h+ 1) be boundary node, simultaneously with point ucentered by, adopt depth-first search strategy, xYin plane, successively to surrounding, expand judging point uneighbor node whether be boundary node and collect frontier point information, find this MCC after all boundary nodes of this one deck, boundary information is broadcasted along boundary node, for z= h+ 1 xYthe border that cross section has existed, exists z= h's xYcross section is successively outwards searched for and definite border along this border from center, fault zone;

2), if two threatening areas have lap, two MCC can start to intersect at some At The Heights, at this moment by two MCC on this height this xYsection is merged into a MCC;

3) according to actual conditions, a given limitation in height region artificially;

4) end of scan successively, and given after limitation in height region, boundary information is existed xZwith yZtwo planes are broadcasted, and determine forbidden zone and critical zone definite border simultaneously.

9. method according to claim 8, is characterized in that, the definite of shortest route comprises:

1) feasibility inspection: at source node s, along+ x,+ ydirection sends detect-message, when one as along+ xthe information of direction runs into another MCC, and it will add (– x,+ y)-and (– x,+ z)-border is as boundary formation; Source node by inspection whether these two detection information can arrive respectively surface [ x _d: x _d, 0: y _d, 0: z _d], [0: x _d, y _d: y _d, 0: z _d]; If can not reach any one in these two surfaces, stop route, because there is not manhatton distance path;

2) present node urouting decision and message transmit, and comprise source node s, be specially:

2.1) all optional directions are increased to working direction set fin, find the MCC of all records;

2.2) for each MCC finding in previous step, from fmiddle removal those make impact point in critical zone and uneighbours along this direction, will arrive the direction of forbidden zone;

2.3) application fully adaptive and Minimal routing are processed from set fworking direction of middle selection;

2.4) routing iinformation is sent to forward to next node along this selected direction.

10. according to the method described in claim 4 or 7, it is characterized in that, uselessonce node definition arrives after this node for path, next step must Yan – x/ – yhuo person – x/ – y/ – Z direction moves, and it is not the shortest making path; When a non-malfunctioning node+ x/+ yor+ x/+ y/+ zwhen the neighbor node of direction is malfunctioning node, be referred to as can ' t-reachnode.