CN-122022102-A - Multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search

CN122022102ACN 122022102 ACN122022102 ACN 122022102ACN-122022102-A

Abstract

A multi-unmanned plane collaborative reconnaissance path planning method based on multi-step prospective search belongs to the technical field of unmanned plane coverage path planning. In each simulation step, all feasible paths of a plurality of future steps are searched and enumerated for each unmanned aerial vehicle in sequence by adopting breadth-first, five indexes of new coverage gain, total coverage quantity, distance rewards led to an uncovered area, boundary avoidance punishment and turning punishment are comprehensively evaluated in a weighted sum mode, and then a dynamic weight updating mechanism is designed to enable the unmanned aerial vehicle to adapt to the reconnaissance tasks under various conditions. The method can effectively improve coverage rate and convergence speed, ensures flight safety, and is suitable for large-scale collaborative reconnaissance scenes such as emergency rescue, disaster monitoring and the like.

Inventors

ZHANG SHAOQING
YU YUQI
XIA WEIGUO
FAN PENGYANG
ZHAO SHUANGYU
GAO HEFU
MA SONG
WANG XINWEI

Assignees

大连理工大学

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (8)

1. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search is characterized by comprising the following steps: The method comprises the following steps of S1, initializing a process, namely setting a three-dimensional space coordinate range of a reconnaissance area, setting grid resolution, setting parameters of a central coordinate of a hemispherical no-fly zone, setting unmanned aerial vehicle parameters, modeling an environment, and initializing a coverage state array; s2, main loop, executing for each simulation time step: s2.1, updating a coverage state array, namely calculating grids currently covered by each unmanned aerial vehicle sensor; S2.2, path planning, namely planning a next moving direction for each unmanned aerial vehicle, wherein the method specifically comprises the following steps: S2.2.1, generating a candidate direction set, and generating three candidate directions of left turn, straight run and right turn based on the current flight direction, wherein the left turn and the right turn adopt 45-degree steering angles in grid space; S2.2.2 path enumeration, namely starting from depth 0, expanding paths layer by layer, generating candidate directions in each step, and calculating the position of the next step; S2.2.3 Path validity check: when path enumeration is carried out, the validity judgment of the boundary and the no-fly zone is needed for each position where the unmanned aerial vehicle should arrive at in the next step, only the valid paths are reserved for enqueuing, and when the path depth reaches the forward-looking depth, the valid paths are stored in a complete path set; S2.2.4 planning the next best moving direction of each unmanned plane by adopting a multi-step look-ahead breadth-first search algorithm BFS, and for each complete path Calculating a comprehensive profit score; S2.3, updating the position, the direction and the current coverage rate of each unmanned aerial vehicle; s3, a multi-unmanned aerial vehicle cooperative mechanism adopts sequential cooperative programming, namely sequentially programming paths of unmanned aerial vehicles according to the number sequence at each time step, and the unmanned aerial vehicle after programming uses an updated coverage state array of the preamble unmanned aerial vehicle Performing profit evaluation, so that the profit evaluation is automatically distributed to uncovered areas, and implicit coordination is realized; s4, an emergency path planning strategy, when BFS searches for a feasible path, the method sequentially executes (1) degradation look-ahead depth And (2) if the strategy (1) fails, calculating the azimuth vector of the no-fly zone to judge the left and right sides of the no-fly zone, and reversing the steering.
2. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search according to claim 1, wherein the step S1 is specifically: S1.1, setting a three-dimensional space coordinate range of a reconnaissance area: Wherein, the Respectively representing a lower limit coordinate and an upper limit coordinate of the length of the reconnaissance area; respectively representing a lower limit coordinate and an upper limit coordinate of the width of the reconnaissance area; respectively representing a lower limit coordinate and an upper limit coordinate of the height of the reconnaissance area; s1.2 setting the grid resolution as Each grid is a cube, Representing each grid at Length on the shaft; s1.3, setting parameters of central coordinates of a hemispherical no-fly zone: Wherein, the Respectively representing the central coordinates of the hemispherical no-fly zone, the radius of the hemispherical no-fly zone is The safe buffer distance is ; S1.4, setting unmanned aerial vehicle parameters, namely the number of unmanned aerial vehicles The horizontal view angle and the vertical view angle of the forward radar are respectively Distance of furthest detection of radar Depth of look ahead ; S1.5, environment modeling, namely discretizing a reconnaissance area into a three-dimensional grid set: Wherein, the Representing the total number of grids within the scout area, Representing a total of partitioned grid matrices within a scout area, a total of Row, 3 columns, each column representing coordinates in one dimension; S1.6, initializing an overlay state array: Wherein, the Indicating that grid j is uncovered Indicating that grid j has been covered; And S1.7, finally initializing the state of the unmanned aerial vehicle, and setting initial positions and initial orientations for each unmanned aerial vehicle.
3. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step look-ahead search according to claim 2, wherein the step S2.1 is specifically that in the step S2.1, whether the grid is covered by the forward radar sensor is required, and the specific determination method is as follows: (1) Conversion of the coordinate system by the position of the unmanned aerial vehicle in the original coordinate system As the origin, an onboard right-hand coordinate system is constructed: Wherein, the Represents the current flight direction vector of the unmanned aerial vehicle, Representation of Is provided with a die for the mold, A unit vector which respectively represents the machine head direction, the right side direction of the machine body and the lower direction of the machine body, Representing the original coordinate system The unit vector of the axis is defined, Representing a rotation matrix, grid The coordinates in the on-board right-hand coordinate system are: Wherein, the Representation grid Coordinates in the original coordinate system; (2) The judging method comprises the following steps: Firstly, calculating a horizontal azimuth angle and a vertical azimuth angle of a grid to be determined relative to the position of the unmanned aerial vehicle: Wherein, the Representing the coordinates of the grid in the on-board right-hand coordinate system A component of, Components and A component; The tolerance angle is then calculated to introduce angular tolerance compensation: Wherein, the And (3) with Representing the tolerance angle between the grid world sphere and the field of view of the drone camera, And (3) with Representing the distance between the position of the unmanned aerial vehicle and the center of the grid on the horizontal plane and the vertical plane; finally, it is established that the conditions for judging that a certain grid is covered by the unmanned aerial vehicle sensor to be met simultaneously are: 。
4. the multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step look-ahead search according to claim 3, wherein the step S2.2.1 specifically comprises: Representing all candidate directions of the drone by defining a set of discrete directions: For each candidate direction Calculating an included angle cosine for screening out the direction most conforming to the expected steering angle: determination by cross product of two direction vectors Whether left or right of the current flight direction of the unmanned aerial vehicle, for determining the sign of the angle: Wherein, the Representation of And (3) with A vector obtained by cross-product; Representation of A kind of electronic device Component of if Then Representing left turn, recording the included angle as a negative value, otherwise, right turn, recording the included angle as a positive value; Respectively represent the current flight direction of the unmanned plane A kind of electronic device Components and A component; respectively represent candidate directions A kind of electronic device Components and A component.
5. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step look-ahead search according to claim 4, wherein the position of the next step in step S2.2.2 is: Wherein, the Indicating the next position the drone should reach, Representing the current location of the drone, Representing the selected direction vector.
6. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search according to claim 5, wherein the validity judgment of the boundary and the no-fly zone in step S2.2.3 is specifically: (1) Boundary inspection: (2) And (3) checking a no-fly zone, and defining a safety radius: Wherein, the Representing minimum safety distance from unmanned aerial vehicle to center of no-fly zone, and position of unmanned aerial vehicle If and only if the following formula is satisfied indicating that no-fly zones have been entered: 。
7. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step look-ahead searching of claim 6, wherein in step S2.2.4, for each complete path The calculation of the comprehensive profit score is specifically as follows: Wherein, the The new number of covered grids representing the path, Indicating the total number of covered grids on the path, Indicating that the distance awarded item is to be played, A boundary penalty term is represented and is used, The turn smoothness score is indicated as such, Is a weight coefficient; (1) New number of covered grids: Path-to-path Calculating the number of grids currently covered and previously uncovered: Wherein, the The path length is indicated as such, Represent the route No A grid set over which the individual locations can cover; (2) Total number of covered grids: Wherein, the Represent the first The number of elements within the grid set that a location can cover; (3) Distance rewarding item: First define a distance threshold The distance of the end of the path to the nearest uncovered grid is then calculated: The distance rewarding items are: Wherein, the The more the path end is closer to the uncovered area, the higher the prize is; (4) Boundary penalty term: The minimum distance from the end of the path to the boundary is first calculated: the boundary penalty term is: Wherein, the The position coordinates of the end of the path are represented, When the end of the path is too close to the boundary, penalty is applied to avoid trapping the unmanned aerial vehicle near the boundary; (5) Turning smoothness score term: A turning smoothness score rewarding mechanism is introduced: (5.1) statistics of historical turning times the system maintains a flight action historical queue for each unmanned aerial vehicle Recording the last 20 steps of flight actions, for each flight action, using 1 for left turn, 2 for straight run, and 3 for right turn, and counting historical turning times: Wherein, the Taking a value of 1 when the flying action is not straight running as an indication function, otherwise, taking a value of 0; (5.2) turning smoothness score weight: dynamically adjusting the weight coefficient of the turning smoothness score according to the historical number of turns: Wherein, the As a basis weight coefficient of the weight of the object, The dynamic weight coefficient enables the unmanned aerial vehicle which turns frequently to be restrained by larger turning; (5.3) turning smoothness score: calculate a cumulative score for directional consistency between adjacent steps on the path: Wherein, the Representing the total number of steps of the drone path, Represent the route No Step and No. The calculation method of the included angle of the step direction vector comprises the following steps: Wherein, the Represent the first The turning smoothness score gives a high score for straight line flight and a low score for large angle turns; (5.4) final Effect in the Complex score function, the contribution of the turning smoothness score term is ; (6) Adaptive weighting mechanism: In order to adapt to the demand difference of different stages of the scout task, a self-adaptive weight adjustment mechanism is introduced, and weight configuration is dynamically selected according to the coverage rate of the current region: (6.1) calculating the current coverage: (6.2) setting coverage threshold Selecting weight configuration according to the current coverage rate: in the early stages of reconnaissance, i.e. Emphasis is placed on the fast overlay and exploration, at which stage the weight of the new overlay grid number Maximum, turn smoothness score term basis weight is The actual turning smoothness score weight is dynamically adjusted to be according to the historical turning times Applying cumulative penalties to frequent turns while prioritizing the rapid expansion of coverage; Late stage of reconnaissance, i.e Emphasis on fine coverage and trajectory optimization, at this stage, distance rewards term And boundary penalty term weights The basic weight of the turning smoothness score is improved to The actual turning smoothness score weight is dynamically adjusted to The unmanned aerial vehicle is guided to fill the residual blank area preferentially through stronger turning smoothness score items, invalid round-trip is reduced, and track quality is optimized.
8. The multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search according to claim 7, wherein the method for calculating the azimuth vector of the no-fly zone in step S4 is as follows: by cross-product To judge the left and right sides of the no-fly zone, wherein, Respectively, are the azimuth vectors of the no-fly zones Components and A component.

Description

Multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search Technical Field The invention belongs to the technical field of unmanned aerial vehicle coverage path planning, and particularly relates to a multi-unmanned aerial vehicle collaborative reconnaissance path planning method based on multi-step prospective search, which is mainly applicable to scenes of fixed-wing unmanned aerial vehicles in emergency rescue, disaster monitoring and the like which are in urgent need of completing large-area reconnaissance coverage. Background In the multi-unmanned plane cooperative area coverage reconnaissance task, the prior art mainly adopts the following methods: A global path planning method based on an optimization algorithm. For example, CN119739201a (university of aerospace in south kyo) employs a modified gray wolf optimization algorithm (GWO), CN120508118A (university of dulcimer) employs a modified manual lemming algorithm (ALA), and path optimization is performed by constructing multiple objective evaluation functions (including fuel, altitude, threat avoidance, etc.). The method has high computational complexity and depends on a large number of iterations, the real-time online re-planning requirement under the high dynamic countermeasure environment is difficult to meet, the threat avoidance is realized only through punishment items, and a precise modeling and path-level filtering mechanism for the geometry of the no-fly zone is lacked. Based on greedy strategy and ACombined dead zone processing method. For example, CN120252709a (Jiangsu Kovida) proposes a full coverage path planning method that fuses greedy with a, calling a to get rid of the poverty by "dead zone triggering". The method does not consider the limited field angle and the three-dimensional coverage characteristic of the forward radar of the unmanned aerial vehicle, the coverage detection is simplified into 'point access or not', and the real sensor model cannot be reflected. An overlay method based on deep reinforcement learning. For example, CN118605606a (university of Jiangsu science) proposes a SADQN-based unmanned aerial vehicle group coverage method, combined with a guide for initial exploration. However, the method relies on a large number of training samples and experience playback, has slow convergence speed, relies on priority ordering in multi-machine cooperation to solve conflicts, and cannot guarantee 100% obstacle avoidance safety. A special obstacle avoidance method for a no-fly zone. For example, CN119247985A (university of information engineering in south Beijing) proposes a track planning method for "completely avoiding no-fly zones", which approximates the track by straight line segments and constrains the intersection points to ensure no entry into no-fly zones. But the method focuses on a communication relay scene, aims at optimizing the communication rate of users instead of maximizing the coverage rate of areas, and models the no-fly zone as an infinitely high cylinder without considering the hemispherical no-fly zone (such as airports and bases) common in actual combat. The other patent CN119356394A (university of northwest industry) is oriented to a high-speed reentry aircraft, adopts an hp self-adaptive pseudo-spectrum method to solve the optimal track, has complex dynamics model and huge calculation cost, and is not suitable for on-line path planning of a common fixed-wing unmanned plane. The existing method generally adopts an evaluation function with fixed weight to evaluate paths, and cannot adapt to the difference of requirements of different stages of a reconnaissance task. In the early stage of the task, the coverage range needs to be rapidly expanded, the coverage of a new area needs to be emphasized, in the later stage of the task, the coverage rate is higher, the aims of filling a blank area and optimizing track smoothness are fulfilled, and invalid round trip is avoided. The fixed weight strategy has difficulty in satisfying the contradictory requirements of the two phases, resulting in overall inefficiency. In addition, the continuity and turning expense of the track are not considered in the path evaluation of the conventional method, the frequently-turned zigzag track is easy to generate, the energy consumption is increased, and the maneuvering constraint of the fixed-wing unmanned aerial vehicle is not facilitated. Disclosure of Invention The invention aims to solve the problems of low path planning efficiency, easy sinking into local optimum, fixed weight of an evaluation function, inadaptability to task stage change and poor track smoothness in the prior art, provides a multi-unmanned plane collaborative reconnaissance path planning method based on multi-step prospective search, which combines a benefit evaluation function, a forward radar three-dimensional coverage model, a hemispherical no-fly zone obstacle avoidance mechanism and an emerg