A method for full-coverage path planning for drones that integrates greedy strategies and A* algorithm

Abstract

This invention discloses a method for full-coverage path planning of unmanned aerial vehicles (UAVs) that integrates a greedy strategy and the A* algorithm. It belongs to the technical field of full-coverage path planning, and its operation steps are as follows: First, initialize the map and obstacles, and define marker quantities; second, evaluate the attributes of surrounding points and use a greedy strategy to select the next point; then, determine whether a dead zone has been entered. If so, use the A* algorithm to replan a path to escape the dead zone; if not, continue using the greedy strategy to find the next forward point; next, check the coverage completion status, checking whether each point is covered; finally, verify the effectiveness of the full-coverage traversal algorithm in a 3D environment. This invention solves the problems of low efficiency, low coverage rate, high repetition rate, and limitations in complex environments of current technologies when dealing with full-coverage traversal inspections, detections, and cleaning in chemical industrial parks and coal mines, as well as agricultural plant protection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of full coverage path planning, and relates to a full coverage traversal path planning problem. BACKGROUND

[0002] Currently, full coverage traversal is a key function of robots, unmanned aerial vehicles and other automated devices when performing complex tasks such as environmental monitoring, coal mine inspection, cleaning, agricultural plant protection, etc., and its goal is to ensure that every location in the task area is visited. The traditional A-star algorithm is widely used in the field of path planning due to its high path planning efficiency and good optimization performance, but in full coverage tasks, its search space is large and unnecessary repeated paths are easily generated, making it difficult to meet the efficiency requirements in complex environments. At the same time, as a simple and efficient local search method, the greedy strategy can quickly select the current optimal solution, but due to the lack of global vision, it is easy to fall into local optimization, resulting in partial areas not being covered or redundant paths. In complex dynamic environments, a single A algorithm or greedy strategy cannot balance global optimization and local real-time adjustment.

[0003] (Chinese patent: Patent No. CN106979785A): This method combines a biological excitation neural network model with a backtracking mechanism to solve the robot deadlock problem through a dynamic A-star algorithm, and introduces a market mechanism to optimize the selection of backtracking nodes, thereby reducing repeated coverage and total coverage time. Its shortcomings are that it has weak adaptability to dynamic environments, the backtracking mechanism mainly targets global deadlock problems, but the processing strategy for small dead zones in complex terrains is not clear enough, which may lead to reduced efficiency, and it does not focus on coverage uniformity, and the complex algorithm may increase the demand for computing resources, requiring high hardware performance.

[0004] (Chinese patents: Patent Nos. CN118466301A and CN115727850A): The above-mentioned patents both focus on full coverage path planning for multi-robot or multi-unmanned aerial vehicle systems, aiming to improve coverage efficiency and reduce task time consumption through various methods. However, these methods still have certain limitations in specific scenarios. For example, using a backtracking mechanism to solve deadlock problems is less efficient in scenarios with dense or sparse node distribution; the combination of A-star algorithm and reinforcement learning optimizes the monitoring path, but highly depends on the accuracy of the environment model, and once the model is not accurate, the monitoring efficiency will be affected; and the bisection greedy strategy is prone to local optimal in complex environments, making it difficult to achieve global optimal coverage and resulting in low efficiency. SUMMARY

[0005] In view of the above problems, the present application aims to provide a UAV full coverage path planning method (UAV full coverage path planning method fusing greedy strategy and A-star algorithm) fusing greedy strategy and A-star algorithm,

[0006] The technical solution of the present application is: the UAV full coverage path planning method fusing greedy strategy and A-star algorithm, preferably a full coverage traversal path planning research method for gas inspection under chemical plant area and coal mine, and the operation steps are as follows:

[0007] Step (1): initialize the map and obstacles, and define a flag quantity;

[0008] Step (2): calculate the information of points around the current point and evaluate its attributes, add a dynamic updating coverage value mechanism, and improve the moving rule, and use the greedy strategy to select the next advancing direction;

[0009] Step (3): design a dead zone triggering mechanism, if any condition of the dead zone is reached, trigger the event, and call the A-star algorithm to re-plan a path to get out of the dead zone;

[0010] Step (4): check the coverage completion condition;

[0011] Step (5): three-dimensional environment simulation verification.

[0012] Further, the step (1) is specifically: initializing the map and obstacles, establishing a grid map according to the facility layout of the chemical plant area; and traversing each point on the map to determine whether the point is an obstacle, if it is an obstacle, marking it as unpassable;

[0013] Meanwhile, set a starting point and a path variable, and define a flag quantity to detect whether to enter a dead zone, i.e. no accessible uncovered point.

[0014] Further, the step (2) is specifically: obtaining the coordinates of 8 adjacent points in the surrounding of the current point, and determining whether they are within the map boundary;

[0015] For each adjacent point within the boundary, evaluate its coverage attribute: if the coverage value is 0 and it is not an obstacle, mark it as a priority coverage point; if the coverage value is greater than 0 and it is not an obstacle, mark it as an accessed point; if it is an obstacle, mark it as a covered and unpassable point;

[0016] Calculate the heuristic value of each point, select the point with the highest score as the next advancing point; dynamically update the coverage value to avoid excessive coverage.

[0017] Further, first, find all grid point indexes in the current UAV detection range, increase the point coverage value matrix T in the detection range by a step value r0, judge whether the coverage value reaches the upper limit C, if it exceeds, limit it to C, that is, T(x, y) = min(T(x, y) + r0, C), second, the UAV selects a point with the smallest coverage value in the coverage range as the traction point, the traction point is the target point of the UAV movement, and by comparing with the current UAV direction angle, the nearest target point is selected. This method can make the UAV select the most needed area for coverage, rather than blind movement, wherein the angle difference between the UAV and the traction point is θ i = arctan(d yi ,d xi );

[0018] If special circumstances are encountered, improved movement rules are called to solve the path planning selection problem in specific scenarios; first, the information of the surrounding points is judged, wherein the around_points array contains the state of each point around the current position, and the movement direction is determined by judging whether these points are empty; second, assuming that the current point P0(x0, y0) and the surrounding points are P1, P2, P3, P4, P5, P6, P7, and P8, the state of each point can be represented as:

[0019]

[0020] Finally, when running to the end of the path or a T-shaped intersection, the smallest dead zone relative to the path is selected, that is, P1 = 1, P4 = 1, P6 = 1, P2 = 0, P7 = 0, the improved movement rule is called, and the next advancing point is selected by judging the position of the current point P0(x0, y0), that is:

[0021]

[0022] At the same time, according to the innovative mechanism of dynamically updating the coverage value, the coverage value of the priority coverage point is adjusted in real time to ensure that the coverage value accumulates gradually but does not exceed the set limit value, so as to optimize the path selection and improve the coverage efficiency.

[0023] Further, the step (3) is specifically: a dead zone triggering mechanism is designed to judge whether to enter a dead zone, if the surrounding points are all covered or are obstacles, the event is triggered, the trapped information is input, a new path is planned to escape the dead zone by combining and calling the A-star algorithm, and the traversal is restarted.

[0024] Further, the step (4) is specifically: all points in the map are checked to judge whether they have been covered, if not, step (2) is performed, and if full coverage is completed, the path planning is ended.

[0025] Further, the step (5) is specifically: firstly, three-dimensional modeling is carried out on the chemical plant area in Blender. Secondly, the three-dimensional model is imported into the gazebo environment. Finally, traversal simulation is carried out in the three-dimensional environment.

[0026] The beneficial effects of the present application are: 1. The path planning flexibility of the traditional method is insufficient, and the adaptability is poor. In a simple environment, it can realize fast coverage, and in a complex environment, it can also dynamically adjust the path to ensure the integrity of coverage. 2. The ladder method is used to dynamically update the grid point coverage value in the detection range, and the maximum coverage value is limited. The waste of resources caused by the uncontrolled coverage value in the traditional algorithm is solved, and the coverage efficiency and coverage uniformity are guaranteed. 3. By improving the movement rule, the dead zone with smaller turning path is preferred, which effectively reduces the repeated path and improves the efficiency of full coverage traversal. It is especially suitable for complex terrains such as T-shaped intersections, optimizes path selection, and ensures coverage rate and flexibility. This strategy can also adapt to boundary conditions, avoid missing or redundancy, and improve the intelligence and adaptability of the overall algorithm; 4. Design a dead zone trigger mechanism. If the points around the current point all meet the trigger conditions, the event will be triggered immediately, and the A-star algorithm will be called to avoid wasting time for the UAV in this area and improve coverage efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is the full coverage traversal path planning research method flow chart of the present application;

[0028] Figure 2 is the full coverage traversal path planning algorithm flow chart of the present application;

[0029] Figure 3 is the full coverage traversal path planning effect diagram of the present application. DETAILED DESCRIPTION

[0030] The specific technical solutions of the present application will be further described in detail below in combination with specific examples.

[0031] As shown in the figure, the UAV full coverage path planning method of the present application fuses the greedy strategy and A-star algorithm;

[0032] The operation steps are as follows: first, initialize the map and obstacles, and define the flag quantity;

[0033] Secondly, evaluate the properties of the surrounding points and use the greedy strategy to select the next point;

[0034] Then, it is judged whether to enter the dead zone. If yes, use the A-star algorithm to re-plan a path to get out of the dead zone. If no, continue to use the greedy strategy to find the next advancing point;

[0035] Finally, check the coverage completion condition and check whether each point is covered.

[0036] The specific implementation steps are as follows:

[0037] S101: As shown in FIG. (2), it is a flow chart of the full coverage path planning algorithm. First, initialize the map and obstacles, and establish a grid map; traverse each point on the map, judge whether the point is an obstacle, if it is an obstacle, mark it as covered, indicating that it cannot be passed through;

[0038] S102: Define the starting point and path variable, and define a flag variable to judge whether to enter the dead zone, that is, all surrounding points have been covered or are obstacles;

[0039] S201: Calculate the coordinates of the 8 adjacent points around the current point, and judge whether it is within the map boundary;

[0040] S202: Evaluate the coverage properties of the surrounding points: if the coverage value is 0 and not an obstacle, mark it as a priority coverage point; if the coverage value is greater than 0 and not an obstacle, mark it as a visited point; if it is an obstacle, mark it as a covered and impassable point;

[0041] Calculate the heuristic value of each point, and select the point with the highest score as the next advancing point;

[0042] If a special situation is encountered, improve the movement rule to solve the path planning selection problem in a specific scene. First, the information of the surrounding points is judged, and the state of each point around the current position is contained in the around_points array. The movement direction is determined by judging whether these points are empty. Second, assuming that the current point P0(x0, y0) and the surrounding points P1, P2, P3, P4, P5, P6, P7, P8, the state of each point can be represented as:

[0043]

[0044] Finally, when running to the end of the path or a T-junction, select the dead zone with the smallest relative path, that is, P1=1, P4=1, P6=1, P2=0, P7=0, call the improved movement rule, and select the next advancing point by judging the position of the current point P0(x0, y0), that is

[0045]

[0046] S203: A dynamic coverage value update method is added. First, all grid point indices within the current UAV's detection range are found. The coverage value matrix T within the detection range is incremented by a step value r0. It is then checked whether the coverage value has reached the upper limit C. If it does, it is limited to C, i.e., T(x,y) = min(T(x,y) + r0,C). Second, the UAV selects the point with the smallest coverage value within the coverage range as the traction point. The traction point is the target point for the UAV's movement, and the closest target point is selected by comparing it with the current UAV's direction angle. This method allows the UAV to select the most needed area for coverage instead of moving blindly. The angle difference between the UAV and the traction point is θ. i =arctan(d yi ,d xi );

[0047] S301: Design a dead zone detection mechanism. If all points around the current point are either covered or obstacles, it is determined that the point has entered a dead zone, and an event is triggered.

[0048] S302: When it is determined that the current point is trapped in a dead zone, as shown in Figure (3), an event is triggered after a point is trapped in a dead zone. Then, the A* algorithm is called to perform path planning. The next optimal point is selected by f = g + h, instead of simply selecting the point with the shortest straight distance. That is, a path is found from the current point to the nearest uncovered area, the dead zone is jumped out, and the greedy strategy is used to select the next forward point.

[0049] S401: Check if every point on the map is covered. If all points are covered, end the path planning.

[0050] like Figure 3 As shown, (a) is the result of the A* algorithm's round-trip full-coverage traversal path planning, showing a path length of 740, 113 turns, and a repetition rate as high as 50%; (b) is the result of the full-coverage traversal path planning combining the greedy strategy and the A* algorithm, showing a path length of 366.8, 94 turns, and a repetition rate of 3.6%; (c) is the result of the path planning further improved based on the combination of the greedy strategy and the A* algorithm, showing a path length of 366.2, 88 turns, and a repetition rate of only 3.4%. The comparison of results shows that the improved algorithm described in this invention significantly improves path length, number of turns, and repetition rate, effectively solving the problems of high number of turns, long paths, and high repetition rates in the conventional A* round-trip algorithm during full-coverage traversal.

[0051] S501: First, a three-dimensional environment model of the chemical plant area is established in Blender, second, the modeled three-dimensional model is imported into gazebo, then, the quadcopter unmanned aerial vehicle is opened in the imported three-dimensional chemical plant area, finally, the improved fusion algorithm is used for full coverage traversal simulation experiment, and finally the simulation experiment shows that the unmanned aerial vehicle can fly and patrol in the three-dimensional environment.

[0052] The present application solves the problems of low efficiency, low coverage rate, high repetition rate and limitations in complex environments in current technology in response to full coverage traversal inspection, detection, cleaning in chemical industry park, coal mine and agricultural plant protection; first, the greedy strategy and A star algorithm are fused, the local property of the greedy strategy and the global property of the A star algorithm are combined, the flexibility of path planning in the traditional method is compensated, the adaptability is poor, and the full coverage in simple environment can be realized, the path in complex environment can be dynamically adjusted, and the integrity of the coverage is ensured; second, the dynamic updating coverage value method is added, the efficiency is avoided due to excessive coverage; then, the moving rule is improved, the movement in special situations such as T-shaped intersection and map boundary end is solved, the repeated path is reduced, and the efficiency is improved; finally, the dead zone triggering mechanism is designed, if the current point is trapped in death, the event is triggered, the A star algorithm is quickly called to escape, and the local optimal condition is got rid of. The efficiency of full coverage is effectively improved.

Claims

1. A method for full coverage path planning of a UAV by fusing a greedy strategy and an A* algorithm, characterized in that, The operation steps are as follows: Step (1): initialize the map and obstacles, define the flag quantity; Specifically: initialize the map and obstacles, establish a grid map; and traverse each point on the map to determine whether the point is an obstacle, if it is an obstacle, mark it as not passable; At the same time, set the starting point and path variable, define a flag quantity to detect whether to enter the dead zone, i.e. no accessible uncovered point; Step (2): calculate the information of the points around the current point and evaluate its attributes, add a dynamic updating coverage value mechanism, and improve the moving rule, and use the greedy strategy to select the next advancing direction; Specifically: get the coordinates of the 8 adjacent points around the current point, and judge whether they are within the map boundary; For each adjacent point within the boundary, evaluate its coverage attribute: if the coverage value is 0 and not an obstacle, mark it as a priority coverage point; if the coverage value is greater than 0 and not an obstacle, mark it as an already visited point; if it is an obstacle, mark it as a covered and non-passable point; Calculate the heuristic value of each point, and select the point with the highest score as the next advancing point; dynamically update the coverage value to avoid excessive coverage; Specifically, first, find all grid point indexes within the current UAV detection range, increase the coverage value matrix T within the detection range by a step value r0, and judge whether the coverage value reaches the upper limit C, if it exceeds, limit it to C, i.e. T(x,y) = min(T(x,y)+r0,C), second, select a point with the smallest coverage value within the UAV coverage range as the traction point, the traction point is the target point of the UAV movement, and by comparing with the direction angle of the current UAV, select the nearest target point; The above method allows the drone to select the area that needs coverage the most, where the angle difference between the drone and the traction point is θ i = arctan(d yi , d xi ) If a special situation is encountered, call the improved moving rule to solve the path planning selection problem in a specific scenario; first, judge the information of the surrounding points, which include the status of each point around the current position in the around_points array, and decide the moving direction by judging whether these points are empty; second, let the current point P0(x0,y0), the surrounding points are P1, P2, P3, P4, P5, P6, P7, P8, and the state of each point is represented as: Finally, when running to the end of the path or a T-shaped intersection, select the dead zone with the smallest relative path, i.e. P1=1, P4=1, P6=1, P2=0, P7=0, call the improved moving rule, and select the next advancing point by judging the position of the current point P0(x0,y0), i.e. At the same time, according to the innovative mechanism of dynamic updating of coverage value, the coverage value of the priority coverage point is adjusted in real time to ensure that the coverage value accumulates gradually but does not exceed the set limit, so as to optimize the path selection and improve the coverage efficiency; Step (3): design a dead zone triggering mechanism, if any condition of the dead zone is reached, trigger the event and call the A-star algorithm to re-plan a path to escape the dead zone; Step (4): check the completion of coverage; Step (5): three-dimensional environment simulation verification.

2. The method of claim 1, wherein the fusion of the greedy strategy and the A* algorithm is performed by: The step (3) is specifically: designing a dead zone triggering mechanism, judging whether to enter a dead zone, if all surrounding points have been covered or are obstacles, triggering an event, inputting a trapped information, planning a new path to escape from the dead zone by combining and calling A-star algorithm, and starting to traverse again.

3. The method of claim 1, wherein the fusion of the greedy strategy and the A* algorithm is performed by: The step (4) is specifically: checking all points in the map, judging whether all points have been covered, if not, performing the step (2), if completing full coverage, ending the path planning.

4. The method of claim 1, wherein the fusion of the greedy strategy and the A* algorithm is performed by: The step (5) is specifically: importing the built 3D model of the chemical plant in Blender into gazebo, performing full coverage traversal simulation in the environment, and verifying the effectiveness of the algorithm.

Images