Multi-point distribution unmanned aerial vehicle path planning method

Abstract

The invention relates to the technical field of logistics, in particular to a multi-point distribution unmanned aerial vehicle path planning method, which comprises the following steps: firstly, carrying out environment modeling on a distribution environment of multiple distribution points in a three-dimensional environment, and setting a target function as a minimum path total length; a complex three-dimensional multi-target path planning problem is decoupled into two levels of access sequence optimization and three-dimensional fine path planning, and on the premise of ensuring that the unmanned aerial vehicle safely avoids all obstacles, the optimal sequence of accessing a plurality of dispersed distribution points is calculated; and a three-dimensional collision-free trajectory which is shortest in distance and conforms to flight constraints is planned for each path in the sequence, and multi-point distribution unmanned aerial vehicle path planning is realized, so that the overall efficiency, safety and economy of unmanned aerial vehicle logistics distribution are remarkably improved, and the operation cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of logistics, in particular to a multi-point distribution UAV path planning method. BACKGROUND

[0002] With the popularization of UAV logistics distribution, the demand for UAV logistics path planning technology for multi-demand point continuous distribution tasks is more urgent. Unlike traditional point-to-point UAV distribution, the core challenge of multi-demand point distribution is not simply the superposition of multiple independent flight tasks, but a complex coupled problem that requires simultaneous optimization of access sequence and three-dimensional fine path. UAVs not only need to safely and efficiently shuttle in complex urban environments full of obstacles, but also must visit multiple dispersed customer points in a globally optimal or near-optimal order before returning to the warehouse. Existing solutions often isolate path planning and sequence optimization, making it difficult to balance computational complexity and path quality. SUMMARY

[0003] The purpose of the present application is to provide a multi-point distribution UAV path planning method, aiming to solve the technical problem of difficult collaborative optimization of global efficiency and local obstacle avoidance when UAVs perform multi-target point continuous distribution in complex urban environments.

[0004] To achieve the above purpose, the present application provides a multi-point distribution UAV path planning method, comprising the following steps:

[0005] Step 1: Select the distribution environment, confirm the transportation warehouse, distribution points and obstacle information;

[0006] Step 2: Discretize the distribution environment and set the obstacle function;

[0007] Step 3: Build a path search model, set the objective function and select the constraint conditions;

[0008] Step 4: Plan the flight path using a double-layer genetic-A* optimization algorithm combining sequence and path, and perform path quality evaluation.

[0009] Optionally, in the distribution environment, the transportation warehouse coordinates are , distribution points , represent the coordinates of the th distribution point, obstacles, , the distribution path starts from the transportation warehouse , visits all demand points in turn, and finally returns to the transportation warehouse .

[0010] Optionally, the expression for discretizing the delivery environment in step 2 is as follows:

[0011]

[0012] where, is the position of each grid cell, and is the number of grids in three directions;

[0013] The expression of the obstacle function is as follows:

[0014]

[0015] is 1 if there is an obstacle at position .

[0016] Optionally, in step 3, the state space is set, and the state is defined as a three-dimensional grid coordinate, i.e., the position of the UAV , is the th waypoint on the path;

[0017] The action space is set, and the allowed actions of the UAV include 26 directions of movement, with the expression as follows:

[0018]

[0019] where, is defined as 26 discrete movement vectors in the action space, and each vector has components taking values of -1, 0, or 1, representing the relative displacement to the neighbor nodes in the three-dimensional grid.

[0020] Optionally, the objective function in step 3 is to minimize the total length of the path , with the expression as follows:

[0021]

[0022] where, a path is sought such that the total length of the path is the shortest, is the solution to be sought, i.e., the entire path, is the th waypoint on the path;

[0023] The constraint conditions are as follows:

[0024]

[0025]

[0026]

[0027]

[0028]

[0029] in, It is an indicator function, when the position When occupied by obstacles, When position When it is safe to pass, , indicating path Every point on All must be located in free space and must not intersect with any obstacles;

[0030] The minimum and maximum altitudes at which drones are permitted to fly;

[0031] To calculate path points Gather all buildings The closest distance to any building in the area. To maintain the required minimum safe distance, every point along the flight path must maintain a minimum distance from all buildings. A distance of meters;

[0032] This means that the drone's flight path completely covers all logistics demand points, visiting every designated location without omission;

[0033] Starting point of the drone path and the end point All are warehouses .

[0034] Optionally, during the execution of step 4, the access sequence of multiple target points is first optimized, and then a three-dimensional collision-free trajectory is planned in real time for each path segment in the sequence.

[0035] The execution steps in the sequence optimization layer are as follows:

[0036] Individual / chromosome encoding;

[0037] Each coded individual corresponds to a visit sequence. Multiple visit sequences are randomly generated to form the initial population.

[0038] The individual visit sequence is fed into the path optimization layer to solve the fitness function; the shorter the path, the higher the fitness.

[0039] Chromosome sequences undergo selection, crossover, and mutation operations to form new sequence populations;

[0040] Termination after the maximum number of iterations is reached, otherwise continue to return the sequence optimization layer first step to start execution;

[0041] The execution steps of the path optimization layer are as follows:

[0042] The access sequence transmitted from the sequence optimization layer is input;

[0043] According to the sequence order, the three-dimensional collision-free path of each segment path is calculated in turn;

[0044] The distances of all segment paths in the sequence are added up and returned to the first layer as the basis for fitness evaluation of the individual.

[0045] Optionally, the path quality evaluation indicators are as follows:

[0046] Path efficiency:

[0047]

[0048] The ratio of the straight-line distance between the starting point and the ending point to the total length of the actual flight path; the closer the ratio is to 1, the less redundant the path is and the higher the efficiency is;

[0049] Safety indicator:

[0050]

[0051] The calculation result is the minimum distance between the unmanned aerial vehicle and all obstacles in the entire flight path; the larger the value is, the more sufficient the safety margin is, and the safer the flight is;

[0052] Height change smoothness:

[0053]

[0054] The average value of the height change between adjacent path points is calculated to reflect the degree of vertical jolt; the smaller the value is, the more gentle the height change is, and the more stable the flight trajectory is.

[0055] The present application provides a multi-point distribution unmanned aerial vehicle path planning method, which firstly models the distribution environment of multiple distribution points in a three-dimensional environment, sets the target function as minimizing the total path length, and then decouples the complex three-dimensional multi-objective path planning problem into two levels of access sequence optimization and three-dimensional fine path planning. Under the premise of ensuring that the unmanned aerial vehicle safely avoids all obstacles, the optimal order of visiting multiple scattered distribution points is calculated, and a three-dimensional collision-free trajectory with the shortest distance and conforming to the flight constraints is planned for each path in this order. The multi-point distribution unmanned aerial vehicle path planning is realized, thereby significantly improving the overall efficiency, safety and economy of unmanned aerial vehicle logistics distribution, and reducing operating costs. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below only show some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0057] Figure 1 is a schematic diagram of a double-layer genetic-A* optimization algorithm in the method of the present application.

[0058] Figure 2 is a schematic diagram of the flight trajectory of the unmanned aerial vehicle at 16 angles in the specific embodiment of the present application.

[0059] Figure 3 is a schematic diagram of the flight trajectory from the starting point 0 to the distribution point 1 in the specific embodiment of the present application.

[0060] Figure 4 is a schematic diagram of the flight trajectory from the distribution point 1 to the distribution point 4 in the specific embodiment of the present application.

[0061] Figure 5 is a schematic diagram of the flight trajectory from the distribution point 4 to the distribution point 2 in the specific embodiment of the present application.

[0062] Figure 6 is a schematic diagram of the flight trajectory from the distribution point 2 to the distribution point 3 in the specific embodiment of the present application.

[0063] Figure 7 is a schematic diagram of the flight trajectory from the distribution point 3 to the distribution point 5 in the specific embodiment of the present application.

[0064] Figure 8 is a schematic diagram of the flight trajectory from the distribution point 5 to the starting point 0 in the specific embodiment of the present application.

[0065] Figure 9 is a schematic diagram of the flight path elevation profile of the unmanned aerial vehicle in the specific embodiment of the present application. DETAILED DESCRIPTION

[0066] The embodiments of the present application will be described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0067] The present application provides a multi-point distribution unmanned aerial vehicle path planning method, comprising the following steps:

[0068] Step 1: Select the distribution environment, confirm the transportation warehouse, distribution point and obstacle information;

[0069] Step 2: Discretize the delivery environment and set the obstacle function;

[0070] Step 3: Build a path search model, set the objective function and select the constraint conditions;

[0071] Step 4: Use a double-layer genetic-A* optimization algorithm to plan the flight path combined with sequence and path, and perform path quality evaluation.

[0072] The following is further described in combination with specific implementation steps:

[0073] In step 1, considering the three-dimensional environment, a transportation warehouse coordinate is , there are delivery points , , the coordinates of the delivery point are , there are obstacles, . The purpose is to find a path , starting from the warehouse , visiting all demand points in turn, and finally returning to the warehouse , so that the unmanned aerial vehicle meets the flight height constraint and obstacle avoidance while the total flight path is the shortest.

[0074] The execution process of step 2 includes:

[0075] 2.1 Space discretization model

[0076] Discretize the continuous urban airspace into a three-dimensional grid space:

[0077]

[0078] In equation (1), is the position of each grid element, and are the number of grids in three directions.

[0079] 2.2 Obstacle function

[0080] Define the obstacle indicator function:

[0081]

[0082] In equation (2), is 1 indicating there is an obstacle at the position.

[0083] The path search model is constructed in Step 3, and the state space and action space are determined by combining reinforcement learning.

[0084] 3.1 State space

[0085] The state is defined as a three-dimensional grid coordinate:

[0086] The position of the UAV , is the th waypoint on the path.

[0087] 3.2 Action space

[0088] The allowed actions of the UAV include 26 directions of movement:

[0089]

[0090] The action space is defined as 26 discrete movement vectors in formula (3), each vector has components taking values of -1, 0 or 1, representing the relative displacement to the neighbor nodes in the three-dimensional grid.

[0091] 3.3 The model is expressed as follows:

[0092] Objective:

[0093]

[0094] Constraints:

[0095]

[0096]

[0097]

[0098]

[0099]

[0100] The objective function in formula (4) is to minimize the total length of the path , find a path such that the total length of the path is the shortest, is the solution to be found, i.e. the entire path, where is the th waypoint on the path. In formula (5), is an indicator function that is 1 when the position is occupied by an obstacle, and 0 otherwise; is an indicator function that is 1 when the position is safe to pass through, and 0 otherwise. The equation (6) means that every point on the path must be located in free space and must not intersect with any obstacle. In the equation (6), are the minimum and maximum height allowed for the UAV to fly. In the equation (7), is the nearest distance from the path point to any building in the set of all buildings , and is the minimum safety distance required to be kept. Every point on the flight path must keep at least meters distance from all buildings. The equation (8) ensures that the flight path of the UAV covers all the delivery demand points completely and visits every specified location without omission. The equation (9) indicates that both the start point and the end point of the UAV path are the warehouse .

[0101] For the three-dimensional path planning problem under the complex constraints in step 4, the present application proposes a double-layer genetic-A* optimization algorithm combining genetic algorithm and A* search with a hierarchical optimization strategy. First, the visiting sequence of multiple target points is optimized, and then a safe and efficient three-dimensional collision-free trajectory is planned for each path in the sequence in real time, finally the continuous distribution task of multiple target points is completed. The steps of the algorithm are as follows:

[0102] First layer: sequence optimization layer. Genetic algorithm is used to optimize the visiting sequence of demand points, and the optimal visiting sequence arrangement of demand points is globally searched. This layer determines the optimal sequence of the UAV visiting each distribution point.

[0103] Individual (chromosome) coding: use a chromosome (such as [0, 3, 1, 2, 4, 0]) to represent a visiting sequence, where the numbers represent the index of the demand point. The start point and end point are the warehouse 0.

[0104] Population initialization: randomly generate multiple visiting sequences (individuals) to form an initial population.

[0105] Fitness function: the fitness value of an individual is equal to the reciprocal of the total length of the complete three-dimensional path obtained after sending the sequence to the second layer for path planning (i.e. the shorter the path, the higher the fitness).

[0106] ​Genetic manipulation: selection, crossover, and mutation operations on chromosome sequences.

[0107] Termination condition: The number of iterations reaches the maximum.

[0108] Second layer: Path optimization layer. For each access sequence from the first layer, calculate the total distance of the corresponding three-dimensional fine path.

[0109] Input: An individual in the first layer (i.e., a specific access sequence, such as [0, 3, 1, 2, 4, 0]).

[0110] Processing: Calculate the 3D collision-free path for each path segment in sequence.

[0111] Output: The sum of the distances of all segmented paths in the sequence is returned to the first layer as the basis for evaluating the fitness of the individual.

[0112] like Figure 1 The diagram shows the algorithm flowchart of the two-layer genetic-A* optimization algorithm.

[0113] In step 4, a safety assessment of the UAV's flight trajectory is further conducted. This work sets path efficiency, safety indicators, and altitude change smoothness to evaluate the quality of the UAV's flight trajectory.

[0114] Path efficiency:

[0115]

[0116] Safety indicators:

[0117]

[0118] Smoothness of height variation:

[0119]

[0120] Equation (10) defines the path efficiency, i.e. the ratio of the straight-line distance between the start point and the end point to the total length of the actual flight path; the closer the ratio is to 1, the less redundant the path is and the higher the efficiency is. Equation (11) is used to evaluate the safety of the path, and the calculation result is the minimum distance between the UAV and all obstacles in the entire flight path; the larger the value is, the more sufficient the safety margin is, and the safer the flight is. Equation (12) quantifies the stability of the flight, and reflects the degree of vertical jolt by calculating the average value of the height change between adjacent path points; the smaller the value is, the more gentle the height change is, and the more stable the flight trajectory is.

[0121] Further, the present application also proposes specific embodiments for auxiliary description:

[0122] In the embodiment, a 2000m×2000m×200m three-dimensional urban airspace simulation environment is constructed, and various obstacle structures including a central building group, four corner high-rise buildings, diagonal lines and corridor-style buildings are set in the space, and a plurality of medium and small buildings are randomly distributed to simulate a real urban environment. The path planning grid adopts a resolution of 120×120×60, the horizontal resolution is 16.67m / grid, and the height resolution is 3.33m / grid. The safe flight height of the UAV is limited between 50m and 150m, and a minimum safety distance of 5m is provided to avoid obstacles. The experiment takes a warehouse located at (80, 100, 60) as the starting point, and needs to visit five distribution points in sequence, distribution point 1: (500, 500, 80), distribution point 2: (1200, 800, 90), distribution point 3: (800, 1500, 85), distribution point 4: (1500, 400, 95), and distribution point 5: (400, 1600, 75). The first layer genetic algorithm parameters are as follows: population size 50, iteration 100 generations, and mutation rate 0.1. The flight sequence of the UAV to the distribution points obtained by optimization solution is: 0→1→4→2→3→5→0. The total path distance is 5098.91 meters. Figure 2 The flight trajectory of the UAV at 16 angles is shown. Figure 3 The flight trajectory from the starting point 0 to the distribution point 1 is shown, Figure 4 The flight trajectory from the distribution point 1 to the distribution point 4 is shown, Figure 5 The flight trajectory from the distribution point 4 to the distribution point 2 is shown, Figure 6 The flight trajectory from the distribution point 2 to the distribution point 3 is shown, Figure 7 The flight trajectory from the distribution point 3 to the distribution point 5 is shown, Figure 8 The flight trajectory from the distribution point 5 to the starting point 0 is shown. Figure 9The elevation profile of the flight path of the unmanned aerial vehicle is shown, which clearly reflects the movement strategy and height distribution in the vertical direction, and different color segments in the figure distinguish the flight segments connecting different delivery points.

[0123] The experimental results show that the proposed path planning algorithm has good comprehensive performance in complex urban environment. The total length of the planned flight path is 5100.38 meters, and the total length of the straight line distance between each point is 3304.14 meters, and the path efficiency is 0.648. Under the premise of ensuring safety and obstacle avoidance, the reasonable path efficiency is realized. In terms of flight height control, the average flight height is 78.08 meters, which meets the safety flight limit of 50-150 meters, and there is no height violation. The average height change is 3.70 meters, which shows that the path smoothness is good. In terms of safety, the path successfully avoids all obstacles without any collision points, and the average safety distance is 65.25 meters, and the minimum safety distance is 8.65 meters, which are higher than the set safety threshold, proving the reliable obstacle avoidance ability of the algorithm in complex environment. The average direction change is 39.50°, which reflects that the path turning is relatively smooth, which is conducive to the stable flight of the unmanned aerial vehicle. All flight segments do not fly above the building, which further ensures the flight safety.

[0124] In summary, the present application decouples the problem into "global sequence optimization" and "local three-dimensional path planning" two levels, and innovatively introduces the second layer A* algorithm for fine processing. This method can determine the customer point access sequence with the highest global efficiency in complex urban environment, and generate a safe, efficient and flight constraint-compliant three-dimensional collision-free trajectory for each path in this sequence. Finally, this realizes the perfect unity of global path optimization and local obstacle avoidance safety, greatly improves the delivery efficiency, significantly reduces the time and energy cost, and provides key technical support for large-scale commercial application of unmanned aerial vehicle logistics.

[0125] The above disclosure is only one preferred embodiment of the present application, of course, cannot limit the scope of the rights of the present application, and those skilled in the art can understand that the above-mentioned embodiment can be implemented in whole or in part, and the equivalent changes made according to the claims of the present application still belong to the scope covered by the present application.

Claims

1. A method for path planning of a multi-drop delivery drone, characterized in that, The method comprises the following steps: Step 1: Select the delivery environment, confirm the transportation warehouse, delivery point and obstacle information; Step 2: Discretize the delivery environment and set the obstacle function; Step 3: Build a path search model, set the objective function and select the constraint condition; Step 4: Plan the flight path by using the double-layer genetic-A* optimization algorithm combined with the sequence and the path, and perform path quality evaluation.

2. The multi-point delivery UAV path planning method according to claim 1, wherein In the delivery environment, the coordinates of the transport warehouse are , there are delivery points , coordinates of the delivery point are represented, obstacles, , the delivery path , starting from the transport warehouse , sequentially visits all the demand points , and finally returns to the transport warehouse .

3. The multi-point delivery UAV path planning method according to claim 2, wherein The expression for discretizing the delivery environment in step 2 is as follows: ； wherein, is the position of each grid cell, and is the number of grids in three directions; The expression for the obstacle function is as follows: ； is 1 an obstacle is present at the location.

4. The multi-point delivery UAV path planning method according to claim 3, wherein In step 3, the state space is set up, and the state is defined as three-dimensional grid coordinates, that is, the position of the UAV. , It is the first on the path One route point; The action space is set, and the actions allowed by the UAV include movement in 26 directions, and the expression is as follows: ； where the action space is defined as 26 discrete movement vectors, each vector with components taking values -1, 0 or 1, representing relative displacement to a neighbor node in a three-dimensional grid.

5. The multi-point delivery UAV path planning method according to claim 4, wherein The objective function in step 3 is to minimize the total path length , which is expressed as follows: ； wherein a path is sought such that the total length of the path is minimal, for the solution sought, i.e. the entire path, is the i th waypoint on the path; The constraint condition is as follows: ； ； ； ； ； in, It is an indicator function, when the position When occupied by obstacles, When position When it is safe to pass, , indicating path Every point on All must be located in free space and must not intersect with any obstacles; the minimum and maximum altitude allowed for the drone to fly; To calculate path points Gather all buildings any building The closest distance, To maintain the required minimum safe distance, every point along the flight path must maintain a minimum distance from all buildings. A distance of meters; , indicates that the flight path of the UAV completely covers all the logistics demand points and accesses each designated location without omission; The start and end points of the drone path are both warehouses .​​ 6. The multi-point delivery UAV path planning method according to claim 5, wherein In the execution process of step 4, the visiting sequence of multiple target points is optimized first, and then a three-dimensional collision-free trajectory is planned for each path in the sequence in real time; The execution steps of the sequence optimization layer are as follows: Individual / chromosome coding; Each coded individual corresponds to a visiting sequence, and multiple visiting sequences are randomly generated to form an initial population; The individual visiting sequence is input into the path optimization layer to solve the fitness function, and the shorter the path, the higher the fitness; The chromosome sequence is subjected to selection, crossover and mutation operations to form a new sequence population; After the number of iterations reaches the maximum, the process is terminated, otherwise the process returns to the first step of the sequence optimization layer and starts execution; The execution steps of the path optimization layer are as follows: The visiting sequence from the sequence optimization layer is input; According to the sequence order, the three-dimensional collision-free path of each path is calculated in turn; The distances of all segmented paths in the sequence are added up and returned to the sequence optimization layer as the basis for fitness evaluation of the individual.

7. The multi-point delivery UAV path planning method according to claim 6, wherein The path quality evaluation indicators are as follows: Path efficiency: ； The ratio of the straight-line distance between the starting point and the end point to the total length of the actual flight path; the closer the ratio is to 1, the less redundant the path is and the higher the efficiency is; Safety index: ； The calculation result is the minimum distance between the UAV and all obstacles in the entire flight path; the larger the value is, the more sufficient the safety margin is and the safer the flight is; Smoothness of height change: ； The average value of the height change between adjacent path points is calculated to reflect the degree of vertical jolt; the smaller the value is, the more gentle the height change is and the more stable the flight trajectory is.

Images