A real-time trajectory planning method for multi-destination UAVs with intelligent decision-making

Abstract

The invention relates to a real-time track planning method for a multi-destination unmanned aerial vehicle with intelligent decision-making. The invention effectively solves the problem that the existing decision-making method cannot realize track planning under optimal conditions. The technical solution comprises: initializing the unmanned aerial vehicle, the destination and the environmental information; secondly, using the ant colony algorithm to establish a multi-destination decision model with minimum loss as the objective function, so as to effectively realize the autonomous decision-making of the unmanned aerial vehicle; then, using the rolling horizon method based on the extended solution to realize the step-by-step and real-time track planning in a complex environment; finally, adopting the strategy of making decisions while walking to complete the flight mission; the scheme adds an intelligent decision-making mechanism to the rolling horizon method (RHC_eS) based on the extended solution, realizes the autonomous decision-making and real-time track planning of the unmanned aerial vehicle, so as to have the real-time planning capability, faster response speed and better path information feedback capability, and is superior to other decision-making and track planning methods.

Description

Technical Field

[0001] The present invention relates to the technical field of track planning, and in particular to a real-time track planning method for multi-destination unmanned aerial vehicles with intelligent decision-making. Background Art

[0002] The UAV multi-destination material transportation mission in the wartime background has caused great difficulties due to its tense and complex environment and changing material demand. At present, domestic and foreign scholars have proposed many methods for the multi-destination material transportation problem in the wartime environment. However, these methods only make decisions on the destination based on the current environment and demand information when solving the material transportation mission in the wartime environment. In real situations, information such as the environment and demand are constantly changing. Therefore, it is difficult to actively, flexibly and in real time respond to the ever-changing battlefield by only using historical information to make a decision.

[0003] Multi-destination route planning belongs to the small traveling salesman problem, which is the most representative combinatorial optimization problem in graph theory and has been proven to be an NP problem. When considering the traveling salesman problem, most existing algorithms choose to use the path length as the basis for destination decision-making. However, in actual environments, it is necessary to consider multiple influencing factors based on tasks and goals. In addition, most algorithms generally separate decision-making and route planning tasks, and there is little research on real-time decision-making and route planning.

[0004] In view of the above, we provide a multi-destination UAV real-time trajectory planning method with intelligent decision-making to solve the above problems. Summary of the invention

[0005] In view of the above situation, the present invention provides a real-time trajectory planning method for multi-destination UAVs with intelligent decision-making. The scheme adds an intelligent decision-making mechanism to the rolling hill method based on extended solution (RHC_eS), realizes the autonomous decision-making and real-time trajectory planning of the UAV, and has real-time planning capability, faster response speed and better path information feedback capability compared with the existing UAV trajectory planning method.

[0006] A real-time trajectory planning method for multi-destination unmanned aerial vehicles with intelligent decision-making, characterized by comprising the following steps:

[0007] S1: Initialize the drone, destination and environment information;

[0008] S2: Construct a multi-destination intelligent decision-making mechanism;

[0009] S3: Select the best destination;

[0010] S4: Use RHC_eS for trajectory planning;

[0011] S5: Update the drone, destination and environment information;

[0012] S6: Adopt a decision-making-while-walking strategy and execute S2-S5 in a loop until reaching the final destination.

[0013] The above technical solution has the following beneficial effects:

[0014] (1) This scheme adds an intelligent decision-making mechanism to the extended solution-based rolling-horizon method (RHC_eS), realizing autonomous decision-making and real-time trajectory planning of the UAV, thus having real-time planning capability, faster response speed, and better path information feedback capability;

[0015] (2) Compared with other existing UAV trajectory planning methods, the trajectory planning method proposed in this scheme makes its trajectory length almost the same as the planning method with the shortest route as the objective function while achieving the lowest overall loss, thereby achieving the multi-destination trajectory planning task of the UAV under the best conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the real-time trajectory planning method for multi-destination unmanned aerial vehicles with intelligent decision-making of the present invention;

[0017] Figure 2 This is a multi-destination intelligent decision-making flow chart based on the ant colony algorithm of the present invention;

[0018] Figure 3 This is a flow chart of the trajectory planning of the rolling time domain method based on the extended solution of the present invention;

[0019] Figure 4 It is a schematic diagram of the intelligent decision-making and track planning results of multiple destinations (5A, 5B) of the present invention;

[0020] Figure 5 It is a schematic diagram of the intelligent decision-making and track planning results of multiple destinations (5C, 5D) of the present invention;

[0021] Figure 6 It is a schematic diagram of the intelligent decision-making and track planning results of multiple destinations (7A, 7B) of the present invention;

[0022] Figure 7 It is a schematic diagram of the intelligent decision-making and track planning results of multiple destinations (7C, 7D) of the present invention. DETAILED DESCRIPTION

[0023] The above and other technical contents, features and effects of the present invention are described in detail below with reference to the attached Figures 1 to 7 It can be clearly presented in the detailed description of the embodiments that the structural contents mentioned in the following embodiments are all referenced to the drawings in the specification.

[0024] This scheme proposes a real-time trajectory planning method for multi-destination UAVs with intelligent decision-making. First, the UAV, destination, and environment information are initialized;

[0025] Secondly, using the ant colony algorithm, a multi-destination decision-making model with minimum loss as the objective function is established to effectively realize the autonomous decision-making of drones;

[0026] Then, the receding horizon method based on the extended solution is used to realize step-by-step and real-time trajectory planning in complex environments;

[0027] Finally, adopt a decision-making-on-the-go strategy to complete the flight mission.

[0028] 1. The technical solution of the present invention is further described in detail below with reference to the accompanying drawings.

[0029] As attached Figure 1 As shown, it is a flow chart of a real-time trajectory planning method for multi-destination UAVs with intelligent decision-making, which includes the following steps:

[0030] Step 1: Initialize the drone, destination, and environment information

[0031] It includes the acquisition of the initial position of the drone, the location of the destination, the material reserves and material consumption speed range of each destination, and the location information of obstacles in the environment. The obstacles are 40 static circular obstacles with different radii that are randomly generated (except the starting point and the destination) in a 300×300 map environment. Overlapping of obstacles is allowed to simulate obstacle environments of different shapes.

[0032] Step 2: Based on the ant colony algorithm, construct a multi-destination intelligent decision-making mechanism, as shown in the attached figure. Figure 2 As shown in the figure, it is a multi-destination intelligent decision-making flowchart based on ant colony algorithm, which includes the following processes:

[0033] (1) Initialize the parameters involved in the ant colony algorithm, such as pheromones and the number of ant colonies.

[0034] (2) Set the ants to start from a fixed starting point.

[0035] (3) The formula for calculating the transfer probability between destinations is as follows:

[0036]

[0037] Among them, α is the importance of the pheromone factor, γ is the importance of the heuristic factor, and τ ij (t) represents the pheromone concentration of destination i and j at time t, η ij is the heuristic function, J c(i) is the set of cities that ant c is allowed to choose in the next step. The heuristic function formula is as follows:

[0038]

[0039] Where, d ij represents the distance between destinations i and j, and the pheromone update formula in the environment is as follows:

[0040]

[0041] Among them, ρ is the pheromone volatility factor, It represents the pheromone concentration left by ant c between destinations i and j. The calculation formula is as follows:

[0042]

[0043] Among them, Q represents a constant coefficient, M c It represents the total loss of ant c after reaching all destinations.

[0044] (4) Use roulette to select the next destination.

[0045] (5) When the ants have reached all destinations, create an objective function and calculate the minimum loss. The objective function for the total material loss prediction of each destination is as follows:

[0046]

[0047] Among them, the total number of destinations is n, S i is the strategic importance of destination i, which is scored by experts using a nine-point system in the hierarchical analysis method. i is the loss at destination i, and its calculation formula is as follows:

[0048]

[0049] In the formula, q i It represents the remaining amount of materials at destination i, and its calculation formula is as follows:

[0050]

[0051] In the formula, G i and They represent the quantity of materials and the average consumption rate of materials at destination i, respectively. is the estimated total time required for the drone to fly from the starting position to the destination i, and the formula for the average material loss rate is as follows:

[0052]

[0053] Among them, V g_iminand V g_imax They represent the minimum and maximum speeds of material consumption at destination i respectively.

[0054] In order to calculate the estimated time for the drone to arrive at each destination, a material delivery time matrix A2 is established, and the formula is as follows:

[0055]

[0056] Where V max and V min are the maximum and minimum speeds of the drone, respectively. A1 represents the shortest track matrix. In order to reduce the amount of calculation and ensure real-time requirements, the distance between the two destinations uses the Euclidean distance. The formula is as follows:

[0057]

[0058] Where n is the number of destinations (excluding the origin), L i,j represents the distance from destination i to destination j, and when i=j, L i,j =0. Since the Euclidean distance is used, the round-trip distance between the two destinations is the same, that is, the shortest track matrix is ​​a symmetric matrix. The material transportation time matrix A2 is established based on the shortest track matrix and expanded as follows:

[0059]

[0060] Among them, t represents the average flight time of the drone from destination i to destination j, and the material delivery time matrix is ​​also a symmetric matrix.

[0061] (6) Record the destination sequence corresponding to the minimum loss and update the pheromone.

[0062] (7) Repeat (2)-(6) until the number of iterations is reached, and output the minimum loss and the corresponding destination sequence.

[0063] Step 3: According to the destination sequence obtained in step 2, select the best destination as the next destination to be flown to.

[0064] Step 4: Use the extended solution-based rolling horizon method to achieve step-by-step, real-time trajectory planning in complex environments, as shown in the attached figure. Figure 3 As shown in the figure, this is the flow chart of the trajectory planning of the rolling time domain method based on the extended solution, which includes the following processes:

[0065] (1) Construct a path expansion mechanism to generate a reference track in the rolling time window in the target direction (when θ = 0° in equation (12)). The formula is as follows.

[0066]

[0067] In the formula, θ is the angle between the target direction and the flight direction, β is the angle between the target direction and the X-axis direction. When the flight direction is between the target direction and the positive direction of the X-axis, the angle is β-θ. When the flight direction is between the target direction and the positive direction of the Y-axis, the angle is β+θ. max Indicates the maximum flight distance of the drone in one second, T w is a rolling time window.

[0068] (2)Use Determine whether the trajectory generated in (1) is feasible. uav is the wingspan of the drone, and the radius of the obstacle is r d , the position is (x d ,y d ).

[0069] (3) When the target direction route is not feasible, the path expansion mechanism is used to generate multi-directional (θ = 30°, 60°, 90°) reference tracks.

[0070] (4) Perform average crossover and single-point crossover operations on the reference track. The average crossover formula is as follows:

[0071]

[0072] Where (f(x,y)) m represents the mth reference track. The number of reference tracks is consistent with the number of search directions. That is, a reference track is generated in each search direction. Then, f(x, y) and f ac (x,y) performs a single-point crossover, and the formula is defined as follows:

[0073] f sc (x,y)={f(x i ,y i ),f ac} (14)

[0074] Note that the {·} operation is used to represent a single-point intersection of two routes.

[0075] (5) The route f(x,y), f ac (x,y), and f sc After (x, y) is integrated, the directional mutation operation is performed together. The process is as follows: first, the mutation probability pm is used to determine whether to mutate. Then, for the route that needs to be mutated, a mutation segment is randomly selected and a direction is randomly selected from the two directions corresponding to θ = 30° and the target direction corresponding to θ = 0° at the segment. Then, according to the path expansion mechanism, the reference track is regenerated in the selected direction on the mutation segment to obtain the mutated new route f dv (x,y).

[0076] (6) After the expansion, crossover and mutation of all routes, the total number of predicted routes is as follows:

[0077] f z (x,y)=[f(x,y),f ac (x,y),f sc (x,y),f dv (x,y)] (15)

[0078] (7) Using the quadratic Bezier curve to smooth the route, we get:

[0079] f z (x(k),y(k))=(1-k) 2 (x0,y0)+2(1-k)k(x1,y1)+k 2 (x2,y2)k∈[0,1] (16)

[0080] Among them, k is the position parameter, (x0, y0), (x1, y1), (x2, y2) represent the coordinates of the first, middle and end control points of the quadratic Bezier curve respectively, and f z The track points in (x,y) serve as control points of the quadratic Bezier curve.

[0081] (8) Construct the shortest route objective function and use the sequential quadratic programming method to find the shortest route F(X,Y) * And the corresponding optimal route segment f in each time window sz (x,y) * , and finally integrate it into the optimal route f flight The shortest path objective function is as follows:

[0082]

[0083] Among them, ||f sz (x,y)|| represents the total length of the smoothed predicted route, L f is the straight-line distance from the last predicted waypoint to the target point, v max is the maximum flight speed of the drone, r fli g h t represents the radius of curvature of the channel, r uav It represents the minimum turning radius of the UAV. Since the smooth predicted path is composed of curves, its length is difficult to calculate accurately. Therefore, the length of the curve is estimated by using the broken line length. The smooth predicted route length is estimated as shown in the following formula.

[0084]

[0085] Note that when b = w, (xw ,y w ) represents the end point of the smoothed predicted route. The distance from the final destination of the drone to the last waypoint is calculated as follows:

[0086]

[0087] Among them, (x f ,y f ) are the end point coordinates.

[0088] Step 5: Update drone, destination and environment information.

[0089] Step 6: Adopt a decision-making-while-walking strategy and repeat steps 2 to 5 until you reach the final destination.

[0090] 2. The effectiveness and correctness of this solution are demonstrated in combination with specific embodiments. Experiments are conducted in environments with different numbers of destinations. The environment settings are shown in Table 1:

[0091]

[0092] Table 1 Multi-destination parameter information

[0093] Since the parameters of each destination involve the amount of materials, speed, etc., there are too many parameters. In order to facilitate parameter setting and compare the results with different numbers of destinations, when setting the relevant parameters of multiple destinations, 3, 5, and 7 destinations are used in an incremental and inclusive manner. For example, 5 destinations are added on the basis of all parameters of 3 destinations remaining unchanged, and only the parameters of the newly added destinations are set. Therefore, Table 1 only shows the parameter setting results of 7 destinations. The parameters of 3 destinations and 5 destinations are taken from the parameter settings of the first 3 and first 5 destinations. The comparison method is shown in Table 2:

[0094] A Comparison method: Genetic algorithm with real-time decision making and minimum loss as the objective function. B Comparison method: An ant colony algorithm with minimum loss as the objective function and only starting decision-making once. C Comparison method: A genetic algorithm with the shortest route as the objective function and only one start decision. D The method in this paper: Ant colony algorithm with minimum loss as the objective function and real-time decision-making.

[0095] Table 2 Description of this method and the comparative method

[0096] The computer hardware configuration used in the experiment is: a computer with 4GB memory and Windows 10 operating system. The simulation is programmed using MATLAB 2018a software. In order to eliminate the influence of random factors and other accidental factors on the algorithm, the above algorithm is executed independently 30 times. The experimental results are shown in Table 3;

[0097] Attached Figure 4 This is a schematic diagram of the intelligent decision-making and trajectory planning structure obtained by using the 5A and 5B planning methods for multiple destinations (5) in this plan;

[0098] Attached Figure 5This is a schematic diagram of the intelligent decision-making and trajectory planning structure obtained by using the 5C and 5D planning methods for multiple destinations (5) in this plan;

[0099] Attached Figure 6 This is a schematic diagram of the intelligent decision-making and trajectory planning structure obtained by using 7A and 7B planning methods for multiple destinations (7) in this plan;

[0100] Attached Figure 7 This is a schematic diagram of the intelligent decision-making and trajectory planning structure obtained by using the 7C and 7D planning methods for multiple destinations (7) in this plan;

[0101] 3 destinations Destination Order Track length Total loss 3A 123 672.4919 672 3B 132 736.4451 548 3C 213 568.7762 494 3D 312 581.3103 368 5 destinations Destination Order Track length Total loss 5A 41235 867.5825 661 5B 41325 965.6638 499 5C 54213 717.4418 599 5D 42135 733.1881 317 7 destinations Destination Order Track length Total loss 7A 4126735 1040.09 983 7B 4126735 1040.09 983 7C 5426713 928.0233 903 7D 4267135 943.7696 609

[0102] Table 3 Results of multi-destination intelligent decision-making and trajectory planning

[0103] It can be clearly seen from Table 3 above that compared with other methods, the method proposed in this scheme performs better in terms of overall loss, and the track length of this scheme is not much different from the method with the shortest route as the objective function (the two are close in data), so the above experiments prove the effectiveness and correctness of the method proposed in this scheme.

[0104] The above is only for illustrating the present invention. It should be understood that the present invention is not limited to the above embodiments, and various variations that conform to the concept of the present invention are within the protection scope of the present invention.

Claims

1. A real-time trajectory planning method for multi-destination UAVs with intelligent decision-making, characterized in that: The following steps are involved: S1: Initialize the drone, destination and environment information; S2: Construct a multi-destination intelligent decision-making mechanism; The construction process of the multi-destination intelligent decision-making mechanism in S2 is as follows: S2-1: Initialize the parameters such as pheromones and ant colony number involved in the ant colony algorithm; S2-2: Set the ants to start from a fixed starting point; S2-3: The formula for calculating the transfer probability between destinations is as follows: Among them, α is the importance of the pheromone factor, γ is the importance of the heuristic factor, and τ ij (t) represents the pheromone concentration of destination i and j at time t, η ij is the heuristic function, J c (i) is the set of cities that ant c is allowed to choose in the next step. The heuristic function formula is as follows: Where, d ij represents the distance between destinations i and j, and the pheromone update formula in the environment is as follows: Among them, ρ is the pheromone volatility factor, It represents the pheromone concentration left by ant c between destinations i and j. The calculation formula is as follows: Among them, Q represents a constant coefficient, M c It represents the total loss of ant c after reaching all destinations; S2-4: Use roulette method to select the next destination; S2-5: When the ant has reached all destinations, create an objective function and calculate the minimum loss. The objective function for multi-destination prediction loss is as follows: Among them, the total number of destinations is n, S i is the strategic importance of destination i, which is scored by experts using a nine-point system in the hierarchical analysis method. i is the loss at destination i, and its calculation formula is as follows: In the formula, q i It represents the remaining amount of materials at destination i, and its calculation formula is as follows: In the formula, G i and They represent the quantity of materials and the average consumption rate of materials at destination i, respectively. is the estimated total time required for the drone to fly from the starting position to the destination i, and the formula for the average material loss rate is as follows: Among them, V g_imin and V g_imax They represent the minimum and maximum speeds of material consumption at destination i respectively; In order to calculate the estimated time for the drone to arrive at each destination, a material delivery time matrix A2 is established, and the formula is as follows: Where V max and V min are the maximum and minimum speeds of the drone, respectively. A1 represents the shortest track matrix. In order to reduce the amount of calculation and ensure real-time requirements, the distance between the two destinations uses the Euclidean distance. The formula is as follows: Where n is the number of destinations excluding the origin, L i,j represents the Euclidean distance from destination i to destination j, and when i=j, L i,j =0. Since the Euclidean distance is used, the round-trip distance between the two destinations is the same, that is, the shortest track matrix is ​​a symmetric matrix. The material transportation time matrix A2 is established based on the shortest track matrix and expanded as follows: Among them, t represents the average flight time of the drone from destination i to destination j, and the material delivery time matrix is ​​also a symmetric matrix; S2-6: Record the destination sequence corresponding to the minimum loss and update the pheromone; S2-7: Repeat S2-2-S2-6 until the number of iterations is reached, and output the minimum loss and the corresponding destination sequence; S3: Select the best destination; S4: Use the extended solution-based rolling-horizon method RHC_eS for trajectory planning; S5: Update the drone, destination and environment information; S6: Adopt a decision-making-while-walking strategy and repeat steps 2 to 5 until reaching the final destination.

2. A multi-destination UAV real-time trajectory planning method with intelligent decision-making according to claim 1, characterized in that: The S1 includes obtaining the initial position of the UAV, the position of the destination, the material storage and material consumption speed range of each destination, and the position information of obstacles in the environment.

3. The method for real-time trajectory planning of a multi-destination UAV with intelligent decision-making according to claim 1 is characterized in that: The process of using the extended solution-based rolling-horizon method RHC_eS to perform trajectory planning in S4 is as follows: S4-1: Construct a path expansion mechanism to generate the target direction, that is, the reference track within the rolling time window when θ = 0° in equation (12), as follows: In the formula, θ is the angle between the target direction and the flight direction, β is the angle between the target direction and the X-axis direction. When the flight direction is between the target direction and the positive direction of the X-axis, the angle is β-θ. When the flight direction is between the target direction and the positive direction of the Y-axis, the angle is β+θ. max Indicates the maximum flight distance of the drone in one second, T w is a rolling time window; S4-2: Exploitation Determine the generated flight path (x) in step S4-1 d ,y d ) is feasible, where L uav is the wingspan of the drone, and the radius of the obstacle is r d , the location is; S4-3: When the target direction route is not feasible, the path expansion mechanism is used to generate multi-directional θ = 30°, 60°, 90° reference tracks; S4-4: Perform average crossover and single-point crossover operations on the reference track. The average crossover formula is as follows: Where (f(x,y)) m represents the mth reference track. The number of reference tracks is consistent with the number of search directions. That is, a reference track is generated in each search direction. Then, f(x, y) and f ac (x,y) performs a single-point crossover, and the formula is defined as follows: f sc (x,y)={f(x i ,y i ),f ac } (14) Note that the {·} operation is used to represent a single-point intersection of two routes; S4-5: Set the route f(x,y), f ac (x,y), and f sc After (x, y) is integrated, the directional mutation operation is performed together. The process is as follows: first, the mutation probability pm is used to determine whether to mutate. Then, for the route that needs to be mutated, a mutation segment is randomly selected and a direction is randomly selected from the two directions corresponding to θ = 30° and the target direction corresponding to θ = 0° at the segment. Then, according to the path expansion mechanism, the reference track is regenerated in the selected direction on the mutation segment to obtain the mutated new route f dv (x,y); S4-6: Integrate all routes after path expansion, intersection and mutation to obtain the following predicted route group: f z (x,y)=[f(x,y),f ac (x,y),f sc (x,y),f dv (x,y)] (15) S4-7: Use quadratic Bezier curve to smooth the route: f z (x(k),y(k))=(1-k) 2 (x0,y0)+2(1-k)k(x1,y1)+k 2 (x2,y2)k∈[0,1] (16) Among them, k is the position parameter, (x0, y0), (x1, y1), (x2, y2) represent the coordinates of the first, middle and end control points of the quadratic Bezier curve respectively, and f z The track points in (x,y) are used as control points of the quadratic Bezier curve; S4-8: Construct the shortest route objective function and use the sequential quadratic programming method to find the shortest route F(X,Y) * And the corresponding optimal route segment f in each time window sz (x,y) * , and finally integrate it into the optimal route f flight , the shortest route objective function is as follows: Among them, ||f sz (x,y)|| represents the total length of the smoothed predicted route, L f is the straight-line distance from the last predicted track point to the target point, v max is the maximum flight speed of the drone, r flight represents the radius of curvature of the channel, r uav It represents the minimum turning radius of the drone. Since the smooth predicted route is composed of curves, it is difficult to calculate the length accurately. Therefore, the length of the curve is estimated by the length of the broken line. The smooth predicted route length is estimated as shown in the following formula: Note that when b = w, (x w ,y w ) represents the end point of the smoothed predicted route. The distance from the final destination of the drone to the last track point is calculated as follows: Among them, (x f ,y f ) are the end point coordinates.

Images