Unmanned aerial vehicle local path planning method based on improved dynamic window method

Abstract

The invention discloses an unmanned aerial vehicle local path planning method based on an improved dynamic window method, and the method comprises the steps: S1, initializing the state, a target point and an environment map of an unmanned aerial vehicle, and creating a grid map of the environment; s2, a dynamic window optimization algorithm is improved, and the improvement points are as follows: (1) according to the obstacle density, a dynamic window parameter adaptive adjustment strategy is introduced, the speed and acceleration range of the unmanned aerial vehicle are dynamically adjusted, and meanwhile, the distance from the unmanned aerial vehicle to an end point is taken as an influence factor, so that the unmanned aerial vehicle automatically decelerates when approaching the end point; (2) setting a dynamic adjustment sampling resolution, and optimizing the number of candidate paths according to environmental complexity; (3) introducing a sparrow search algorithm for optimizing the index weight of a trajectory evaluation function; and S3, inputting the created path grid map into the improved dynamic window optimization algorithm, and outputting an optimal path. The method can be more effectively applied to local path planning of the unmanned aerial vehicle, and the safety and effectiveness of autonomous navigation of the unmanned aerial vehicle in a complex environment are improved.

Description

Technical Field

[0001] The present invention relates to a method for local path planning of an unmanned aerial vehicle (UAV) based on an improved dynamic window method. Background Art

[0002] Due to its characteristics of high speed, flexibility, and wide coverage, UAVs have shown important application values in military, civilian, and commercial fields. With the increasing demands for UAVs in aspects such as autonomous flight, swarm control, and path planning, having a flexible, reliable, and stable path planning ability is the key to realizing the autonomous navigation of UAVs. Currently, when UAVs encounter complex environmental challenges, problems such as getting trapped in local optima and unsatisfactory obstacle avoidance path effects are prone to occur. This is mainly because the local path optimization algorithm of UAVs is limited by the sampling rate, range search, and fixation of evaluation indicators in a dynamic environment and cannot adapt to a flexible and variable navigation environment with different degrees of dense obstacles. Summary of the Invention

[0003] The present invention provides a method for local path planning of an unmanned aerial vehicle based on an improved dynamic window method to solve the problems existing in the above-mentioned prior art. The present invention can be more effectively applied to the local path planning of UAVs and improve the safety and effectiveness of the autonomous navigation of UAVs in complex environments.

[0004] The technical solutions adopted by the present invention are as follows:

[0005] A method for local path planning of an unmanned aerial vehicle based on an improved dynamic window method, comprising the following steps:

[0006] S1: Initialize the state of the UAV, the target point, and the environmental map, and create a grid map of the environment;

[0007] S2: Improve the dynamic window optimization algorithm, and the improvement points are:

[0008] (1) According to the density of obstacles, introduce a dynamic window parameter adaptive adjustment strategy to dynamically adjust the speed and acceleration ranges of the UAV. At the same time, take the distance of the UAV from the end point as an influencing factor, so that the UAV automatically decelerates when approaching the end point;

[0009] (2) Set a dynamically adjustable sampling resolution and optimize the number of candidate paths according to the environmental complexity;

[0010] (3) Introduce a sparrow search algorithm to optimize the index weights of the trajectory evaluation function;

[0011] S3: Input the created path grid map into the improved dynamic window optimization algorithm and output the optimal path.

[0012] Further, in S1, the initial velocities of the drone on the X, Y, and Z axes are (0, 0, 0), and the drone is considered to have reached the target when it arrives within 1 m of the target point.

[0013] Further, for improvement point (1), a threshold for obstacle density is set. When the obstacle density is higher than this threshold, the speed and acceleration ranges are reduced to restrict the movement of the drone; when the obstacle density is lower than this threshold, the speed and acceleration ranges are expanded to improve the movement flexibility of the drone.

[0014] Further, the mathematical function for adaptively adjusting the dynamic window parameters is:

[0015]

[0016] where vs0 is the current dynamic window parameter of the drone, and vs i is the dynamic window parameter after adaptively adjusting the dynamic window parameter;

[0017] ρ od is the obstacle density within the specified range of the drone, ρ0 is the obstacle density threshold, and ρ0 = 0.5 is taken;

[0018] δ is the speed range reduction coefficient when the obstacle density is higher than the threshold, and δ = 0.5 is taken. When the obstacle density around the drone is greater than the obstacle density threshold, the dynamic window is adjusted;

[0019] d i is the distance of the drone from the target point; d0 is the threshold distance from the target point, and d0 = 20 is taken; ε is the speed range reduction coefficient when too close to the target, and ε = 0.7 is taken.

[0020] Further, for improvement point (2), the sampling resolution is defined as:

[0021]

[0022] where V is the speed sampling resolution; V0 is the default sampling resolution, and V0 = 0.1 is taken; ρ od is the obstacle density within the specified range of the drone; ρ0 is the obstacle density threshold.

[0023] Further, for improvement point (3), specifically:

[0024] The motion model of the drone during flight is defined as:

[0025]

[0026] where Δt is the adjacent time interval; x t 、y t and zt They are the abscissa, ordinate and altitude coordinate of the UAV at time t respectively; v t and v z They are the XY-axis plane velocity and Z-axis velocity of the UAV respectively; θ t and ω t They are the steering angle and steering angular velocity of the UAV respectively;

[0027] In the trajectory prediction after sampling, the trajectory and its retrograde are evaluated by using a trajectory evaluation function, and the speed corresponding to the optimal trajectory is selected; in the selection of the weights of the trajectory evaluation function in each iteration, the sparrow search algorithm is introduced, and the most suitable weight set for the current situation is optimized through the sparrow search algorithm, so as to calculate the optimal path of the UAV in the next time period.

[0028] Furthermore, the trajectory evaluation function is composed of four sub-indicators, namely the heading angle, the distance from obstacles, the speed and the smoothness. The trajectory evaluation function is defined as:

[0029] F = w1·heading(ν,ω) + w2·dist(ν,ω) + w3·velocity(ν,ω) + w4·smooth(ν,ω)

[0030] Where: heading(v,ω) is the heading angle index, dist(v,ω) is the index of the distance from obstacles, velocity(v,ω) is the linear velocity index of the predicted trajectory, and smooth(v,ω) is the smoothness index; w1, w2, w3 and w4 are the weight parameters of these four indicators respectively.

[0031] Furthermore, the most suitable weight set for the current situation is optimized through the sparrow search algorithm, so as to calculate the optimal path of the UAV in the next time period. The process is as follows:

[0032] (1) Determine the search range of the dynamic weights:

[0033] W = [w1, w2, w3, w4],

[0034] {w1 + w2 + w3 + w4 = 1 | 0.05 < w1 < 0.2, 0.1 < w2 < 0.4, 0.05 < w3 < 0.2, 0.05 < w4 < 0.2}

[0035] Where, w1, w2, w3 and w4 are the weights of the heading angle index, the distance from obstacles index, the speed and the smoothness index respectively, and W is the weight set;

[0036] (2) Initialize the sparrow population, which is used to randomly initialize the positions of the sparrow population. The position of each sparrow represents each weight combination. The specific formula is:

[0037] wi,j = lb j + (ub j - lb j )·rand(N, dim)

[0038] where w i,j is the value of the i-th sparrow's weight at the j-th weight; ub j and lb j are the upper and lower limits of the j-th weight respectively, both being a 1×dim vector representing the maximum and minimum values of each dimension; rand(N, dim) represents generating an N×dim random matrix with each element between [0, 1];

[0039] Calculate the fitness value for each weight combination, that is, the objective function value of each weight, and find the optimal weight combination. The specific formula is:

[0040] F i = w i,1 ·heading(ν, ω) + w i,2 ·dist(ν, ω) + w i,3 ·velocity(ν, ω) + w i,4 ·smooth(ν, ω)

[0041] (3) Update the weight set. Each sparrow moves closer to the position of the current optimal sparrow, and at the same time introduce randomness to maintain diversity. The update formula for the i-th sparrow at the t-th iteration is as follows:

[0042] w i t+1 = w i t + r × (w* - w i t )

[0043] where w it+1 is the updated weight; w it is the weight of the current iteration; w* is the currently found optimal weight, and r is a random number matrix, taking rand(N, dim), representing the random change amount of different weights;

[0044] To ensure that the position of the sparrow is within the search space range, it is necessary to limit the position of the sparrow. If the position of the sparrow exceeds the lower limit lb or the upper limit ub, then limit it within the boundary. The specific formula is:

[0045] w i t+1 = max(w i t+1 , lb)

[0046] w it+1 = min(w i t+1 , ub)

[0047] (4) Termination condition check: The iterative process includes a termination condition check, which is:

[0048] F* = min(F1, F2, ···, F N )

[0049] In each iteration, the sparrow with the smallest fitness value is found, and the algorithm stops iterating and outputs the optimal solution.

[0050] The present invention has the following beneficial effects:

[0051] (1) By introducing a dynamic window adaptive adjustment strategy, the movement speed and acceleration range are intelligently adjusted according to the density of surrounding obstacles, improving the algorithm's ability to adapt to complex and changeable environments.

[0052] (2) By improving the dynamic speed sampling strategy, the algorithm optimizes the number of candidate paths according to the environmental complexity, improving the accuracy of the optimal path and the calculation efficiency.

[0053] (3) By integrating the dynamic weight strategy of the sparrow search algorithm, the weights of the sub-indices of the trajectory evaluation function are dynamically adjusted according to environmental changes, improving the flexibility of path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the present invention.

[0055] Figure 2a and Figure 2b show the comparison results of the local path planning trajectories of the UAV in a complex dynamic environment. Among them, (a) is a top view and (b) is an oblique view,

[0056] Figure 3 is a comparison graph of the UAV's speed change curve. DETAILED DESCRIPTION OF THE INVENTION

[0057] The present invention will be further described below with reference to the accompanying drawings.

[0058] As Figure 1 shown, a method for local path planning of an unmanned aerial vehicle (UAV) based on an improved dynamic window method (SSA-DWA) of the present invention includes the following steps:

[0059] S1: Initialize the UAV state, target point, and environmental map. The initial speed of the UAV on the X, Y, and Z axes is (0, 0, 0). The UAV is considered to have reached the target when it is within 1 m of the target point, and a grid map of the environment is created.

[0060] S2: Improve the dynamic window optimization algorithm with the following improvement points:

[0061] (1) By introducing a dynamic window parameter adaptive adjustment strategy, specifically:

[0062] In a high-density obstacle environment, narrow the speed and acceleration ranges to restrict the movement of the UAV and improve safety;

[0063] In a low-density obstacle environment, expand the speed and acceleration ranges to improve the movement flexibility of the UAV and enhance the UAV path planning efficiency;

[0064] Take the distance of the UAV from the end point as an influencing factor to make the UAV automatically decelerate when approaching the end point.

[0065] (2) By setting a dynamically adjustable sampling resolution, the number of candidate paths can be optimized according to the environmental complexity: in a high-obstacle density environment, generate more candidate paths by increasing the sampling density to ensure finding a better path. In an obstacle-dense environment, reduce the number of candidate paths by decreasing the sampling density to reduce the algorithm's computational amount and improve the computational efficiency.

[0066] (3) Introduce the SSA algorithm and apply the SSA algorithm to optimize the trajectory evaluation function index weights in the DWA algorithm, and use the SSA algorithm to dynamically adjust according to the real-time requirements of the task to find the optimal weight combination.

[0067] S3: Input the path raster map created in S1 into the improved dynamic window optimization algorithm and output the optimal path.

[0068] Regarding improvement point 1:

[0069] Due to the self-constraints of the UAV during flight, the actual achievable longitudinal speed v and the steering angular velocity range ω within adjacent Δt intervals are restricted. The speed space value within a certain time window is defined as:

[0070]

[0071] Among them, v max and v min respectively represent the upper and lower limits of the UAV linear velocity value; ω max and ω min respectively represent the upper and lower limits of the UAV angular velocity value; v and ω respectively represent the maximum plus and minus linear velocity and the maximum steering plus and minus angular velocity of the UAV.

[0072] The heading angle of the UAV within a unit window time is:

[0073]

[0074] Among them, θ is the current heading angle of the UAV.

[0075] In order to enable the UAV to stop before hitting an obstacle when it senses danger, the speed range of the UAV is:

[0076]

[0077] Among them, d(v, ω) is the d nearest distance between the corresponding predicted trajectory and the UAV.

[0078] In order to enable the UAV to intelligently adjust its moving speed according to the surrounding environment to ensure efficient and safe flight, the present invention introduces a dynamic window parameter adaptive adjustment strategy, specifically:

[0079] Set the obstacle density threshold. When the obstacle density is higher than this threshold, narrow the speed and acceleration ranges to restrict the movement of the UAV; when the obstacle density is lower than this threshold, expand the speed and acceleration ranges to improve the movement flexibility of the UAV. At the same time, take the distance of the UAV from the end point as an influencing factor to make the UAV automatically decelerate when approaching the end point.

[0080] Define the speed adjustment range as:

[0081]

[0082] Among them, vs0 is the current dynamic window parameter of the UAV, and vs i is the dynamic window parameter after the dynamic window parameter adaptive adjustment strategy.

[0083] ρ od is the obstacle density within the specified range of the UAV, ρ0 is the obstacle density threshold, and take ρ0 = 0.5.

[0084] δ is the speed range reduction coefficient when the obstacle density is higher than the threshold, take δ = 0.5. When the obstacle density around the UAV is greater than the obstacle density threshold, the dynamic window is adjusted.

[0085] d i is the distance of the UAV from the target point, d0 is the threshold distance from the target point, and take d0 = 20.

[0086] ε is the speed range reduction coefficient when too close to the target, take ε = 0.7.

[0087] When the UAV is about to reach the target point, in order to ensure the safety of the UAV, it is necessary to adjust the dynamic window of the UAV.

[0088] Regarding improvement point 2:

[0089] Due to the presence of dynamic obstacles in the environment, the environment is variable. There are relatively more high-resolution speed samples, resulting in more generated candidate paths, while there are relatively fewer low-resolution speed samples, resulting in fewer generated candidate paths. To flexibly balance the accuracy and computational complexity of the optimal path, the dynamic speed sampling strategy is improved to adjust the speed sample density.

[0090] The sampling resolution adjustment formula proposed by the present invention is:

[0091]

[0092] where V is the speed sampling resolution, V0 is the default sampling resolution, and V0 = 0.1 is taken. ρ od is the obstacle density within the specified range of the UAV and can be adjusted according to the environmental complexity. ρ0 is the obstacle density threshold.

[0093] By dynamically adjusting the sampling resolution, the number of candidate paths can be optimized according to the environmental complexity: in a high-obstacle density environment, more candidate paths are generated by increasing the sampling density to ensure finding a better path. In an obstacle-dense environment, the number of candidate paths is reduced by decreasing the sampling density, reducing the computational complexity of the algorithm and improving the computational efficiency.

[0094] Regarding improvement point 3:

[0095] The motion model of the UAV during flight is defined as:

[0096]

[0097] where Δt is the adjacent time interval; x t , y t and z t are the abscissa, ordinate, and altitude coordinate of the UAV at time t, respectively; v t and v z are the XY-axis plane speed and Z-axis speed of the UAV, respectively; θ t and ω t are the turning angle and turning angular velocity of the UAV, respectively.

[0098] In the trajectory prediction after sampling, a trajectory evaluation function is needed to evaluate the trajectory and reverse, so as to select the speed corresponding to an optimal trajectory. To solve the problem that the traditional dynamic window method lacks consideration of smoothness in path planning, the present invention also introduces B-splines. On the basis that the trajectory evaluation function of the traditional dynamic window method consists of three sub-indicators (namely, the heading angle, the distance from obstacles, and the speed), B-splines are introduced into the trajectory evaluation function to add a smoothness indicator. By optimizing the control points of the B-spline, the smoothness of the path can be effectively improved, thereby reducing the situation of sharp turning. This not only improves the coherence of the path, but also enables the UAV to execute tasks more smoothly and accurately during obstacle avoidance, reduces the collision risk, and improves flight safety.

[0099] In the dynamic path planning of UAVs in complex environments, the trajectory evaluation function can help UAVs adjust the path in real time in a dynamically changing environment, avoid obstacles, avoid dangerous areas, evaluate and guide UAVs to select the optimal path, and ensure the efficient completion of flight tasks.

[0100] The B-spline evaluation index is introduced, and combined with the three indexes of the UAV's heading angle, the distance from the nearest obstacle, and the flight speed, a new objective evaluation function is constructed. The new objective evaluation function makes the path planning more accurate and flexible, and effectively improves the adaptability and task execution efficiency of UAVs in complex environments. The formula is as follows:

[0101] F = w1·heading(ν,ω) + w2·dist(ν,ω) + w3·velocity(ν,ω) + w4·smooth(ν,ω)

[0102] Where: heading(v,ω) is the heading angle index, dist(v,ω) is the index of the distance from obstacles, velocity(v,ω) is the linear velocity index of the predicted trajectory, smooth(v,ω) is the smoothness index, which is generated by the second derivative of the B-spline; w1, w2, w3, and w4 are the weight parameters of these four indexes respectively. Through the optimization and solution of the objective evaluation function, the UAV can achieve safe, fast, and accurate path planning in complex dynamic environments.

[0103] To enhance the ability of UAVs to adapt to the flexible and changeable task requirements in dynamic environments, the SSA algorithm (Sparrow Search Algorithm) is introduced in the weight selection of the trajectory function in each iteration to dynamically adjust different index weights according to environmental changes.

[0104] The specific principle for adjusting the evaluation index weights proposed by the present invention is as follows:

[0105] When the drone approaches an obstacle, to ensure the safety of the drone, obstacle avoidance should be the top consideration, and at this time, the weight of the obstacle avoidance sub-index should be increased; while in an open area, the risk of the drone colliding is significantly reduced, and at this time, the weights of the drone speed and target orientation should be appropriately increased to improve the flight efficiency.

[0106] The specific implementation process is as follows:

[0107] By automatically inputting the current parameters of the drone, the distance to the obstacle, the end point and other data into the SSA algorithm to optimize the most suitable weight set for the current situation, thereby calculating the optimal path of the drone within the next time period.

[0108] (1) First, determine the search range of the dynamic weight:

[0109] W = [w1, w2, w3, w4],

[0110] {w1 + w2 + w3 + w4 = 1 | 0.05 < w1 < 0.2, 0.1 < w2 < 0.4, 0.05 < w3 < 0.2, 0.05 < w4 < 0.2}

[0111] Among them, w1, w2, w3, and w4 are the weights of the heading angle index, the distance to the obstacle index, the speed, and the smoothness index respectively, and W is the weight set.

[0112] (2) Then, initialize the sparrow population, which is used to randomly initialize the positions of the sparrow group. The position of each sparrow represents each weight combination. The specific formula is:

[0113] w i,j = lb j +(ub j - lb j )·rand(N, dim)

[0114] Among them, w i,j is the value of the jth weight of the ith sparrow; ub j and lb j are the upper and lower limits of the jth weight respectively, both of which are a 1×dim vector, representing the maximum and minimum values of each dimension; rand(N, dim) represents generating an N×dim random matrix, and each element is between [0, 1]. N is 50, and dim is 4.

[0115] Calculate the fitness value for each weight combination, that is, the objective function value of each weight, and find the optimal weight combination. The specific formula is:

[0116] F i = w i,1 ·heading(ν, ω)+ w i,2 ·dist(ν, ω)+ wi,3 ·velocity(ν,ω)+w i,4 ·smooth(ν,ω)

[0117] (3) Then update the weight set. Each sparrow approaches the position of the current optimal sparrow, and randomness is introduced to maintain diversity. The update formula for the i-th sparrow at the t-th iteration is as follows:

[0118] w i t+1 = w i t + r × (w* - w i t )

[0119] where w it+1 is the updated weight; w it is the weight of the current iteration; w* is the optimal weight found so far, and r is a random number matrix, taking rand(N, dim), representing the random variation of different weights.

[0120] To ensure that the position of the sparrow is within the search space, it is necessary to limit the position of the sparrow. If the position of the sparrow exceeds the lower limit lb or the upper limit ub, it is restricted within the boundary. The specific formula is:

[0121] w i t+1 = max(w i t+1 , lb)

[0122] w i t+1 = min(w i t+1 , ub)

[0123] (4) Termination condition check: The iteration process includes the check of the termination condition, which is:

[0124] F* = min(F1, F2, ···, F N )

[0125] In each iteration, find the sparrow with the minimum fitness value, and the algorithm stops iterating and outputs the optimal solution.

[0126] Figure 2a and Figure 2b show the comparison results of the local path planning trajectories of UAVs in a complex dynamic environment. Among them Figure 2a is the top view, Figure 2bThis is an inclined view. The sphere with pink-purple grid lines is a dynamic obstacle, and the sphere with black grid lines is a static obstacle. The line connecting the centers of every two spherical grayish-white obstacle frames represents the movement range of the dynamic obstacle. It is not an actual existing obstacle but only visualizes the trajectory of the dynamic obstacle. The vertical color bar on the right represents the height change of the obstacle. The change in color gradient is only used to enhance the visualization effect, making obstacles at different heights more intuitive in the figure and helping to observe and analyze the path planning process of the drone. The green trajectory in the figure is the DWA algorithm, the red trajectory is the AWDWA algorithm, the yellow trajectory is the SDWA algorithm, the blue trajectory is the IAPF algorithm, and the black trajectory is the SSA-DWA algorithm. The improved SSA-DWA algorithm can successfully avoid obstacles and find a better path in a complex dynamic and static obstacle environment.

[0127] Figure 3 It shows the curve of the drone's speed changing with time in a complex dynamic environment. The X-axis represents the iteration time, and the Y-axis represents the drone's speed. The green curve is the DWA algorithm, the yellow curve is the SDWA algorithm, the red curve is the AWDWA algorithm, the blue curve is the IAPF algorithm, and the black curve is the SSA-DWA algorithm. From the changing trend of the curves, it can be seen that the improved SSA-DWA algorithm has the smallest fluctuation during flight, can effectively control the speed fluctuation of the drone, and improve flight stability.

[0128] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be regarded as the protection scope of the present invention.

Claims

1. An improved dynamic window-based local path planning method for unmanned aerial vehicles, characterized in that: It includes the following steps: S1: Initialize the state of the drone, the target point, and the environmental map, and create a grid map of the environment; S2: Improve the dynamic window optimization algorithm. The improvement points are as follows: (1) According to the obstacle density, introduce a dynamic window parameter adaptive adjustment strategy to dynamically adjust the speed and acceleration range of the drone. At the same time, take the distance of the drone from the end point as an influencing factor to make the drone decelerate automatically when approaching the end point; (2) Set the dynamic adjustment sampling resolution and optimize the number of candidate paths according to the environmental complexity; (3) Introduce the sparrow search algorithm to optimize the index weights of the trajectory evaluation function; S3: Input the created path grid map into the improved dynamic window optimization algorithm and output the optimal path.

2. The improved dynamic window method-based UAV local path planning method according to claim 1, wherein: In S1, the initial velocities of the drone on the X, Y, and Z axes are (0, 0, 0). The drone is considered to have reached the target when it is within 1 m of the target point.

3. The local path planning method for unmanned aerial vehicles based on the improved dynamic window approach according to claim 1, wherein: For improvement point (1), set an obstacle density threshold. When the obstacle density is higher than this threshold, narrow the speed and acceleration ranges to limit the movement of the drone; when the obstacle density is lower than this threshold, expand the speed and acceleration ranges to improve the movement flexibility of the drone.

4. The method for local path planning of an unmanned aerial vehicle based on an improved dynamic window method according to claim 3, characterized in that: The mathematical function for the adaptive adjustment of dynamic window parameters is: Among them, vs0 is the current dynamic window parameter of the UAV, and vs i is the dynamic window parameter after adaptive adjustment of the dynamic window parameter; ρ od It represents the density of obstacles within the specified range of the UAV, and ρ0 is the obstacle density threshold, with ρ0 = 0.5 taken; δ is the speed range reduction coefficient when the obstacle density is higher than the threshold. Take δ = 0.

5. When the obstacle density around the drone is greater than the obstacle density threshold, adjust the dynamic window; d i d is the distance of the UAV from the target point; d0 is the threshold distance from the target point, and d0 = 20 is taken; ε is the speed range reduction coefficient when the distance from the target is too close, and ε = 0.7 is taken.

5. The method for local path planning of an unmanned aerial vehicle based on an improved dynamic window method according to claim 1, wherein: For improvement point (2), the sampling resolution is defined as: where V is the velocity sampling resolution; V0 is the default sampling resolution, and V0 = 0.1; ρ od is the obstacle density within the specified range of the UAV; ρ0 is the obstacle density threshold.

6. The method for local path planning of an unmanned aerial vehicle based on an improved dynamic window approach according to claim 1, characterized in that: For improvement point (3), specifically: The motion model of the drone during flight is defined as: where Δt is the adjacent time interval; x t , y t and z t are respectively the abscissa, ordinate and altitude coordinate of the UAV at time t; v t and v z are respectively the UAV XY-axis plane speed and Z-axis speed; θ t and ω t are respectively the UAV steering angle and steering angular velocity; In the trajectory prediction after sampling, evaluate the trajectory through the use of the trajectory evaluation function, and select the speed corresponding to the optimal trajectory. Introduce the sparrow search algorithm in the weight selection of the trajectory evaluation function in each iteration, and find the most suitable weight set for the current situation through the sparrow search algorithm, so as to calculate the optimal path of the drone in the next time period.

7. The method for local path planning of an unmanned aerial vehicle based on the improved dynamic window approach according to claim 6, wherein: The trajectory evaluation function is composed of four sub-indicators, namely the heading angle, the distance from the obstacle, the speed, and the smoothness. The trajectory evaluation function is defined as: F = w1·heading(ν, ω) + w2·dist(ν, ω) + w3·velocity(ν, ω) + w4·smooth(ν, ω) Where: heading(v, ω) is the heading angle index, dist(v, ω) is the index of the distance from the obstacle, velocity(v, ω) is the linear velocity index of the predicted trajectory, and smooth(v, ω) is the smoothness index; w1, w2, w3, and w4 are the weight parameters of these four indicators respectively.

8. The method for local path planning of an unmanned aerial vehicle based on the improved dynamic window approach according to claim 7, wherein: The process of finding the most suitable weight set for the current situation through the sparrow search algorithm and calculating the optimal path of the drone in the next time period is as follows: (1) Determine the search range of the dynamic weights: W = [w1, w2, w3, w4], {w1 + w2 + w3 + w4 = 1|0.05 < w1 < 0.2, 0.1 < w2 < 0.4, 0.05 < w3 < 0.2, 0.05 < w4 < 0.2} Among them, w1, w2, w3, and w4 are the weights of the heading angle index, the distance to the obstacle index, the speed, and the smoothness index respectively, and W is the weight set; (2) Initialization of the sparrow population, which is used to randomly initialize the positions of the sparrow population. The position of each sparrow represents each weight combination. The specific formula is: w i,j = lb j + (ub j - lb j )·rand(N,dim) where w i,j is the value of the j-th weight for the i-th sparrow; ub j and lb j are the upper and lower bounds of the j-th weight respectively, both being a 1×dim vector representing the maximum and minimum values of each dimension; rand(N,dim) represents generating an N×dim random matrix with each element between [0,1]; Calculate the fitness value for each weight combination, that is, the objective function value of each weight, and find the optimal weight combination. The specific formula is: F i = w i,1 ·heading(ν, ω) + w i,2 ·dist(ν, ω) + w i,3 ·velocity(ν, ω) + w i,4 ·smooth(ν, ω) (3) Update the weight set. Each sparrow moves closer to the position of the current optimal sparrow, and at the same time, randomness is introduced to maintain diversity. The update formula for the i-th sparrow at the t-th iteration is as follows: w i t+1 = w i t + r × (w* - w i t ) Among them, w it+1 is the updated weight; w it is the weight of the current iteration; w* is the optimal weight found currently, r is a random number matrix, taking rand(N, dim), representing the random variation of different weights; To ensure that the position of the sparrow is within the search space range, it is necessary to limit the position of the sparrow. If the position of the sparrow exceeds the lower limit lb or the upper limit ub, it will be restricted within the boundary. The specific formula is: w i t+1 = max(w i t+1 , lb) w i t+1 = min(w i t+1 , ub) (4) Termination condition check: The iteration process includes the check of the termination condition, which is: F* = min(F1, F2, ···, F N ) In each iteration, find the sparrow with the smallest fitness value, and the algorithm stops iterating and outputs the optimal solution.

Images