Unmanned aerial vehicle path planning method based on fuzzy adaptive parameter RRT algorithm

Abstract

The invention relates to an unmanned aerial vehicle path planning method based on a fuzzy adaptive parameter RRT algorithm, and belongs to the technical field of path planning. Comprising the following steps: S1, system initialization and environment perception: defining a starting point and a target point of an unmanned aerial vehicle in a three-dimensional task space, loading information of a static obstacle and a dynamic obstacle, and configuring a sensor to obtain environment parameters; s2, designing a discrete fuzzy inference system: defining input variables as obstacle density RN and total node number TN, and defining output variables as step length SS and target deviation probability RP; constructing a fuzzy rule base for dynamically reasoning an output variable according to the input variable; s3, executing a fuzzy adaptive parameter RRT algorithm, optimizing and improving the discrete fuzzy inference system, and adjusting adaptive parameters; and S4, path optimization: optimizing the initial path by adopting a cost function rewiring method, and reducing the length of the path. In the dynamic process of path planning, according to the number of random nodes and the total number of nodes of the current random tree, a fuzzy inference system is designed to adaptively adjust the selection probability and the step length of an algorithm, and the path planning efficiency and the success rate are effectively improved. The problems in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to a method for unmanned aerial vehicle path planning based on a fuzzy adaptive parameter RRT algorithm, belonging to the technical field of path planning. Background Art

[0002] The Rapidly-exploring Random Tree (RRT) algorithm is a path planning method based on spatial sampling proposed by LaValle in 1998. This algorithm constructs a random tree in the task space in the way of node expansion to obtain a safe and feasible path. Compared with the path planning method based on the cost function, the RRT algorithm has disadvantages in terms of path length, smoothness, etc., but it has a simple structure, few algorithm parameters, high computational efficiency, and theoretical path convergence completeness. At present, the research of researchers on RRT mainly focuses on the sampling method of random nodes, the expansion method of new nodes, etc., but there has been no clear selection method for the main parameters of the RRT algorithm.

[0003] The main parameters of the RRT algorithm are the selection probability p g and the extension step size ε. In the prior art, there is a lack of theoretical basis for the selection of these two, and generally they are obtained by empirical trial and error. It is impossible to dynamically adjust the algorithm parameters p g and ε according to the real-time situation of path planning in order to obtain better optimization performance; at the same time, in complex scenarios, the fixed parameters of the standard RRT algorithm lead to long planning time and high failure rate. Poor adaptability to dynamic obstacles: Static parameters cannot respond to real-time obstacle changes, resulting in an increased risk of path collision.

[0004] For example, Chinese Patent Application Publication No. CN112945254A discloses a method for unmanned vehicle curvature continuous path planning based on a rapidly expanding random tree, which is a smooth path planning scheme for any given upper limit curvature. This scheme can clearly consider the smoothness of the path during the planning process, but it is impossible to dynamically adjust the algorithm parameters p g and ε according to the real-time situation of path planning, so as to obtain better optimization performance. Summary of the Invention

[0005] The purpose of the present invention is to propose a method for unmanned aerial vehicle path planning based on a fuzzy adaptive parameter RRT algorithm. On the basis of analyzing the implementation mechanism of the RRT algorithm, combined with a fuzzy inference system, the fuzzy parameter setting of the RRT algorithm is carried out. During the dynamic process of path planning, according to the number of random nodes and the total number of nodes of the current random tree, a fuzzy inference system is designed to adaptively adjust the selection probability and step size of the algorithm, solving the problems that occur in the prior art.

[0006] The UAV path planning method based on the fuzzy adaptive parameter RRT algorithm described in the present invention includes the following steps:

[0007] S1: System initialization and environment perception: Define the starting point q of the UAV and the target point q in the three-dimensional task space, load the information of static and dynamic obstacles, and configure sensors to obtain environmental parameters; start and the target point q goal , load the static obstacle and dynamic obstacle information, and configure the sensor to obtain the environmental parameters;

[0008] S2: Design of discrete fuzzy inference system: Define the input variables as obstacle density RN and total number of nodes TN, and the output variables as step size SS and target bias probability RP; construct a fuzzy rule base for dynamically inferring the output variables according to the input variables;

[0009] S3: Execution of the fuzzy adaptive parameter RRT algorithm, optimize and improve the discrete fuzzy inference system, and adjust the parameters adaptively;

[0010] S4: Path optimization: Optimize the initial path by using the cost function re-wiring method to reduce the path length;

[0011] S5: Simulation implementation, perform simulation implementation on the above steps.

[0012] Preferably, step S3: The execution of the fuzzy adaptive parameter RRT algorithm specifically includes the following sub-steps:

[0013] S31: Algorithm initialization, set the initial parameters and update the selection probability p of the discrete fuzzy inference system g and the step size ε;

[0014] S32: Determine whether the current random tree has reached the target position:

[0015] If not, execute step S33;

[0016] If so, execute step S38;

[0017] S33: Generate a random number p, if the random number p < p g , then execute step S5; otherwise execute step S6;

[0018] S34: Use the target position q goal as the random node q rand , select the node q rand nearest to q near from the current random tree, and extend ε to obtain a new node q new ;

[0019] S35: Randomly generate a node q rand in the task space, and calculate the candidate new node q new ;

[0020] S36: Detect q new Check if there is a threat in the path between q near and q:

[0021] If there is no threat, add q new to the random tree and update the output N of the discrete fuzzy inference system s , then go to step S37;

[0022] If there is a threat, update the output N of the discrete fuzzy inference system r and go to step S37;

[0023] S37: Determine whether the parameter update condition is satisfied:

[0024] If it is satisfied, return to step S31 to update the selection probability p g and the step size ε of the discrete fuzzy inference system;

[0025] If it is not satisfied, return to step S32;

[0026] S38: Reverse search from the random tree to obtain the final path from the starting point q start to the target point q goal .

[0027] Preferably, in step S1, sensors are configured to obtain environmental parameters: the obstacle density RN is detected in real time by a lidar, and the total number of nodes TN is counted by a random tree node counter.

[0028] Preferably, the optimization and improvement of the discrete fuzzy inference system in step S3 specifically include the following:

[0029] Environmental perception: Automatically update the random node density RN and the total number of nodes TN;

[0030] Fuzzy inference: Dynamically calculate the step size SS and the target bias probability RP;

[0031] Path expansion: Generate a random node q with the target bias probability RP rand , otherwise select q goal , expand the step size SS from the nearest node q near , and perform collision detection;

[0032] Path optimization: Use the cost function of RRT to rewire the path to reduce the path length.

[0033] Preferably, the input variables and output variables are defined as:

[0034] The universe of discourse of the random node density RN is [0, 10], and the membership function is triangular;

[0035] The domain of the total number of nodes TN is [0, 1000], and the membership function is Gaussian;

[0036] The domain of the step size SS is [1, 5], and the membership function is triangular;

[0037] The domain of the target bias probability RP is [0, 1], and the membership function is triangular.

[0038] Preferably, the fuzzy rule base includes:

[0039] If RN = high and TN = small, then SS = medium, RP = high;

[0040] If RN = low and TN = large, then SS = large, RP = low;

[0041] If RN = medium and TN = medium, then SS = medium, RP = medium.

[0042] Preferably, the dynamic obstacle is a moving vehicle, and the speed range is 0 - 5 m / s.

[0043] Preferably, the lidar and the inertial measurement unit IMU are fused with data, and the trajectory of the dynamic obstacle is predicted through Kalman filtering.

[0044] Preferably, the fuzzy inference is implemented through an ANFIS adaptive neuro - fuzzy system, and the time consumed by the fuzzy inference does not exceed 3% of the total calculation time.

[0045] Compared with the existing technology, a UAV path planning method based on a fuzzy - adaptive parameter RRT algorithm of the present invention has the following beneficial effects:

[0046] ① Dynamic parameter optimization: Based on the fuzzy inference system, p g and ε are adjusted in real - time, and the planning efficiency is improved by 26.5% compared with the traditional RRT; the dynamic adjustment of the target bias probability RP improves the path convergence speed and reduces the failure rate. The fuzzification strategy can set the parameters of the RRT algorithm more reasonably, effectively improving the path planning efficiency and success rate.

[0047] ② Dynamic adaptability: Respond to environmental changes in real - time through fuzzy rules, effectively reducing the collision risk.

[0048] ③ Efficiency balance: Reduce the step size in high - obstacle - density areas to improve safety, and increase the step size in open areas to accelerate the search.

[0049] Low computational load: The time consumed by ANFIS inference only accounts for 2 - 3% of the overall calculation, which is suitable for embedded systems. Description of the Drawings

[0050] Figure 1 It is a composition diagram of the fuzzy inference system described in Embodiment 1 of the present invention;

[0051] Figure 2 It is the flowchart of the steps of the fuzzy adaptive parameter RRT algorithm described in Embodiment 1 of the present invention;

[0052] Figure 3 It is the path planning diagram described in Embodiment 2 of the present invention. Detailed implementation manners

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0054] Embodiment 1:

[0055] The method for unmanned aerial vehicle path planning based on the fuzzy adaptive parameter RRT algorithm described in the present invention includes the following steps:

[0056] S1: System initialization and environment perception: Define the starting point q start and the target point q goal of the unmanned aerial vehicle in the three-dimensional task space, load the information of static obstacles and dynamic obstacles, and configure sensors to obtain environmental parameters;

[0057] S2: Design of discrete fuzzy inference system: Define the input variables as obstacle density RN and total number of nodes TN, and the output variables as step size SS and target bias probability RP; construct a fuzzy rule base for dynamically inferring the output variables according to the input variables;

[0058] S3: Execute the fuzzy adaptive parameter RRT algorithm, optimize and improve the discrete fuzzy inference system, and adjust the adaptive parameters;

[0059] S4: Path optimization: Optimize the initial path by using the cost function re-wiring method to reduce the path length;

[0060] S5: Simulation implementation, perform simulation implementation on the above steps.

[0061] As Figure 2 shown, the execution of the fuzzy adaptive parameter RRT algorithm specifically includes the following sub-steps:

[0062] S31: Algorithm initialization, set initial parameters and update the selection probability p g and step size ε of the discrete fuzzy inference system;

[0063] S32: Determine whether the current random tree has reached the target position:

[0064] If not, execute step S33;

[0065] If so, execute step S38;

[0066] S33: Generate a random number p. If the random number p < p g , then execute step S5; otherwise, execute step S6;

[0067] S34: Use the target position q goal as the random node q rand , and select the node q rand nearest to q near from the current random tree, and extend it by ε to obtain a new node q new ;

[0068] S35: Randomly generate a node q rand in the task space, and calculate the candidate new node q new ;

[0069] S36: Detect whether there is a threat in the path between q new and q near :

[0070] If there is no threat, add q new to the random tree, update the output N s of the discrete fuzzy inference system, and turn to step S37;

[0071] If there is a threat, update the output N r of the discrete fuzzy inference system and turn to step S37;

[0072] S37: Judge whether the parameter update condition is satisfied:

[0073] If it is satisfied, return to step S31 to update the selection probability p g and the step size ε of the discrete fuzzy inference system;

[0074] If it is not satisfied, return to step S32;

[0075] S38: Backward search from the random tree to obtain the final path from the starting point q start to the target point q goal .

[0076] As Figure 1 shown, the fuzzy inference system first fuzzifies the precise quantity input from the outside world and transforms it into a fuzzy set on the universe of discourse; then, according to the fuzzy rules, selects a suitable fuzzy inference method to obtain the fuzzy inference result for the current fuzzy input, defuzzifies it, and finally obtains the final precise output quantity.

[0077] The RRT algorithm is applied to path planning of moving objects such as unmanned aerial vehicles and mobile robots. Therefore, it is proposed to perform RRT algorithm parameter p gThe fuzzy setting with ε is used to establish a discrete fuzzy inference system with two inputs and two outputs. The input variables are N r and N s , and the outputs are p g and ε.

[0078] During the planning process, overly frequent changes to the algorithm parameters will also lead to a decrease in optimization efficiency. Therefore, the stability of the algorithm parameters should be ensured within a certain range. To this end, for a certain parameter setting, a certain number of q rand selection times and the total number of nodes are allowed. Thresholds and are set. When N r is an integer multiple of , or N s is an integer multiple of , the algorithm parameters are updated.

[0079] The fuzzy set of N r T(N r ) ∈ {PS, PM, PB}, representing "positive small", "positive medium", and "positive big" respectively. The progress of the planning process is characterized by the total number of nodes of the random tree. Similarly, the number of partitions of N s is 3, and T(N s ) ∈ {PS, PM, PB}.

[0080] According to the selection range of p g , the number of fuzzy partitions of p g is selected as 3, and the set T(p g ) ∈ {PS, PM, PB}. The number of fuzzy partitions of the step size ε is 3, and T(ε) ∈ {PS, PM, PB}.

[0081] The fuzzy inference of the RRT algorithm parameters is a two-input two-output system, which can be split into two two-input single-output systems for design. After determining q near , during the extension process towards q new , when N r is small, a larger p g can be adopted to enhance the tendency to extend towards the target position, and at the same time, the step size can be appropriately increased to enhance the extension speed of the random tree; when N r is large, it indicates that q rand has been generated multiple times, and no new q new has been expanded. At this time, a smaller p g needs to be adopted to increase the exploration ability of the RRT, and at the same time, the step size is reduced to avoid the obstacle area. When the total number of nodes N s is small, the exploration ability of the random tree for unknown areas is enhanced; in the later stage of the planning process, the tendency of the random tree towards the target position should be strengthened.

[0082] Simulation experiment

[0083] The fuzzy adaptive parameter RRT algorithm and the standard RRT algorithm of the present invention are respectively used for path planning in a complex task environment. There are 100 obstacle regions in the task environment, and the sizes and positions of the obstacles are randomly generated. 100 digital maps are randomly generated, and the two algorithms are respectively run 100 times. The average value of the 100 times of planning for each is used as the final simulation result.

[0084] Select the parameter p of the standard RRT algorithm g = 0.5 and ε = 6. The parameter settings of the fuzzy adaptive parameter RRT algorithm N r The universe of discourse of r is N s ∈ {0, 5, 10, 15, 20}, N g The universe of discourse of p g is set to p g ∈ {0.3, 0.4, 0.5}, and the universe of discourse of ε is ε ∈ {4, 6, 8}. The fuzzy operation uses a single-point fuzzy set. The fuzzy set of the output quantity is obtained by using the fuzzy inference method, and then the weighted average method is used to defuzzify the fuzzy set of the output quantity. The results of the simulation experiment are shown in Table 1.

[0085] Table 1 Experimental results of the fuzzy adaptive parameter RRT algorithm of the present invention

[0086] Algorithm Name Time (seconds) Length Total Number of Nodes Number of Failures Standard RRT 2.875 203.1 161.15 19 Fuzzy RRT 2.112 196.7 160.92 11

[0087] As can be seen from Table 1, after adopting the fuzzy parameters, the simulation time of the fuzzy adaptive parameter RRT algorithm is better than that of the standard RRT algorithm, and the simulation time is shortened by 26.5%. This shows that online updating the RRT algorithm parameters will not increase the computational complexity of the algorithm additionally, but also make the algorithm parameter settings more reasonable, further improving the optimization efficiency of the algorithm. The path lengths and total number of nodes obtained by the two algorithms do not differ significantly, and the path length obtained by the fuzzy adaptive parameter RRT algorithm is slightly smaller than that of the standard RRT. Set the upper limit of the total number of nodes N s to 1000. When N s > 1000 and no feasible path is obtained, it is considered that the path planning fails this time. From the simulation results, it can be seen that the fuzzy adaptive parameter RRT algorithm significantly reduces the number of planning failures and improves the planning success rate. On the one hand, the reason for the planning failure is that due to the random maps used, there is a possibility that the target position and the starting position are not in the same connected region. On the other hand, it is due to setting the upper limit of N s resulting in insufficient expansion of the random tree nodes.

[0088] The specific implementation process is as follows:

[0089] (1) Scene definition

[0090] Environment: 50m×50m three-dimensional space, containing dynamic obstacles (moving vehicles) and static obstacles (buildings).

[0091] Start and end points:

[0092] Start point q start =(5,5,10)

[0093] End point q goal =(45,45,10)

[0094] Sensor configuration:

[0095] LiDAR (detecting obstacle density RN)

[0096] Tree node counter (counting the total number of nodes TN in the current random tree).

[0097] (2) Design of fuzzy inference system

[0098] Input variables:

[0099] Random node density (RN): The obstacle density is divided into low (0 - 3), medium (4 - 7), and high (8 - 10), and a triangular membership function is used.

[0100] Total number of nodes (TN): The tree scale is divided into small (0 - 300), medium (300 - 600), and large (600 - 1000), and a Gaussian membership function is used.

[0101] Output variables:

[0102] Step size (SS): The range is 1 - 5m, and the membership function is small (1 - 2), medium (2 - 3), and large (3 - 5).

[0103] Standard deviation deviation probability (RP): The range is 0 - 1, and the membership function is low (0 - 0.3), medium (0.3 - 0.7), and high (0.7 - 1).

[0104] Fuzzy rule base (partial example):

[0105] Rule 1: If RN = high and TN = small, then SS = medium, RP = high (prioritize exploration under dense obstacles).

[0106] Rule 2: If RN = low and TN = large, then SS = large, RP = low (quickly converge under sparse obstacles).

[0107] Rule 3: If RN = medium and TN = medium, then SS = medium, RP = medium (balanced mode).

[0108] (3) Fuzzy improvement of the RRT algorithm

[0109] Adaptive parameter adjustment process:

[0110] Environmental perception: Update the obstacle density RN and the number of tree nodes TN every 10 ms.

[0111] Fuzzy inference: Dynamically calculate SS and RP through ANFIS (Adaptive Neuro-Fuzzy Inference System).

[0112] ANFIS Input Design

[0113] Table 2 ANFIS Input Table

[0114] Input Variable Fuzzy Set Partition Physical Meaning Obstacle Distance (OD) Near, Mid, Far Distance from the Current Node to the Nearest Obstacle Goal Direction Deviation (GD) Small, Medium, Large Angle Deviation between the Current Node's Orientation and the Goal Point Obstacle Density (ODen) Low, Medium, High Degree of Obstacle Distribution Density around the Current Node

[0115] ANFIS Output Design

[0116] Table 3 ANFIS Output Table

[0117] Output Variable Range Function Step Size Adjustment Factor (SSF) [0.5,1.5] Dynamically Control the Expansion Step Size Goal Bias Probability (GBP) [0.1,0.8] Probability of Controlling the Sampling Point to Bias towards the Goal

[0118] Dynamic Parameter Adjustment

[0119]

[0120]

[0121] Path Extension:

[0122] Generate a random node q with probability RP rand , otherwise select q goal .

[0123] From the nearest node q near Extend the step size SS and perform collision detection.

[0124] Path Optimization: Use the cost function of RRT (Rewiring) to reduce the path length.

[0125] When the step size is large (SSF = 1.5), expand the search radius to find more optimization opportunities. When the step size is small (SSF = 0.5), focus on local fine-tuning. Only perform Rewiring on local nodes to avoid the high computational cost of global search. By combining the Rewiring mechanism with fuzzy / ANFIS dynamic parameter adjustment, the path planning algorithm significantly improves the path quality while ensuring real-time performance. Rewiring shortens the path by 10% - 15% on average through local path reconnection.

[0126] (4) Simulation Implementation (Python + PyBullet)

[0127] # Fuzzy Inference Module

[0128] def fuzzy_adaptation(RN,TN):

[0129] # Calculate SS and RP using scikit-fuzzy: ml-citation{ref="3,7" data="citationList"}

[0130] return SS, RP

[0131] # Core loop of RRT*

[0132] while not reach_goal:

[0133] RN = get_obstacle_density()

[0134] TN = tree.node_count

[0135] SS, RP = fuzzy_adaptation(RN, TN)

[0136] q_rand = sample_node(RP)

[0137] q_new = extend(q_near, q_rand, SS)

[0138] if collision_free(q_new):

[0139] tree.add_node(q_new)

[0140] rewire(q_new)

[0141] Table 4 Comparison table of performance indicators

[0142] Index Traditional RRT Fuzzy Adaptive RRT Average Path Length (m) 62.4 51.2 Convergence Time (s) 25.7 18.3 Number of Collisions 3-5 0-1 Computing Overhead (CPU%) 12％ 15％

[0143] The comparison of performance indicators after simulation implementation is shown in Table 4. It is significantly better than the traditional RRT in terms of average path length, convergence time, number of collisions, and computational overhead.

[0144] As Figure 3 shown, the path result is obtained in one planning process. The abscissa and ordinate represent the X-axis and Y-axis coordinates respectively, where the coordinate point (0, 0) is the starting position and the coordinate point (100, 100) is the target position. In the initial stage of random tree growth, more branches are generated to try to find potential paths passing through the target position; in the later stage of the planning process, it tends to extend towards the target position to accelerate the optimization speed.

[0145] In summary, the UAV path planning method based on the fuzzy adaptive parameter RRT algorithm of the present invention has the following key advantages:

[0146] Dynamic adaptability: Respond to environmental changes in real time through fuzzy rules to reduce the risk of collisions.

[0147] Efficiency balance: Reduce the step size in areas with high obstacle density to improve safety, and increase the step size in open areas to accelerate the search.

[0148] Low computational load: The time-consuming of ANFIS inference only accounts for 2-3% of the overall calculation, which is suitable for embedded systems.

[0149] As mentioned above, it is only the preferred specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent replacements or changes, and should be covered within the protection scope of the present invention.

Claims

1. A UAV path planning method based on a fuzzy adaptive parameter RRT algorithm, characterized in that, It includes the following steps: S1: System initialization and environmental perception: Define the starting point q of the UAV in the three-dimensional task space start and the target point q goal , load the information of static and dynamic obstacles, and configure sensors to obtain environmental parameters; S2: Design of discrete fuzzy inference system: The input variables are the obstacle density RN and the total number of nodes TN, and the output variables are the step size SS and the target bias probability RP; construct a fuzzy rule base for dynamically inferring the output variables based on the input variables; S3: Execution of fuzzy adaptive parameter RRT algorithm to optimize and improve the discrete fuzzy inference system and adjust the adaptive parameters; S4: Path optimization: Optimize the initial path using the cost function re - wiring method to reduce the path length; S5: Simulation implementation, simulate and implement the above steps.

2. The method for path planning of an unmanned aerial vehicle based on the fuzzy adaptive parameter RRT algorithm according to claim 1, wherein The step S3: Execution of fuzzy adaptive parameter RRT algorithm specifically includes the following sub - steps: S31: Algorithm initialization, setting initial parameters and updating the selection probability p of the discrete fuzzy inference system g and the step size ε; S32: Determine whether the current random tree has reached the target position: If not, execute step S33; If so, execute step S38; S33: Generate a random number p. If the random number p < p g , then execute step S5; otherwise, execute step S6; S34: Use the target position q goal as the random node q rand , and select the node q rand closest to q near from the current random tree, and extend to obtain a new node q new ; S35: Randomly generate a node q within the task space rand , and calculate the candidate new node q new ; S36: Detect q new Whether there is a threat to the path between q near and q: If there is no threat, add q new to the random tree and update the output N of the discrete fuzzy inference system s , and turn to step S37; If there is a threat, update the output N of the discrete fuzzy inference system r And turn to step S37; S37: Determine whether the parameter update condition is met: If satisfied, return to step S31 to update the selection probability p of the discrete fuzzy inference system g and the step size ε; If not, return to step S32; S38: Backward search from the random tree to obtain the final path from the starting point q start to the target point q goal ​ 3. The method for UAV path planning based on the fuzzy adaptive parameter RRT algorithm according to claim 1, characterized in that, In the step S1, configure sensors to obtain environmental parameters: Detect the obstacle density RN in real - time through a lidar, and count the total number of nodes TN through a random tree node counter.

4. The method for UAV path planning based on the fuzzy adaptive parameter RRT algorithm according to claim 2, wherein The optimization and improvement of the discrete fuzzy inference system in the step S3 specifically include the following: Environmental perception: Automatically update the random node density RN and the total number of nodes TN; Fuzzy inference: Dynamically calculate the step size SS and the target bias probability RP; Path extension: Generate a random node q with the target bias probability RP rand , otherwise select q goal , from the nearest node q near Expand the step size SS and perform collision detection; Path optimization: Use the cost function re - wiring of RRT to reduce the path length.

5. The method for UAV path planning based on the fuzzy adaptive parameter RRT algorithm according to claim 1, characterized in that, The input variables and output variables are defined as: The universe of discourse of the random node density RN is [0, 10], and the membership function is triangular; The universe of discourse of the total number of nodes TN is [0, 1000], and the membership function is Gaussian; The universe of discourse of the step size SS is [1, 5], and the membership function is triangular; The universe of discourse of the target bias probability RP is [0, 1], and the membership function is triangular.

6. The method for UAV path planning based on the fuzzy adaptive parameter RRT algorithm according to claim 5, wherein The fuzzy rule base includes: If RN = high and TN = small, then SS = medium, RP = high; If RN = low and TN = large, then SS = large, RP = low; If RN = medium and TN = medium, then SS = medium, RP = medium.

7. The UAV path planning method based on the fuzzy adaptive parameter RRT algorithm according to claim 1, characterized in that The dynamic obstacle is a moving vehicle with a speed range of 0 - 5m / s.

8. The method for UAV path planning based on the fuzzy adaptive parameter RRT algorithm according to claim 3, wherein The lidar and the inertial measurement unit IMU fuse data to predict the trajectory of the dynamic obstacle through Kalman filtering.

9. The method for path planning of an unmanned aerial vehicle based on the fuzzy adaptive parameter RRT algorithm according to claim 4, wherein The fuzzy inference is implemented through an ANFIS (Adaptive Neuro - Fuzzy Inference System), and the fuzzy inference time does not exceed 3% of the total calculation time.

Images