Unmanned aerial vehicle path navigation planning method based on urban low-altitude airspace risk constraints

Abstract

The application discloses a kind of unmanned plane path navigation planning methods based on city low-altitude airspace risk constraint, belong to the technical field of unmanned plane path planning, comprising: initialization fast exploration random tree algorithm;Fast exploration random tree algorithm is randomly sampled and node extension is used;Local optimization is carried out using the algorithm of elephant trunk, and extended node is generated;Using collision detection function, obstacle detection between extended nodes is carried out;Tree is constructed, and optimal path is searched and output;The application combines the random expansion characteristics of fast exploration random tree and the local optimization ability of elephant trunk search, so as to generate the optimal safe path in three-dimensional environment quickly, can effectively reduce the generation of redundant nodes, improve path planning efficiency and convergence speed, make it more suitable for unmanned plane in city airspace fast path planning.

Description

Technical Field

[0001] This invention belongs to the technical field of UAV path planning, specifically relating to a UAV path navigation planning method based on urban low-altitude airspace risk constraints. Background Technology

[0002] As a primary vehicle for the low-altitude economy, drones are widely used in passenger transport and logistics delivery. Urban airspace scenarios place great importance on rapid path planning for drones. Effective path planning helps drones achieve safe and efficient flight in complex urban environments, avoiding collisions, shortening flight time, and ensuring timely mission completion.

[0003] Traditional path planning algorithms (such as Rapidly Exploring Random Tree, RRT) often suffer from problems such as path redundancy, low node expansion efficiency, poor path smoothness, and slow convergence speed when dealing with high-density obstacles. This leads to longer UAV flight time, increased energy consumption, and even affects flight safety. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned shortcomings in the prior art by providing a UAV path navigation planning method based on urban low-altitude airspace risk constraints, thereby solving the problems of path redundancy, low node expansion efficiency, poor path smoothness, and slow convergence speed in traditional path planning algorithms.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for unmanned aerial vehicle (UAV) path navigation planning based on urban low-altitude airspace risk constraints includes the following steps:

[0007] S1. Set the drone path navigation planning parameters, including the search tree T and the starting point q of the drone path. start and the endpoint q goal ;

[0008] S2. Randomly sample nodes q using the fast exploratory random tree algorithm. rand And search for nodes q in the search tree T. rand The nearest node q near This leads to the generation of node q. new ;

[0009] S3, Expand node q new As the location of the longhorn beetle, the longhorn beetle whisker algorithm is used to update the longhorn beetle position and generate an extended node q. bettle ;

[0010] S4, Connect the extended node q bettle With extension node qnew ;

[0011] If there are no obstacles between the connected paths, then node q will be expanded. bettle Add to the search tree T and connect it with node q near Connect the links, then execute S5; if there are obstacles in the connected paths, return to execute S3.

[0012] S5. Calculate the expanded node q bettle With the endpoint q goal The distance D between them; if the distance D is less than the threshold ε, the optimal path is output; if the distance D is greater than the threshold ε, S2 to S5 are iteratively executed until the distance D is less than the threshold ε.

[0013] Furthermore, S2 specifically includes:

[0014] Randomly sample node q in the urban airspace environment rand Search in the search tree T for the distance from the sampling node q rand The nearest node q near If node q near With sampling node q rand If there are no obstacles in the connection, then the fast exploratory random tree algorithm with a path search step size Δq is used to start from node q. near To sampling node q rand Expand in the direction to generate an extended node q. new .

[0015] Furthermore, step S3 includes the following sub-steps:

[0016] S31, Expand node q new As the location of the longhorn beetle, and randomly generate the unit vector that the longhorn beetle must search randomly;

[0017] S32. Based on the location of the longhorn beetle and the unit vector that the longhorn beetle needs to randomly search, calculate the search positions of the two tentacles of the longhorn beetle and the cost function of the corresponding search positions;

[0018] S33. Update the beetle position according to the cost function, and then generate the extended node q. bettle .

[0019] Furthermore, in S31, the longhorn beetle must randomly search for a unit vector as follows:

[0020]

[0021] In the formula, For the longhorn beetle, a random unit vector must be searched; rnd(.) is a random vector, and k is the position dimension.

[0022] Furthermore, S32 specifically includes:

[0023] Calculate the search positions of the longhorn beetle's two tentacles:

[0024]

[0025] In the formula, X r and X l The search positions of the right and left antennae of the longhorn beetle are respectively; d is the sensing distance of the longhorn beetle's antennae; X is the extended node q. new Location;

[0026] Calculate the cost function f(X) for the search positions of the two tentacles of the longhorn beetle:

[0027]

[0028] In the formula, α1 and α2 are the distance costs, respectively. Risk Cost Weighting coefficients; C i i = 1, 2, ..., k, representing the drone in RP i-1 RP i The overall risks and costs of the flight segment.

[0029] Furthermore, the cost of distance and risk costs They are respectively:

[0030]

[0031] in,

[0032] In the formula, ||RP i-1 ,RP i ||For drones from waypoint RP i-1 Fly to RP i The distance traveled; v UAV R is the cruising speed of the drone; R(i,j,h) is the equivalent safety level of the drone at coordinates (i,j,h); R norm This is the normalized risk value for the route in this area; Med(R(i,j,h)) is the median of all values ​​in the risk matrix for this flight segment.

[0033] Furthermore, the equivalent security level R(i,j,h) is specifically as follows:

[0034] R(i,j,h)=P(LOC)·R f (i,j,h)

[0035] Among them, R f =R i ⊙M

[0036] R i=D peo ·A cas

[0037] In the formula, P(LOC) is the probability that the drone will fail and crash while in flight, and R... f (i,j,h) represents the final ground-to-ground risk value for the drone crashing within the grid at location (i,j,h); R f This is the final ground risk value matrix; R i D is the initial ground risk matrix; peo A represents the ground population density in the area where the drone crashed. cas M represents the area of ​​casualties caused by the drone crash; M is the matrix of the effectiveness of ground risk mitigation measures, which includes two types: M1 and M2. Type M1 is to reduce the number of people directly exposed to the risk area, and type M2 is to reduce the impact of the drone hitting the ground.

[0038] Furthermore, in step S33, the position of the longhorn beetle is updated to x according to the cost function. t+1 This leads to the generation of extended node q. bettle Update the location of the longhorn beetle. t+1 for:

[0039] x t+1 =x t +δ t *d*sign[f(x r )-f(x l )]

[0040] δ t+1 =δ t *η

[0041] In the formula, x t The position of the longhorn beetle before the update; δ t Let δ be the search step size of the beetle whisker algorithm at time t. t+1 f(x) represents the search step size of the beetle whisker algorithm at time t+1; η is the reduction coefficient, and d is the sensing distance of the beetle whisker angle; r ) and f(x l ) are the cost functions corresponding to the search positions of the right and left tentacles of the longhorn beetle, respectively.

[0042] The UAV path navigation planning method based on urban low-altitude airspace risk constraints provided by this invention has the following beneficial effects:

[0043] 1. This invention utilizes an improved RRT method (RRT-BAS) optimized by the Beetle Antennae Search (BAS) algorithm to solve the path planning problem for UAVs in urban airspace. The Beetle Antennae Search algorithm mimics the antennae-based sensing mechanism of a longhorn beetle, dynamically adjusting its movement direction and step size based on the detection results of its left and right antennae to gradually find the optimal path. The BAS algorithm adjusts and relocates nodes along the path, reducing redundant nodes and minimizing path turns, resulting in a smoother path. Simultaneously, the Beetle Antennae Search algorithm exhibits high search efficiency and local optimization capabilities, making it suitable for path planning in complex environments and high-dimensional problems. It effectively compensates for the slow convergence speed of the RRT algorithm due to random node sampling.

[0044] 2. This invention improves the path planning efficiency of UAVs in complex urban airspace environments. The algorithm combines the random expansion characteristics of Rapid Exploratory Random Tree (RRT) with the directional optimization capabilities of Beetle Beard Search (BAS) to quickly generate optimal safe paths in a three-dimensional environment. This method effectively avoids obstacles, reduces public safety risks associated with flight, and ensures the feasibility and safety of the path, especially in urban airspace environments with complex terrain, numerous obstacles, dense populations, and dynamically changing risks. Attached Figure Description

[0045] Figure 1 This is a visualization of the three-dimensional ground risk matrix of the UAV in this invention.

[0046] Figure 2 This is an analysis of the crash process of the UAV after the failure of the present invention.

[0047] Figure 3 This is a flowchart of the UAV path navigation planning method based on urban low-altitude airspace risk constraints according to the present invention.

[0048] Figure 4 The path result diagrams of various algorithms are shown for the verification of this invention.

[0049] Figure 5 The following graph shows the convergence speed results of various algorithms during the verification of this invention. Detailed Implementation

[0050] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0051] Example 1

[0052] This embodiment provides a UAV path navigation planning method based on urban low-altitude airspace risk constraints, utilizing an improved RRT method (RRT-BAS) optimized by the Beetle Antennae Search (BAS) algorithm. The Beetle Antennae Search, mimicking the antennae sensing mechanism of a longhorn beetle, dynamically adjusts its movement direction and step size based on the detection results of its left and right antennae to gradually find the optimal path. A cost function integrating the distance cost and risk cost of the UAV flight path is used as the objective function of the BAS mechanism. By leveraging the excellent local optimization capabilities of the Beetle Antennae Search, the method overcomes the shortcomings of the RRT algorithm, such as high overall cost and slow convergence speed due to random sampling. Figure 3 Specifically, it includes the following:

[0053] Step S1: Initialize the fast exploratory random tree algorithm;

[0054] Configure the drone path navigation planning parameters, including the initial search tree T and the starting point q of the drone path. start and the endpoint q goal The starting point of this implementation is q. start and the endpoint q goal These two locations form the basis of the entire route planning process, and RRT's route expansion will be based on these two locations.

[0055] Step S2: Quickly explore the random sampling and node expansion of the random tree algorithm;

[0056] Specifically, in this embodiment, nodes q are randomly sampled in the urban airspace environment. rand Search in the search tree T for the distance from the sampling node q rand The nearest node q near If node q near With sampling node q rand If there are no obstacles in the connection, then the fast exploratory random tree algorithm with a path search step size Δq is used to start from node q. near To sampling node q rand Expand in the direction to generate an extended node q. new .

[0057] Step S3: Use the beetle whisker algorithm to perform local path optimization, update the beetle position, and generate extended node q. bettle Specifically, it includes the following sub-steps:

[0058] Step S31: The longhorn beetle is located at the extended node q. new Its position is X:

[0059] X(x1,x2,…,x n )

[0060] In the formula, xi Let i be the position of the beetle's center of mass at time i.

[0061] A longhorn beetle must be randomly generated and its unit vector must be randomly searched.

[0062]

[0063] In the formula, For the longhorn beetle, a random unit vector must be searched; rnd(.) is a random vector, and k is the position dimension.

[0064] Step S32: Based on the location of the longhorn beetle and the unit vector that the longhorn beetle must randomly search, calculate the search positions of the two tentacles of the longhorn beetle and the cost function of the corresponding search positions;

[0065] Specifically, in order to simulate the activity of the beetle's antennae, this embodiment provides the search positions of the right and left antennae of the longhorn beetle to simulate the search behavior;

[0066] The search locations for the two tentacles of the longhorn beetle are:

[0067]

[0068] In the formula, X r and X l The search positions of the right and left antennae of the longhorn beetle are respectively; d is the sensing distance of the longhorn beetle's antennae; X is the extended node q. new Location;

[0069] The longhorn beetle determines its next action based on the path cost function sensed by its left and right antennae, calculating the extended q... new The cost function f(X) on both sides of the search direction r ) and f(X l Specifically, the three-dimensional airspace is divided into several unit cubes, and the UAV's selectable waypoints RP are assumed to be... k Given the vertices of a cube, the drone's path can be represented as Path = {RP1, RP2, ..., RP...} k The route cost function f(X) can be expressed as the sum of costs for a series of flight segments. Considering both path length and flight risk, the cost function f(X) for the search position of the longhorn beetle's two tentacles is:

[0070]

[0071] In the formula, α1 and α2 are the distance costs, respectively. Risk Cost Weighting coefficients; C i i = 1, 2, ..., k, representing the drone in RP i-1 RP i The overall risks and costs of the flight segment.

[0072] Distance Cost and risk costs They are respectively:

[0073]

[0074] To ensure that the distance cost and the risk cost are on the same order of magnitude, this invention normalizes the risk assessment value R(i,j,h). norm The normalized risk value for this route is represented by the minimum of the mean and median of all values ​​in the risk matrix for this flight segment, indicating the concentrated distribution trend of the risk level in this area.

[0075]

[0076] In the formula, ||RP i-1 ,RP i ||For drones from waypoint RP i-1 Fly to RP i The distance traveled; v UAV R is the cruising speed of the drone; R(i,j,h) is the equivalent safety level of the drone at coordinates (i,j,h); R norm This is the normalized risk value for the route in this area; Med(R(i,j,h)) is the median of all values ​​in the risk matrix for this flight segment.

[0077] The specific calculation process of the equivalent security level R(i,j,h) in this embodiment is as follows:

[0078] The primary safety risk posed by drone operation is the expected casualties resulting from drone system failure and crashes on urban ground (hereinafter referred to as "ground risk"). This is the most significant impact of drone flights in urban airspace on public safety. A three-dimensional safety risk matrix R can be constructed in urban airspace. n,m,l (i = 1, 2, ..., n; j = 1, 2, ..., m; h = 1, 2, ..., l), such as Figure 1 As shown:

[0079]

[0080] In the formula, R(i,j,h) represents the equivalent safety level of the UAV at coordinates (i,j,h), with units of casualties per flight hour, and can be calculated by the following formula:

[0081] R(i,j,h)=P(LOC)·R f (i,j,h)

[0082] In the formula, P(LOC) is the probability that the drone will fail and crash while in flight, and R... f(i,j,h) represents the final ground-to-ground risk value of the drone crashing within the grid at location (i,j,h), which can be represented by the final ground-to-ground risk matrix R. f express.

[0083] The final ground risk matrix R of drones f This represents the ground-to-ground risk impact on ground personnel when a drone crashes on the ground. It consists of two main parts: the initial ground-to-ground risk matrix R of the drone. i The effect matrix M of ground risk mitigation measures for drones.

[0084] R f =R i ⊙M

[0085] Among them, the initial ground risk matrix R of the UAV i This refers to the inherent risk to the ground, which is an inherent attribute of drone operation and is only related to the drone's characteristic parameters (maximum characteristic size, maximum takeoff mass, and cruise speed) and the population density on the ground in the crash area.

[0086] R i (i,j,h)=D peo ·A cas

[0087] In the formula, D peo A represents the ground population density in the area where the drone crashed. cas This indicates the area of ​​casualties caused by the drone crash.

[0088] M represents the effectiveness of ground risk mitigation measures, and is divided into two types: M1 and M2. The M1 type reduces the number of people directly exposed to the risk area, and can be represented by quantifying the protective effect of ground elements such as buildings on people from drone impacts.

[0089]

[0090] In the formula, P s The ground shading coefficient is given by parameter α when P s When the value is 6, the mortality rate P f =50% of the required impact energy value, generally α = 34kJ, and parameter β is the energy threshold required for human lethality, generally β = 34J. E crash Let be the kinetic energy of the drone upon impact, and k be a correction factor applicable when the instantaneous kinetic energy of the drone upon impact is low. It can be calculated using the following formula:

[0091]

[0092] M2 (Mitigation of the Impact of Drones) is a method to reduce the impact of drones hitting the ground, and a widely discussed approach is to equip drones with parachutes. This study characterizes the mitigation effect of M2 measures by analyzing the kinematic process of the parachute system after the drone fails in mid-air.

[0093] Taking the multi-rotor drone developed in the invention research as an example, Figure 2 The fall process of a multi-rotor UAV equipped with a parachute after failure was analyzed. Under ideal conditions, it is subjected to gravity mg and air resistance F. airdrag The impact can be divided into two stages. In the diagram, v represents the direction of velocity, which is opposite to the direction of air resistance. t1 is the parachute deployment time, and t2 is the time it takes for the drone to crash. In the first stage, the drone loses power and transitions from uniform horizontal flight to a ballistic descent. During this stage, the parachute system is not activated. In the second stage, the parachute system deploys, and the drone enters a parachute descent.

[0094] In such Figure 2 During the fall shown, the three important variables affecting the risk to Earth are the horizontal velocity of the drone at the time of impact. Vertical velocity The total fall time t2 is calculated using the following formula:

[0095]

[0096] In the formula, m is the mass of the drone. and Let ρ and g represent the instantaneous horizontal and vertical velocities of the UAV at time t1 (the moment the parachute system is fully deployed), respectively, where ρ represents air density, g is gravitational acceleration, and C is the velocity of the UAV. d,p S represents the overall air resistance coefficient of the parachute. p This indicates the overall effective windward area of ​​the parachute.

[0097] The total time t2 of the drone crash is related to the altitude of the failure location, and can be expressed as:

[0098]

[0099] In the formula, C d,u S represents the air resistance coefficient of the drone. u This indicates the effective windward area of ​​the drone's fuselage.

[0100] This embodiment utilizes the above calculation process to calculate the three-dimensional ground risk matrix R of UAVs operating in urban airspace. n,m,l Then, the equivalent security level R(i,j,h) is calculated.

[0101] Step S33: Update the beetle's position based on the cost function value fed back from the antennae, select the direction with the lower cost for expansion, and update the beetle's position to x. t+1 Generate extended node q bettle :

[0102] x t+1 =x t +δ t *d*sign[f(x r )-f(x l )]

[0103] δ t+1 =δ t *η

[0104] In the formula, x t The position of the longhorn beetle before the update; δ t Let δ be the search step size of the beetle whisker algorithm at time t. t+1 f(x) represents the search step size of the beetle whisker algorithm at time t+1; η is the reduction coefficient, and d is the sensing distance of the beetle whisker angle; r ) and f(x l The cost functions for the search positions of the right and left tentacles of the longhorn beetle are respectively.

[0105] Step S4: Use a collision detection function to detect obstacles between extended nodes;

[0106] Specifically, connect the extended node q bettle With extension node q new If there are no obstacles between the connected paths, then node q will be expanded. bettle Add to the search tree T and connect it with node q near Connect the links, then execute S5; if there are obstacles in the connected paths, return to execute S3.

[0107] Step S5: Construct a tree, search for and output the optimal path;

[0108] Calculate the extended node q bettle With the endpoint q goal The distance D between them; if the distance D is less than the threshold ε, the optimal path is output; if the distance D is greater than the threshold ε, S2 to S5 are iteratively executed until the distance D is less than the threshold ε.

[0109] The RRT-BAS algorithm in this embodiment fully combines the random expansion capability of RRT with the direction selection capability of BAS. The addition of BAS makes the expansion process no longer completely random, but incorporates local optimization capabilities. Through path optimization and asymptotic optimization, the quality of path planning is further improved. By reconnecting and neighborhood search, the RRT algorithm can continuously optimize the path and gradually converge to the global optimum; making the expanded path more consistent with the trend of the optimal path and reducing the generation of redundant nodes.

[0110] To verify the effectiveness of the algorithm proposed in this invention, RRT, BAS, RRT*, and RRT*-BAS were selected as baseline algorithms for comparative analysis. Figure 4 , Figure 5 As shown in the figure. Simulation results show that, considering both path quality and convergence speed, the RRT-BAS algorithm of this invention is significantly superior to the RRT, BAS, RRT*, and RRT*-BAS algorithms.

[0111] Although specific embodiments of the invention have been described in detail with reference to the accompanying drawings, this should not be construed as limiting the scope of protection of this patent. Various modifications and variations that can be made by a person skilled in the art without inventive effort within the scope described in the claims still fall within the scope of protection of this patent.

Claims

1. A method for unmanned aerial vehicle (UAV) path navigation planning based on urban low-altitude airspace risk constraints, characterized in that, Includes the following steps: S1. Set the drone path navigation planning parameters, including the search tree. The starting point of the drone's path and the end point ; S2. Randomly sample nodes using the fast exploratory random tree algorithm. and in the search tree Searching for nodes The nearest node This leads to the expansion of node generation. ; S3, Expand the node As the location of the longhorn beetle, the longhorn beetle whisker algorithm is used to update the longhorn beetle position and generate extended nodes. ; S4, Connect Expansion Nodes With extended nodes ; If there are no obstacles between the connected paths, then the node will be expanded. Add to search tree In, and with nodes Connect the links, then execute S5; if there are obstacles in the connected paths, return to execute S3. S5, Calculate the expanded node and the finish line The distance D between them; if the distance D is less than the threshold If the distance D is greater than the threshold, then output the optimal path; Then iterate through steps S2 to S5 until the distance D is less than the threshold. ; S3 includes the following steps: S31, Expand the node As the location of the longhorn beetle, and randomly generate the unit vector that the longhorn beetle must search randomly; S32. Based on the location of the longhorn beetle and the unit vector that the longhorn beetle needs to randomly search, calculate the search positions of the two tentacles of the longhorn beetle and the cost function of the corresponding search positions; S33. Update the beetle position according to the cost function, and then generate extended nodes. ; Specifically, S32 includes: Calculate the search positions of the longhorn beetle's two tentacles: In the formula, and Search locations for the right and left tentacles of the longhorn beetle, respectively; The sensing distance of the longhorn beetle's whiskers; For expanding nodes Location; For longhorn beetles, a random unit vector must be searched. Cost function for calculating the search position of the two tentacles of a longhorn beetle : In the formula, , Distance cost Risk and cost Weighting coefficients; , i =1,2,…,k, indicating that the drone is in The overall risks and costs of the flight segment; Distance Cost and risk costs They are respectively: in, In the formula, For drones from route nodes Fly to The distance traveled; The cruising speed of the drone; For the drone at coordinates The equivalent security level of the location, and i =1,2,…, n ; j =1,2,…, m ; h =1,2,…, l ; This is the normalized risk value for the path in this region; This is the median of all values ​​in the risk matrix for this flight segment.

2. The UAV path navigation planning method based on urban low-altitude airspace risk constraints according to claim 1, characterized in that, S2 specifically includes: Random sampling nodes in urban airspace environment In the search tree Searching for sampling nodes The nearest node If node With sampling nodes If the connecting lines have no obstacles, then the path search step size is determined using the fast exploration random tree algorithm. From node To sampling node Expand in direction to generate extended nodes. .

3. The UAV path navigation planning method based on urban low-altitude airspace risk constraints according to claim 1, characterized in that, In S31, the longhorn beetle must randomly search for a unit vector as follows: In the formula, For random vectors, For location dimension.

4. The UAV path navigation planning method based on urban low-altitude airspace risk constraints according to claim 3, wherein the equivalent safety level Specifically: in, In the formula, This represents the probability that a drone will malfunction and crash while in flight. The location where the drone crashed on the ground is Final ground risk value within the grid; This is the final ground risk value matrix; This is the initial ground risk matrix; The density of people on the ground in the area where the drone crashed. The area of ​​casualties caused by the drone crash; A matrix of the effectiveness of land risk mitigation measures. include M 1 and M 2. Two types, M Type 1 aims to reduce the number of people directly exposed to high-risk areas. M Type 2 is to reduce the impact of drones hitting the ground.

5. The UAV path navigation planning method based on urban low-altitude airspace risk constraints according to claim 1, wherein in step S33, the position of the longhorn beetle is updated according to the cost function. This leads to the generation of extended nodes. Update the location of the longhorn beetle. for: In the formula, The position of the longhorn beetle before the update; For the longhorn beetle algorithm in The search step size at any given moment. For the longhorn beetle algorithm in The search step size at any given moment; The reduction factor is... The sensing distance of the longhorn beetle's whiskers; and These are the cost functions corresponding to the search positions of the right and left tentacles of the longhorn beetle, respectively.

Images