Unmanned aerial vehicle path planning and obstacle avoidance optimization method based on improved elite colony algorithm

Abstract

The invention discloses an unmanned aerial vehicle path planning and obstacle avoidance optimization method based on an improved elite colony algorithm, and the method comprises the following steps: carrying out three-dimensional grid environment modeling and obstacle generation, and obtaining a discrete model of a whole three-dimensional space according to a layering + two-dimensional grid method; elite strategy and path generation: ants select paths by using a random proportion strategy, set a transition probability function and introduce a fluctuation coefficient to prevent excessive concentration of path weights caused by pheromone volatilization; dynamic volatilization rate self-adaptive adjustment: carrying out self-adaptive adjustment according to pheromone distribution, and constructing a dynamic volatilization rate model; path cost calculation and dynamics constraint are integrated, a multi-objective planning method is applied, the path length and flight stability are optimized at the same time, and a path cost function is constructed; and performing path post-processing and smooth optimization, including processing a balanced path to improve the flyability and performing collision detection, thereby establishing an unmanned aerial vehicle dynamics constraint model, and ensuring the flight safety on the premise of path optimization.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous navigation of unmanned aerial vehicles (UAVs), and specifically relates to a three-dimensional path planning method for UAVs based on an improved elite ant colony algorithm, which is particularly suitable for autonomous obstacle avoidance and trajectory optimization of UAVs in complex environments. Background Art

[0002] With the rapid development of drone technology, its application in military reconnaissance, agricultural plant protection, logistics distribution, disaster relief and other fields is becoming more and more extensive. Therefore, drone path planning has become a key issue in practical applications. In the early stages, it focused on geometric rules and simple search algorithms. At present, the introduction of optimization theory and intelligent algorithms is the mainstream. The future trend is the integration of AI-driven and high-precision environmental perception to enhance autonomy, adaptability and multi-machine collaboration capabilities.

[0003] However, existing drone path planning methods have the following technical drawbacks: (1) Traditional ant colony algorithms converge slowly in three-dimensional environments and are prone to falling into local optimal solutions; (2) Existing methods are insufficiently adaptable to dynamic obstacles, resulting in low efficiency in real-time replanning; (3) The generated paths often do not conform to the drone's dynamic constraints, making actual flight difficult; and (4) The path is not smooth enough, increasing the drone's energy consumption and mechanical losses. For example, invention patent CN114610076A discloses a trajectory planning method that combines an artificial potential field method with angle constraints from the A-star algorithm, but does not explicitly address dynamic obstacles or real-time environmental changes.

[0004] To address the above problems, the present invention proposes a path planning method based on an improved elite ant colony algorithm. Combining the idea of ​​the Bresenham algorithm, the method sets constraints for three-dimensional motion, so that the method can autonomously solve the obstacle avoidance problem under different terrains and complete the path smoothing, thereby ensuring the flyability of the UAV. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention aims to provide a method for optimizing UAV path planning and obstacle avoidance based on an improved elite ant colony algorithm. This invention proposes an improved elite ant colony algorithm that integrates a dynamic performance coefficient, an elite strategy, and Bresenham line detection to significantly enhance the efficiency and adaptability of path planning. It also introduces three-dimensional dynamic constraints (including climb rate and turning radius) to enhance the feasibility and safety of the path. Through layered grid modeling and adaptive parameter adjustment, it handles obstacles and optimizes the algorithm's performance in dynamic environments, enabling UAVs to quickly generate safe paths. The algorithm also smooths the path to ensure feasibility in actual flight.

[0006] The technical solution adopted by the present invention to solve its technical problems is:

[0007] The UAV path planning method and obstacle avoidance optimization method based on the improved elite ant colony algorithm include the following steps:

[0008] Step 1. 3D grid environment modeling and obstacle generation

[0009] Step 2. Elite strategy and path generation

[0010] Step 3. Dynamic volatility adaptive adjustment

[0011] Step 4. Path cost calculation and dynamic constraint integration

[0012] Step 5. Path post-processing and smoothing optimization

[0013] Furthermore, the step 1. 3D grid environment modeling and obstacle generation is as follows:

[0014] The entire three-dimensional flight space is divided into several parallel horizontal layers along the height direction (i.e., the Z-axis), with each layer representing a two-dimensional plane at a given height. Using a two-dimensional grid method, each layer is divided into m×n grid cells, with each cell representing information such as whether a location is flyable or whether there are obstacles. A discrete model of the entire three-dimensional space is obtained using the "layering + two-dimensional grid" approach. After establishing an O-XYZ three-dimensional rectangular coordinate system centered at the origin O, the UAV's flight range is determined, and system modeling is performed based on this. The resulting modeling results in a three-dimensional grid network structure that not only accurately reflects the distribution of terrain and obstacles but also facilitates the implementation of subsequent path search algorithms. This model structure facilitates the storage and updating of pheromones during path planning, significantly improving path search efficiency and planning accuracy, particularly when using biomimetic optimization algorithms such as the ant colony algorithm.

[0015] Assuming that obstacles are uniformly distributed, the probability that a grid is an obstacle is set to P o , the probability of free space is 1-P o , expressed as:

[0016]

[0017] Where X~U(0,1) is a uniformly distributed random variable.

[0018] Furthermore, the step 2. elite strategy and path generation is as follows:

[0019] When the algorithm is initialized, the pheromone level on each path is set to a constant (Q is a constant), representing the pheromone concentration on the path from position i to position j. The pheromone level on each path is set to a constant value (Q is a constant). As the iterations proceed, the ants use a random proportional strategy to select paths, and its transition probability function is defined as follows:

[0020]

[0021] Where, Defined as the probability that the kth ant moves from node i to node j at time t; heuristic factor and Characterizes the priority of inter-node transfer, and its value is determined by the specific problem. It remains fixed during the algorithm operation and uses the traditional processing method η ij =1 / d ij (d ij The set of allowed nodes A = {0, 1, …, n-1}-T defines the ant's feasible movement range (T is the tabu table), which dynamically records the nodes the ant has visited to avoid repeated traversal. After a single path is constructed, the system will evaluate the current optimal path based on the tabu table data and reset the table.

[0022] The algorithm parameter α controls the relative influence of the pheromone trajectory and reflects the intensity of the pheromone concentration accumulated in the subsequent path selection process. When the α value is larger, the positive feedback effect of the pheromone will be significantly enhanced, so that the ant colony will show a stronger tendency to choose a path. The parameter β reflects the decision weight of the heuristic information. The larger its value, the more the system tends to choose the local optimal solution, thereby making the transmission probability Move closer to the greedy strategy.

[0023] To maintain a dynamic balance of pheromone concentration, the algorithm introduces a fluctuation coefficient ρ∈[0,1] to prevent excessive concentration of path weights due to pheromone volatilization. When the pheromone concentration on a path is abnormally high, an increase in ρ accelerates pheromone volatilization, thereby balancing path selection probabilities. Conversely, when the pheromone concentration is too low, a decrease in ρ strengthens the pheromone retention effect. After each iteration, the system performs both pheromone updates and volatilization operations, preserving pheromone traces from high-quality paths and maintaining the algorithm's exploration capabilities through the volatilization mechanism. This dual regulation mechanism ensures an effective balance between algorithm development and exploration performance, ensuring the global optimal solution is achieved. It can be expressed as:

[0024] τ ij (t+n)=(1-ρ)τ ij (t)+Δτ ij (3)

[0025]

[0026] Where, represents the amount of pheromone released during the iteration process. If the kth ant does not pass through the path, the pheromone increase is 0. The increment of pheromone passing through the path can be expressed as:

[0027]

[0028] Where Q is the pheromone constant, L k represents the complete path length of ant k.

[0029] Furthermore, the step 3. adaptively adjusting the dynamic volatility rate is as follows:

[0030] Dynamic volatility ρ is based on pheromone distribution (variance σ 2 ) is adaptively adjusted, if σ 2 If σ is larger (larger path differences), then reduce volatility to obtain more path possibilities (improve exploration). 2 If the path converges, the volatility is increased to accelerate the convergence (improve utilization). The dynamic volatility is expressed as:

[0031]

[0032] Where, pheromone variance σ 2 =Var(τ ij |τ ij >0) reflects the degree of dispersion of pheromone distribution. If the variance is high, it means that the path differences are large, and it is necessary to retain diversity and reduce the volatility. On the contrary, it means that the path is driven and redundant information needs to be volatilized to increase the volatility. is a smooth connection between the minimum and maximum volatility, when σ 2 →0,ρ→ρ max When σ 2 →∞,ρ→ρ min The coefficient of 0.1 is adjusted through experiments to control the sensitivity of the variance to the volatility.

[0033] Furthermore, the step 4. integrating the path cost calculation with the dynamic constraints is as follows:

[0034] The UAV path needs to optimize both path length and flight stability. The path cost function is as follows:

[0035] f(t)=w1·N steps +w2∑|Δz| (7)

[0036] Where, the path length N steps It directly affects flight time and energy consumption. Altitude changes ∑|Δz| are caused by frequent climbs and descents, which increase the power burden and may exceed the drone's performance. If the drone is sensitive to altitude (such as a heavy-duty aircraft), increase w2. If shortening the path is the priority, decrease w2.

[0037] Furthermore, the step 5. path post-processing and smoothing optimization is as follows:

[0038] After the above steps, the resulting path may contain redundant nodes (such as a zigzag path), which requires post-processing to smooth the path and improve flyability. This includes two steps:

[0039] (1) Parametric line generation

[0040] For two points p1(x1,y1,z1) and p2(x2,y2,z2), generate discrete path points:

[0041] x2(t)=x1(t)+tΔx (8)

[0042] y2(t)=y1(t)+tΔy (9)

[0043] z2(t)=z1(t)+tΔz (10)

[0044] Where t∈[0,1], and the step size is determined by the maximum coordinate difference.

[0045] (2) Collision Detection

[0046] Collision detection includes four aspects: discretization detection for each interpolation point to check whether it is within the obstacle; early termination, if any node is unreachable, the path is deemed invalid; path simplification, deleting redundant nodes, and reducing the number of control instructions; and ensuring safety, ensuring that the smoothed path still avoids obstacles and complies with the drone's flight constraints.

[0047] To ensure that the generated path conforms to the physical motion capabilities of the drone, it is necessary to model the drone's dynamic constraints. In path planning, these constraints are usually implemented through simplified mathematical or collective models. The following is the drone dynamic constraint model involved in the three-dimensional improved ant colony algorithm:

[0048] (1) Maximum climb rate constraint

[0049] Limit the vertical height change of the drone per unit time:

[0050] |z k+1 -z k |≤Δz max ·Δt (11)

[0051] When generating neighborhood nodes, directly filter out nodes that do not satisfy |Δz|<h cr The direction of movement, h cr Indicates the maximum height difference between adjacent nodes. For example, in actual applications, if the current node height is z k =10m, then the next step can only be z k+1∈[8,12], while avoiding the UAV stall or power system overload due to excessive vertical speed.

[0052] (2) Minimum turning radius constraint

[0053] The lift force L is decomposed into a vertical component Lcosφ and a horizontal component Lsinφ. The gravity mg balances the vertical component of the lift force:

[0054] Lcosφ=mg (12)

[0055] The centripetal force has a horizontal component of lift provided by:

[0056] Lsinφ=mv 2 / R (13)

[0057] Eliminating the lift force L, we can obtain:

[0058] tanφ=v 2 / gR (14)

[0059] Where v is the flight speed of the drone (m / s), g is the acceleration due to gravity (9.81m / s 2 ),φ max Is the maximum roll angle of the drone. The drone is subject to the maximum roll angle limit and must meet the turning radius R≥R min , the minimum turning radius is

[0060] R min =v 2 / gtanφ max (15)

[0061] In a three-dimensional path, the yaw angle is constrained by the moving direction vector. If the current moving direction vector is Then the next direction vector Need to meet:

[0062]

[0063] If the maximum deflection angle of the UAV is ψ max =45°, path nodes with sharp turns are prohibited.

[0064] (3) Speed ​​and acceleration constraints

[0065] Limit the speed and acceleration of the drone to ensure the power system can execute the following path:

[0066]

[0067] Speed ​​is indirectly constrained by path step length. Assuming a fixed time per step, fewer steps result in higher average speed. Path length is penalized in the cost function, encouraging the generation of short paths to meet the speed cap. Acceleration constraints are implemented by limiting the severity of direction changes (e.g., neighborhood searches only allow adjacent nodes).

[0068] (4) Energy consumption constraints

[0069] Total energy consumption versus climbing height Δh and horizontal displacement Related:

[0070]

[0071] Where c1·Δh represents the vertical energy penalty, Represents the horizontal displacement cost.

[0072] (5) 3D path smoothness constraint

[0073] Limit the path curvature to ensure smooth tracking of the drone:

[0074]

[0075] Where κ represents the curvature of the path, which is used to quantify the degree of curvature of the path. The larger the curvature, the more sharply the path bends; conversely, the flatter the path. The tangent vector at a point on the path indicates the direction the drone is flying at that point. Redundant nodes are removed through line detection. After the path is smoothed, the curvature is significantly reduced, avoiding sharp turns.

[0076] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0077] (1) This algorithm introduces the elite strategy, Bresenham algorithm ideas, and three-dimensional motion constraints on the basis of the original ant colony algorithm. This allows the algorithm to autonomously adjust the obstacle avoidance when encountering obstacles on different terrains, and use routes that are more suitable for real life to complete relatively complex obstacle avoidance tasks;

[0078] (2) Technically, by increasing the dynamic volatility coefficient and the elite strategy, the number of algorithm iterations can be reduced and the path quality can be significantly improved. In 50 sets of random obstacle scene tests, 93% of the cases successfully generated safe paths;

[0079] (3) Safety: The maximum climb rate and steering angle are limited to avoid the risk of the drone stalling and to implement motion constraints. Through the obstacle avoidance mechanism, the grid status detects obstacles in real time to ensure path safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 It is the flowchart of the improved elite ant colony algorithm;

[0081] Figure 2 It is a schematic diagram of three-dimensional space division modeling;

[0082] Figure 3 This is a schematic diagram of the traditional elite ant group iteration;

[0083] Figure 4 This is a schematic diagram of the improved elite ant group iteration;

[0084] Figure 5 This is a schematic diagram of the traditional ant colony algorithm path;

[0085] Figure 6 It is a schematic diagram of the improved ant colony algorithm path;

[0086] Figure 7 This is a schematic diagram comparing the traditional ant colony algorithm and the improved ant colony algorithm. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solutions and advantages of the present invention clearer, the following Figure 1-7 , the present invention is described in further detail.

[0088] Reference Figure 1 , based on the improved elite ant colony algorithm flowchart, it is characterized by the smooth processing of all pheromone updates and drone path reachability (ensuring the feasibility of its physical entity movement).

[0089] like Figure 2 As shown, the specific steps of the three-dimensional environment modeling are as follows:

[0090] Step 1. 3D map construction using the grid method

[0091] First, the entire 3D flight space is divided into several parallel horizontal layers along the height direction (i.e., the Z-axis), with each layer representing a 2D plane at a given height. Then, using a 2D grid method, each layer is divided into m×n grid cells, with each cell representing information such as whether a location is flyable or whether there are obstacles. This "layered + 2D grid" approach ultimately results in a discrete model of the entire 3D space.

[0092] After establishing the O-XYZ three-dimensional rectangular coordinate system with the coordinate origin O as the center, the flight range of the UAV can be determined, and system modeling operations can be performed on this basis. Figure 2 As shown in the figure, the modeling results form a three-dimensional grid network structure, which not only accurately reflects the distribution of terrain and obstacles, but also facilitates the implementation of subsequent path search algorithms. Furthermore, this model structure facilitates the storage and updating of pheromones during path planning, especially when using biomimetic optimization algorithms such as ant colony algorithms, which can significantly improve path search efficiency and planning accuracy.

[0093] Step 2. Obstacle generation logic

[0094] Based on a uniformly distributed random model, the probability of each grid being an obstacle is P0 (the obstacle generation probability threshold). It should be marked as an obstacle when n ≤ P0. Here, 1 represents an obstacle, 0 represents free space, and max(x, y, z) represents the grid state. When n ≤ P0, the random number falls within the probability interval [0, P0] and should be marked as obstacle 1, expressed as:

[0095]

[0096] n is a random variable that obeys uniform distribution, denoted as n~U(0,1), that is, n takes random values ​​with equal probability in the interval [0,1).

[0097] Step 3. Parameter design

[0098] Construct a three-dimensional grid map with a size of [202020], where each grid is marked as a feasible area (0) or an obstacle (1); set the starting point

[111] and the end point [202020], and randomly generate obstacles (with probability P o =0.5), the safe flight height threshold is greater than or equal to 5 meters; initialization parameters: number of ants N = 50, maximum number of iterations T max =80, pheromone factor α = 1.5, inspiration factor β = 2.0, pheromone intensity Q = 100, maximum climbing rate is 2, obstacle generation probability is 0.15, elite ant ratio is 0.2, minimum volatility coefficient is 0.1, and maximum volatility coefficient is 0.5.

[0099] Figure 3 and Figure 4 The number of iterations is set to 80, and the iteration graphs of the traditional ant colony algorithm and the improved elite ant colony algorithm are compared in the number of iterations to obtain the optimal solution (blue), and the path change size of the elite ant colony algorithm and the traditional ant colony algorithm is compared with the change of the number of iterations during the simulation process (red).

[0100] like Figure 3 The traditional ant group iteration diagram shows that the iteration of the traditional ant group tends to be flat after 30. Figure 4 The improved elite ant colony iteration graph shows that the improved elite ant colony's convergence rate flattens out after about a dozen iterations. While ensuring optimality, this simulation experiment employed a strategy to expedite algorithm convergence. Comparing the optimal path iterations, it's clear that the improved elite ant colony converges faster than the traditional ant colony. This demonstrates that the improved elite ant colony algorithm can accelerate convergence to the optimal solution for path planning, reducing computation time and finding the optimal solution more quickly.

[0101] Figure 5 It is the traditional ant colony algorithm path. Figure 6 The figure shows the improved ant colony algorithm path. Gray dots represent three-dimensional obstacles, the red path represents the optimal flight trajectory, the green circle represents the starting point, and the purple cross represents the end point. The iteration process is displayed in real time. Comparing the original and smoothed paths, the improved ant colony algorithm path shows a 30% improvement in path length. Convergence iterations are also reduced by 50% in the data comparison. The time required for the algorithm to complete the task also decreases significantly with the reduction in iterations. These comparisons demonstrate that the improved elite ant colony algorithm significantly outperforms the traditional ant colony algorithm in terms of path length, number of inflection points, number of convergence iterations, and time consumption.

[0102] Table 1 Comparison between traditional ant colony algorithm and improved elite ant colony algorithm

[0103]

[0104] Figure 7 This is a comparison chart between the traditional ant colony algorithm and the improved ant colony algorithm. It can be seen that the path length of the smoothed path (blue dotted line) is shorter than the original path length. Compared with the original unsmoothed path, the smoothed path removes unnecessary lengthy paths, reducing the flight time of the drone. At the same time, the smoothed path optimizes some path planning problems where the route turns are very small and the drone is difficult to walk, allowing the drone to fly more smoothly. At the same time, this also optimizes the constraints that may appear in the actual environment and is more suitable for actual tasks.

[0105] Simulations show that the improved code significantly improves the efficiency, stability, and practicality of path planning through dynamic parameter adjustment, elite strategies, and path post-processing. It also enhances the code's engineering usability through enhanced visualization and error handling. These improvements make the algorithm more suitable for real-world UAV missions in complex 3D environments.

Claims

1. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm is characterized by: The following steps are involved: Step 1. 3D grid environment modeling and obstacle generation; Step 2. Elite strategy and path generation; Step 3. Dynamic volatilization rate adaptive adjustment; Step 4. Path cost calculation and integration of dynamic constraints; Step 5. Path post-processing and smoothing optimization.

2. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm according to claim 1, characterized in that: The step 1 is specifically as follows: the entire three-dimensional flight space is divided into several parallel horizontal layers along the height direction (i.e., the Z-axis direction), each layer represents a two-dimensional plane at a given height; each layer is divided into m×n grid units using a two-dimensional grid method, and each grid is used to indicate information such as whether the location is flyable or whether there are obstacles; a discrete model of the entire three-dimensional space is obtained according to the "layering + two-dimensional grid" method; after establishing an O-XYZ three-dimensional rectangular coordinate system centered on the coordinate origin O, the flight range of the UAV is clarified, and system modeling operations are performed on this basis; the modeling results form a three-dimensional grid network structure, which not only accurately reflects the distribution of terrain and obstacles, but also facilitates the implementation of subsequent path search algorithms; this model structure is conducive to the storage and updating of pheromones during the path planning process, especially when using bionic optimization algorithms such as ant colony algorithms, which can significantly improve path search efficiency and planning accuracy; assuming that obstacles are uniformly distributed, the probability of a grid being an obstacle is set to P o , the probability of free space is 1-P o , expressed as Where X~U(0,1) is a uniformly distributed random variable.

3. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm according to claim 1, characterized in that: The second step is as follows: When the algorithm is initialized, the pheromone level on each path is set to a constant (Q is a constant), which represents the pheromone concentration on the path from position i to position j; the pheromone level on each path is set to a constant value (Q is a constant); as the iteration proceeds, the ants use a random proportion strategy to select a path, and its transition probability function is defined as follows: Where, Defined as the probability that the kth ant moves from node i to node j at time t; heuristic factor and Characterizes the priority of inter-node transfer, and its value is determined by the specific problem; Keep it fixed during the algorithm operation and use the traditional processing method η ij =1 / d ij (d ij is the distance between nodes); the set of allowed access A = {0, 1, ..., n-1}-T defines the feasible movement range of the ant (T is the tabu table), which dynamically records the nodes visited by the ant to avoid repeated traversal; after a single path is constructed, the system will evaluate the current optimal path based on the tabu table data and reset the table; The algorithm parameter α controls the relative influence of the pheromone trajectory and reflects the intensity of the pheromone concentration accumulated in the subsequent path selection process. When the α value is larger, the positive feedback effect of the pheromone will be significantly enhanced, so that the ant colony will show a stronger tendency to choose a path; the parameter β reflects the decision weight of the heuristic information. The larger its value is, the more the system tends to choose the local optimal solution, thereby making the transmission probability Move closer to a greedy strategy; In order to maintain the dynamic balance of pheromone concentration, the algorithm introduces a fluctuation coefficient ρ∈[0,1] to prevent excessive concentration of path weights caused by pheromone volatilization: when the pheromone concentration of a path is abnormally high, an increase in the ρ value will accelerate the pheromone volatilization process, thereby balancing the path selection probability; conversely, when the pheromone concentration is too low, a decrease in the ρ value will strengthen the pheromone retention effect; after each iteration, the system will perform pheromone update and volatilization operations to retain pheromone traces of high-quality paths and maintain the algorithm's exploration capability through the volatilization mechanism; this dual adjustment mechanism ensures an effective balance between algorithm development and exploration performance, and provides a guarantee for obtaining the global optimal solution, which can be expressed as: t ij (t+n)=(1-ρ)τ ij (t)+Δτ ij (3) Where, represents the amount of pheromone released during the iteration process; if the kth ant does not pass through the path, the pheromone increase is 0. The increment of pheromone passing through the path can be expressed as: Where Q is the pheromone constant, L k represents the complete path length of ant k.

4. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm according to claim 1, characterized in that: The specific steps of step 3 are as follows: the dynamic volatility rate ρ is calculated based on the pheromone distribution (variance σ 2 ) is adaptively adjusted, if σ 2 If σ is larger (larger path differences), then reduce volatility to obtain more path possibilities (improve exploration). 2 If the path converges, the volatility is increased to accelerate the convergence (improve utilization). The dynamic volatility is expressed as: Where, pheromone variance σ 2 =Var(τ ij |τ ij >0) reflects the degree of dispersion of pheromone distribution; if the variance is high, it means that the path difference is large, and diversity needs to be retained and the volatility rate needs to be reduced; otherwise, it means that the path is driven and redundant information needs to be volatilized to increase the volatility rate; the exponential function is a smooth connection between the minimum and maximum volatility, when σ 2 →0,ρ→ρ max When σ 2 →∞,ρ→ρ min When , information is retained and diversity is maintained; the coefficient of 0.1 is adjusted through experiments to control the sensitivity of the variance to the volatility rate.

5. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm according to claim 1, characterized in that: The details of step 4 are as follows: the UAV path needs to optimize both path length and flight stability. The path cost function is as follows: f(t)=w1·N steps +w2∑|Δz| (7) Where, the path length N steps It directly affects flight time and energy consumption. Altitude changes ∑|Δz| means frequent climbs / descents increase the power burden and may exceed the performance of the drone. If the drone is sensitive to altitude (such as a heavy-duty model), increase w2. If shortening the path is the priority, reduce w2.

6. The UAV path planning and obstacle avoidance optimization method based on the improved elite ant colony algorithm according to claim 1, characterized in that: The fifth step is as follows: After the first to fourth steps, the path obtained may contain redundant nodes (such as a zigzag path), and the path needs to be smoothed through post-processing to improve the flyability. It specifically includes two parts: (1) Parametric line generation For two points p1(x1,y1,z1) and p2(x2,y2,z2), generate discrete path points: x2(t)=x1(t)+tΔx (8) y2(t)=y1(t)+tΔy (9) z2(t)=z1(t)+tΔz (10) Where t∈[0,1], the step size is determined by the maximum coordinate difference; (2) Collision Detection Collision detection includes four aspects: discretization detection for each interpolation point to check whether it is within the obstacle; Early termination: if any node is unreachable, the path is deemed invalid; path simplification is performed, redundant nodes are deleted, and the number of control instructions is reduced; Ensure safety by ensuring that the smoothed path still avoids obstacles and complies with the flight constraints of the drone; To ensure that the generated path conforms to the physical motion capabilities of the drone, it is necessary to model the drone's dynamic constraints. In path planning, these constraints are usually implemented through simplified mathematical or collective models. The following is the drone dynamic constraint model involved in the three-dimensional improved ant colony algorithm: (1) Maximum climb rate constraint Limit the vertical height change of the drone per unit time: |from k+1 -With k |≤Δz max Δt (11) When generating neighborhood nodes, directly filter out nodes that do not satisfy |Δz|<h cr The direction of movement, h cr Indicates the maximum height difference between adjacent nodes; for example, in actual applications, if the current node height is z k =10m, then the next step can only be z k+1 ∈[8,12], while avoiding the UAV stall or power system overload due to excessive vertical speed; (2) Minimum turning radius constraint The lift force L is decomposed into a vertical component Lcosφ and a horizontal component Lsinφ. The gravity mg balances the vertical component of the lift force: Lcosφ=mg (12) The centripetal force has a horizontal component of lift provided by: Lsinφ=mv 2 / R (13) Eliminating the lift force L, we can obtain: tanφ=v 2 / gR (14) Where v is the flight speed of the drone (m / s), g is the acceleration due to gravity (9.81m / s 2 ),φ max The maximum roll angle of the drone; the drone is subject to the maximum roll angle limit and must meet the turning radius R ≥ R min , the minimum turning radius is R min =in 2 / gtanφ max (15) In a three-dimensional path, the yaw angle is constrained by the moving direction vector. If the current moving direction vector is Then the next direction vector Need to meet: If the maximum deflection angle of the UAV is ψ max =45°, path nodes with sharp turns are prohibited; (3) Speed ​​and acceleration constraints Limit the speed and acceleration of the drone to ensure the power system can execute the following path: Speed ​​is indirectly constrained by path step length. Assuming that the time per step is fixed, the fewer steps, the higher the average speed. The path length is penalized in the cost function to encourage the generation of short paths to meet the speed limit. Acceleration constraints are achieved by limiting the severity of direction changes (for example, neighborhood search only allows adjacent nodes). (4) Energy consumption constraints Total energy consumption versus climbing height Δh and horizontal displacement Related: Where c1·Δh represents the vertical energy penalty, represents the horizontal displacement cost; (5) 3D path smoothness constraint Limit the path curvature to ensure smooth tracking of the drone: Where κ represents the curvature of the path, which is used to quantify the degree of curvature of the path; The greater the curvature, the more sharply the path bends; conversely, the flatter the path; The tangent vector at a point on the path indicates the flight direction of the drone at that point; redundant nodes are removed through straight line detection; after the path is smoothed, the curvature is significantly reduced to avoid sharp turns.

Images