A three-dimensional path planning method for unmanned aerial vehicles based on dynamic risk entropy adjustment
By introducing a dynamic risk entropy adjustment mechanism, the problem of balancing exploration and stability in complex environments for UAV path planning is solved. This enables path planning that enhances exploration capabilities in high-risk areas and maintains stability in low-risk areas, thereby improving the obstacle avoidance and trajectory generation capabilities of UAVs in urban low-altitude environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN UNIVERSITY
- Filing Date
- 2026-02-06
- Publication Date
- 2026-05-29
AI Technical Summary
Existing UAV path planning methods struggle to balance exploration and stability in complex environments, resulting in insufficient success rate and stability of flight path planning.
By introducing an environmental risk quantification model and a dynamic risk entropy adjustment mechanism, the exploration intensity is dynamically adjusted by calculating the risk entropy value. Combined with the risk perception reward function, the path planning strategy is optimized, enabling UAVs to enhance their exploration capabilities in high-risk areas and maintain stability in low-risk areas.
It significantly improves the safety, success rate and efficiency of path planning, adapts to dynamic obstacle avoidance and trajectory generation in complex urban low-altitude environments, and has strong adaptability and robustness.
Smart Images
Figure CN122111090A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of deep reinforcement learning and risk perception optimization technology, and in particular to a UAV 3D path planning method based on dynamic risk entropy adjustment, which is applicable to dynamic obstacle avoidance and safe and efficient trajectory generation in complex urban low-altitude airspace environments. Background Technology
[0002] With the development of the low-altitude economy, the demand for drones in delivery, inspection and monitoring, and emergency rescue is constantly increasing. During missions, drones need to perform real-time path planning in complex three-dimensional airspace environments to avoid obstacles and no-fly zones, ensuring flight safety and efficient mission completion.
[0003] Existing path planning methods are mainly divided into two categories: one is rule-based traditional planning methods, such as the A* algorithm and artificial potential field method. These methods are feasible in static or simple environments, but lack adaptability in dynamic environments, are prone to getting trapped in local optima, and struggle to handle high-dimensional continuous action spaces. The other category is intelligent planning methods based on deep reinforcement learning, such as the SAC algorithm. These methods can optimize policies in continuous action spaces and have a certain degree of adaptability. However, traditional SAC algorithms typically use fixed entropy weights to balance exploration and exploitation. This can cause unnecessary randomness in low-risk areas and may lack sufficient exploration in high-risk areas, resulting in insufficient success rate and stability of flight path planning.
[0004] Therefore, existing technologies lack a drone path planning method that can adaptively adjust the exploration intensity according to the level of environmental risk, so as to improve the dynamic obstacle avoidance capability and flight efficiency in complex urban low-altitude environments. Summary of the Invention
[0005] This invention addresses the challenge of balancing exploration and stability in complex environments using existing UAV path planning methods. It proposes a UAV 3D path planning method based on dynamic risk entropy adjustment. By introducing an environmental risk quantification model and a dynamic risk entropy adjustment mechanism, the UAV can enhance its exploration capabilities in high-risk areas and maintain trajectory stability in low-risk areas. This effectively improves the safety, success rate, and efficiency of path planning, making it suitable for dynamic obstacle avoidance and trajectory generation in complex urban low-altitude airspace.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] This invention proposes a UAV 3D path planning method based on dynamic risk entropy adjustment, comprising:
[0008] Step 1: Obtain the current status information and environmental perception data of the drone;
[0009] Step 2: Construct a two-dimensional continuous motion space for outputting flight control commands;
[0010] Step 3: Based on the environmental perception data, calculate the risk entropy value that integrates obstacle spatial density and proximity to the no-fly zone;
[0011] Step 4: Based on the risk entropy value, dynamically adjust the exploration intensity of the path planning strategy using a preset exploration intensity adjustment function;
[0012] Step 5: Using the dynamically adjusted path planning strategy and a reward function that includes a risk perception term, generate flight control commands in the continuous action space;
[0013] Step 6: Control the UAV to perform three-dimensional flight according to the flight control commands.
[0014] Furthermore, step 1 also includes: modeling the UAV flight mission as a Markov decision process MDP = (S, A, P, R, γ), where the state space S represents the UAV's flight state and environmental perception information, the action space A represents the control commands that the UAV can execute, the state transition function P represents the environmental dynamics model, the reward function R is used to guide the UAV to achieve safe and efficient path planning, and γ is a discount factor.
[0015] Furthermore, in step 2, the action space is defined as a vector containing two continuous variables: ,in, This indicates the change in yaw angle. This represents the change in pitch angle. Both are continuous floating-point numbers and are limited to a reasonable range through normalization to ensure flight stability and controllability.
[0016] Further, step 3 includes:
[0017] Step 3.1: Using the layered spatial discretization method, calculate the obstacle density in the current flight layer and the adjacent ±10m spatial layers to obtain the obstacle spatial density index;
[0018] Step 3.2: Use an exponential decay function to characterize the distance relationship between the UAV and the no-fly zone boundary to obtain the no-fly zone proximity index;
[0019] Step 3.3: Weighted fusion of obstacle spatial density and no-fly zone proximity to obtain risk entropy value.
[0020] Furthermore, the exploration intensity adjustment function is expressed as:
[0021]
[0022] in This represents the dynamically adjusted entropy weight; and , respectively, are the lower and upper bounds of the entropy weight, used to control the minimum and maximum exploration intensity of the strategy in low-risk and high-risk regions; k is a nonlinear adjustment coefficient; when At lower levels, the entropy weights approach [value missing]. The drone maintained stable flight; when At higher levels, the entropy weights rapidly approach [the desired value]. To enhance exploration capabilities.
[0023] Furthermore, the reward function is expressed as:
[0024]
[0025] In the formula,
[0026]
[0027]
[0028]
[0029]
[0030] in For the reward function, , , and These represent the path length constraint, smoothness constraint, safety constraint, and risk perception constraint, respectively. This indicates the distance between the drone's current position and the target point. This indicates the distance between the drone and the target point at the previous moment. for The corresponding weighting coefficients, This represents the altitude difference between the drone at the current moment and the previous moment. for The corresponding weighting coefficients, This represents the change in the yaw angle of the drone. This represents the change in the pitch angle of the UAV. for The corresponding weighting coefficients, It is a positive penalty intensity coefficient. Indicates the threat zone.
[0031] In another aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a UAV three-dimensional path planning method based on dynamic risk entropy adjustment as described above.
[0032] The present invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the UAV three-dimensional path planning method based on dynamic risk entropy adjustment as described above.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] This invention proposes a UAV 3D path planning method based on dynamic risk entropy adjustment, which has significant technical advantages. First, by introducing a dynamic risk entropy mechanism, the algorithm can adaptively adjust its exploration behavior, enhancing exploration capabilities in high-risk areas and ensuring flight stability in low-risk areas, thus significantly improving the success rate and stability of path planning. Second, compared to traditional path planning methods, this invention exhibits a higher planning success rate in complex dynamic environments, effectively avoiding risks and finding suitable flight paths in situations with dense obstacles and significant environmental dynamic changes, while achieving shorter planning time and flight range. Third, the dynamic risk entropy mechanism achieves a precise balance between exploration and utilization by quantifying environmental risks, enabling the algorithm to quickly converge to a better solution in multi-objective and multi-interference scenarios, demonstrating strong adaptability and robustness. Finally, this method has strong practicality and promotional value, providing an efficient real-time path planning solution for urban low-altitude UAVs, especially in complex airspace dynamic obstacle avoidance and multi-task execution, offering a new solution for UAV autonomous navigation. In summary, this invention demonstrates good engineering application prospects and research value in real-time UAV path planning. Attached Figure Description
[0035] Figure 1 A basic flowchart of a UAV 3D path planning method based on dynamic risk entropy adjustment is provided for an embodiment of the present invention;
[0036] Figure 2 A model structure diagram of a UAV 3D path planning method based on dynamic risk entropy adjustment provided in an embodiment of the present invention;
[0037] Figure 3 A two-dimensional motion space diagram of the design provided for embodiments of the present invention;
[0038] Figure 4 A training flowchart for a UAV 3D path planning method based on dynamic risk entropy adjustment is provided in this embodiment of the invention.
[0039] Figure 5 Environmental simulation diagrams provided for experiments in embodiments of the present invention;
[0040] Figure 6 The charts show a comparison of the metrics of this invention and the SAC algorithm; where (a) is a comparison of average rewards and (b) is a comparison of success rates.
[0041] Figure 7 The diagram shows the effect of the path planning of the present invention and the SAC algorithm; where (a) corresponds to the SAC algorithm and (b) corresponds to the present invention.
[0042] Figure 8 This is a comparison chart of the qualitative indicators of the present invention and the SAC algorithm. Detailed Implementation
[0043] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:
[0044] like Figure 1 , Figure 2 As shown, a UAV 3D path planning method based on dynamic risk entropy adjustment includes:
[0045] S101: Acquire the current status information and environmental perception data of the drone;
[0046] S102: Construct a two-dimensional continuous action space for outputting flight control commands;
[0047] S103: Based on the environmental perception data, calculate the risk entropy value that integrates obstacle spatial density and proximity to the no-fly zone;
[0048] S104: Based on the risk entropy value, dynamically adjust the exploration intensity of the path planning strategy using a preset exploration intensity adjustment function;
[0049] S105: Using a dynamically adjusted path planning strategy, combined with a reward function that includes a risk perception term, generate flight control commands in the continuous action space;
[0050] S106: Control the UAV to perform three-dimensional flight according to the flight control command.
[0051] Furthermore, the method specifically includes:
[0052] (1) Constructing an environment model for UAV path planning
[0053] In urban low-altitude scenarios, the environment consists of obstacle areas, no-fly zones, and flight target points. Unmanned aerial vehicles (UAVs) need to perform autonomous path planning within a three-dimensional airspace to avoid collisions with dynamic obstacles. The UAV flight mission is modeled as a Markov Decision Process (MDP) = (S, A, P, R, γ), where: the state space S represents the UAV's flight state and environmental perception information; the action space A represents the control commands that the UAV can execute; the state transition function P represents the environmental dynamics model; the reward function R guides the UAV to achieve safe and efficient path planning; and γ is a discount factor.
[0054] (2) Constructing the state space and action space
[0055] The state space S consists of multi-dimensional features of the UAV and its environment, including current position, velocity, target point orientation, and environmental risk quantification features. In the autonomous path planning task of the UAV, the action space defines the control behaviors that the agent can execute at each decision step. Based on the flight attitude control requirements of the UAV, a two-dimensional continuous action space is designed, such as... Figure 3 As shown, this is used to guide the drone to flexibly adjust its flight direction and attitude in three-dimensional space. Specifically, the action space is defined as a vector containing two continuous variables: .in, This indicates the change in yaw angle. This represents the change in pitch angle. Both are continuous floating-point numbers, and are normalized to limit them within a reasonable range to ensure flight stability and controllability. A yaw angle is also set, taking into account the stability of the flight path, maneuverability, and the performance characteristics of the UAV itself. [−30 30 The range of values and pitch angle of ] [−15 , 15 The range of values for ].
[0056] (3) Constructing an environmental risk quantification model
[0057] The risk quantification model integrates obstacle spatial density and no-fly zone proximity indices to calculate environmental risk entropy. Obstacle spatial density is obtained through hierarchical grid statistics, and no-fly zone proximity is characterized by an exponential decay function to represent the distance relationship between the UAV and the no-fly zone boundary.
[0058] Obstacle spatial density is based on a discretized model of layered space, indicating the proportion of building obstacle grids within a 5×5 grid in the current flight layer and adjacent ±10m spatial layers. The calculation formula is as follows:
[0059]
[0060] in This refers to the total number of grid cells occupied by buildings within the current flight level and its adjacent ±10m space levels. This represents the total number of grid cells within the corresponding detection range.
[0061] For no-fly zone proximity, it quantifies the relationship between the drone's current position and the no-fly zone boundary. It is defined as the normalized distance from the current position to the nearest no-fly zone boundary, and an exponential decay function characterizes the proximity effect of the risk.
[0062]
[0063] Where d is the shortest distance to the boundary of the no-fly zone. The attenuation coefficient is... The distance threshold is the risk impact threshold.
[0064] After weighted fusion and normalization, the risk entropy value is generated, as shown in the formula:
[0065]
[0066] in, Indicates obstacle density. Indicates the distance between the drone and the boundary of the no-fly zone. The weights are adjustable.
[0067] (4) Design a dynamic risk entropy adjustment mechanism
[0068] To address the issue of insufficient exploration capabilities of drones in high-risk environments, a dynamic risk entropy adjustment mechanism is introduced to dynamically adjust the entropy weight coefficient α.
[0069] Specifically, based on environmental risk entropy values Construct an exploration intensity adjustment function to achieve adaptive adjustment of entropy weights:
[0070]
[0071] in, and , where are the upper and lower limits of the entropy weight, and k is the nonlinear adjustment coefficient. When At lower levels, the entropy weights approach [value missing]. The drone maintained stable flight; when At higher levels, the entropy weights rapidly approach [the desired value]. To enhance exploration capabilities.
[0072] When environmental risks are high Increase the size of the algorithm to enhance its exploration capabilities; when the environmental risk is low. As the number of resources decreases, the algorithm tends to utilize resources stably, thus achieving a dynamic balance between exploration and utilization.
[0073] (5) Reward function design
[0074] To address the complexity and high safety requirements of UAV 3D path planning tasks in urban low-altitude environments, a multi-dimensional composite reward function was constructed. This function aims to effectively guide the agent to learn an optimal control strategy that balances flight efficiency, path smoothness, and environmental safety through a reinforcement learning framework. Specifically, the reward function consists of a weighted average of the following four core components.
[0075] 1. Path length constraint
[0076]
[0077] The distance between the current position of the intelligent agent (drone) and the target point is... The distance from the target point at the previous moment was , These are the weighting coefficients.
[0078] 2. Smoothness constraint
[0079]
[0080] in The height difference between the agent's current moment and its previous moment. These are the weighting coefficients.
[0081] 3. Security Constraints
[0082]
[0083] This penalty term constrains excessive yaw and pitch angle adjustments, where the agent's yaw angle change is... The change in pitch angle is , These are the weighting coefficients.
[0084] 4. Risk Perception Items
[0085]
[0086] in It is a positive penalty intensity coefficient.
[0087] The risk perception item is used to dynamically adjust rewards in conjunction with the dynamic risk entropy adjustment mechanism, thereby increasing the severity of punishment in high-risk areas.
[0088] The total reward function is defined as:
[0089]
[0090] (6) Training and Path Planning Execution
[0091] During the training phase, the agent is trained using an experience replay mechanism and a soft update strategy. A dynamic risk entropy adjustment mechanism is employed to achieve high exploration of high-risk regions and stable utilization of low-risk regions. Figure 4 As shown.
[0092] During the execution phase, the UAV generates action commands [Δψ, Δθ] in real time based on the current state input, and completes path planning and dynamic obstacle avoidance tasks.
[0093] It is worth noting that this method supports inputting state information and environmental perception data from multiple UAVs, and outputting control commands in a continuous two-dimensional motion space. This method possesses efficient collaborative and dynamic obstacle avoidance capabilities, and is adaptable to complex mission scenarios and real-time deployment on embedded UAV platforms.
[0094] Furthermore, to verify the effectiveness of the present invention, the following experiments and effect verifications were conducted:
[0095] Through simulation experiments in complex urban low-altitude environments, the proposed UAV 3D path planning method (DRE-SAC algorithm) has demonstrated a higher path planning success rate and shorter planning time compared to traditional methods. In scenarios with dense obstacles and significant environmental dynamics, it can effectively avoid risks and generate stable and efficient flight paths, exhibiting strong adaptability and engineering application value.
[0096] To verify the effectiveness of the present invention, the following experiment was conducted:
[0097] 1. Experimental Environment
[0098] Hardware specifications: CPU is a 13th Gen Intel(R) Core(TM) i7-13650HX (2.60 GHz), memory is 32GB, and GPU is an NVIDIA GeForce RTX 4060;
[0099] Software platform: Python version 3.9.7, CUDA version 10.8, and models are built and trained based on the PyTorch 1.10.1 deep learning framework.
[0100] 2. Experimental Simulation Environment
[0101] The experiment was conducted in a simulation environment, using randomly generated buildings as obstacles and setting up no-fly zones to create a more complex path planning experimental scenario. The experimental area was set as a three-dimensional space of 500m × 500m × 100m, with 20 buildings whose radii were randomly generated between 10m and 50m and whose heights varied between 10m and 100m. Figure 5 As shown. To ensure the rationality of building distribution, the algorithm employs a collision detection mechanism to avoid overlap between buildings. Furthermore, two no-fly zones were set up in the experimental environment, their locations and extents determined according to actual task requirements. The existence of these no-fly zones further increases the challenge of path planning, requiring the algorithm to dynamically avoid restricted areas and ensure flight safety.
[0102] This experiment used the SAC algorithm and the DRE-SAC algorithm to train for 1000 rounds respectively, and obtained the curves of reward value and success rate change, as shown below. Figure 6 As shown. From Figure 6 As can be seen in (a), the DRE-SAC algorithm can generate curves with relatively high reward values, and its optimization speed is similar to that of the SAC algorithm, with a convergence speed close to the time step. Figure 6 As shown in (b), the DRE-SAC algorithm exhibits stronger convergence stability, with smaller fluctuations in path planning success rate after convergence, demonstrating better stability. Its path planning success rate is close to 75%, while the original SAC algorithm's success rate after convergence is around 50% (success rate calculation method: number of successful path planning rounds / total number of training rounds). In summary, in the training experiment, the DRE-SAC algorithm shows significantly improved convergence stability and planning success rate compared to the original algorithm, thus enhancing online trajectory planning capabilities.
[0103] from Figure 7 As can be seen in (a), under the first obstacle layout, the SAC algorithm exhibits poor trajectory smoothness and contains turning nodes with significant angle changes, resulting in insufficient flight control precision. In contrast, from... Figure 7 As shown in (b), the DRE-SAC algorithm exhibits smaller maneuver amplitude and a smoother path. The reason for the poor path selection of the SAC algorithm is that the original algorithm lacks sufficient understanding of the environmental state when processing input state data, fails to effectively identify key information in the state, and lacks dynamic perception of changes in complex flight environments. The DRE-SAC algorithm enhances its ability to process the current environmental state by introducing dynamic risk entropy, thereby better understanding and coping with the constantly changing flight environment, and optimizing the accuracy and stability of path planning.
[0104] In simulation experiments, cylindrical abstract structures are used, with the core advantage of significantly improving the computational efficiency of path planning algorithms. Their regular geometric shape simplifies complex collision detection to calculating the distance to the central axis, far exceeding the detection speed of polygonal models. Simultaneously, cylinders naturally provide a uniform safety buffer, and their rotational invariance simplifies algorithm implementation, perfectly aligning with the simulation testing objective of efficiently verifying algorithm performance while preserving core spatial features. Figure 5 The image shown is an example of how we create a simulation environment.
[0105] The improved algorithm proposed in this invention incorporates several hyperparameters, some of which are sensitive to environmental factors. Improper parameter settings may hinder the algorithm's convergence. Therefore, based on extensive numerical simulations, the hyperparameters of the algorithm are listed in the table in this section.
[0106] Table 1 List of Algorithm Hyperparameters
[0107]
[0108] Figure 8 Quantitative results are presented for path length, planning time, and number of inflection points. Simulations were conducted using the same computer under identical conditions to calculate the time required for UAV path planning. From the three comparative graphs of the experimental data, in terms of path planning distance, the SAC algorithm achieves a path planning distance of 820.84m, while the DRE-SAC algorithm achieves 780.35m, demonstrating that the improved algorithm significantly shortens the path length and exhibits superior path planning efficiency. Regarding planning time, the SAC algorithm takes 7.70s, while the DRE-SAC algorithm takes only 6.60s, indicating an improvement in planning efficiency. As for the average turning angle, the SAC algorithm achieves an average turning angle of 11.10°, while the improved SAC algorithm achieves 8.60°, effectively reducing the average turning angle. This is significant for reducing energy consumption and improving flight control stability during UAV flight. In summary, considering these three key indicators, the DRE-SAC algorithm demonstrates superior performance compared to the original SAC algorithm in terms of UAV path planning length, time, and turning angle related to flight stability.
[0109] Experimental results show that the DRE-SAC algorithm exhibits strong adaptability in simulated environments, enabling it to adjust the UAV's flight strategy according to environmental changes during real-time path planning. The method proposed in this invention has strong practicality, providing an efficient real-time path planning solution for urban low-altitude UAVs. Especially when facing dynamic obstacles and multi-target tasks in complex urban airspace, the risk entropy-adjusted SAC algorithm offers a new solution for UAV autonomous navigation decision-making.
[0110] In another aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a UAV three-dimensional path planning method based on dynamic risk entropy adjustment, as shown in S101-S106.
[0111] The present invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a UAV three-dimensional path planning method based on dynamic risk entropy adjustment, as shown in S101-S106.
[0112] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A UAV 3D path planning method based on dynamic risk entropy adjustment, characterized in that, include: Step 1: Obtain the current status information and environmental perception data of the drone; Step 2: Construct a two-dimensional continuous motion space for outputting flight control commands; Step 3: Based on the environmental perception data, calculate the risk entropy value that integrates obstacle spatial density and proximity to the no-fly zone; Step 4: Based on the risk entropy value, dynamically adjust the exploration intensity of the path planning strategy using a preset exploration intensity adjustment function; Step 5: Using the dynamically adjusted path planning strategy and a reward function that includes a risk perception term, generate flight control commands in the continuous action space; Step 6: Control the UAV to perform three-dimensional flight according to the flight control commands.
2. The UAV three-dimensional path planning method based on dynamic risk entropy adjustment according to claim 1, characterized in that, Step 1 further includes: modeling the UAV flight mission as a Markov decision process MDP = (S, A, P, R, γ), where the state space S represents the UAV's flight state and environmental perception information, the action space A represents the control commands that the UAV can execute, the state transition function P represents the environmental dynamics model, the reward function R is used to guide the UAV to achieve safe and efficient path planning, and γ is a discount factor.
3. The UAV three-dimensional path planning method based on dynamic risk entropy adjustment according to claim 1, characterized in that, In step 2, the action space is defined as a vector containing two continuous variables: ,in, This indicates the change in yaw angle. This represents the change in pitch angle. Both are continuous floating-point numbers and are limited to a reasonable range through normalization to ensure flight stability and controllability.
4. The UAV three-dimensional path planning method based on dynamic risk entropy adjustment according to claim 1, characterized in that, Step 3 includes: Step 3.1: Using the layered spatial discretization method, calculate the obstacle density in the current flight layer and the adjacent ±10m spatial layers to obtain the obstacle spatial density index; Step 3.2: Use an exponential decay function to characterize the distance relationship between the UAV and the no-fly zone boundary to obtain the no-fly zone proximity index; Step 3.3: Weighted fusion of obstacle spatial density and no-fly zone proximity to obtain risk entropy value.
5. The UAV three-dimensional path planning method based on dynamic risk entropy adjustment according to claim 1, characterized in that, The exploration intensity adjustment function is expressed as follows: in This represents the dynamically adjusted entropy weight; and , respectively, are the lower and upper bounds of the entropy weight, used to control the minimum and maximum exploration intensity of the strategy in low-risk and high-risk regions; k is a nonlinear adjustment coefficient; when At lower levels, the entropy weights approach [value missing]. The drone maintained stable flight; when At higher levels, the entropy weights rapidly approach [the desired value]. To enhance exploration capabilities.
6. The UAV three-dimensional path planning method based on dynamic risk entropy adjustment according to claim 1, characterized in that, The reward function is expressed as follows: In the formula, in For the reward function, , , and These represent the path length constraint, smoothness constraint, safety constraint, and risk perception constraint, respectively. This indicates the distance between the drone's current position and the target point. This indicates the distance between the drone and the target point at the previous moment. for The corresponding weighting coefficients, This represents the altitude difference between the drone at the current moment and the previous moment. for The corresponding weighting coefficients, This represents the change in the yaw angle of the drone. This represents the change in the pitch angle of the UAV. for The corresponding weighting coefficients, It is a positive penalty intensity coefficient. Indicates the threat zone.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a UAV three-dimensional path planning method based on dynamic risk entropy adjustment as described in any one of claims 1 to 6.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a UAV three-dimensional path planning method based on dynamic risk entropy adjustment as described in any one of claims 1 to 6.