Multi-unmanned aerial vehicle path planning method for power transmission line inspection

Abstract

The invention discloses a multi-unmanned aerial vehicle path planning method for power transmission line inspection, and belongs to the field of path planning. According to the method, mobile edge computing (MEC) and unmanned aerial vehicle airborne intelligent decision making are combined, and the method comprises the steps that firstly, an MEC server synchronizes a cluster state and issues a global demand matrix; the unmanned aerial vehicle dynamically updates the global risk matrix through real-time environment perception; and fusing the real-time state, the risk and the demand matrix of each unmanned aerial vehicle airborne unit, and autonomously making a decision to generate an optimal short-term planning path. Before the path is executed officially, the path is reported to an MEC server for cooperative conflict detection, and the unmanned aerial vehicle can execute flight only after it is confirmed that no conflict exists. According to the method, the unmanned aerial vehicle cluster is endowed with autonomous perception and rapid avoidance capability for dynamic unknown risks, path planning becomes a real-time evolution intelligent decision-making process, and the safety of cluster collaborative operation is ensured through centralized conflict detection, so that the efficiency, robustness and reliability of a power transmission line inspection task are remarkably improved.

Description

Technical Field

[0001] This application relates to the technical field of unmanned aerial vehicle path planning, and more specifically, to a multi-unmanned aerial vehicle path planning method for transmission line inspection. Background Art

[0002] With the continuous advancement of the construction of China's smart grid and the increasing expansion of the scale of the transmission line network, the demand for efficient and accurate inspection of transmission lines has increased sharply. The traditional inspection mode of transmission lines mainly relies on manual climbing, ground telescope observation or fixed on-line monitoring equipment. These methods not only have inherent bottlenecks in inspection efficiency, cost control and coverage breadth, but also pose severe safety challenges to inspection personnel when facing complex terrains (such as mountains and rivers) or extreme weather conditions, and even the tasks cannot be carried out. Unmanned aerial vehicles (UAVs), that is, drones, have become an indispensable important tool in the field of transmission line inspection due to their excellent maneuverability and high-efficiency data acquisition capabilities. By carrying professional payloads such as high-definition cameras, infrared thermal imagers and lidar, drones can efficiently complete key tasks such as defect identification of transmission lines, abnormal temperature rise monitoring, and insulator pollution analysis, thus improving inspection efficiency while greatly ensuring operation safety. However, applying drones to large-scale transmission line inspection still faces a series of technical challenges, and the most core one is the path planning problem in complex environments. Especially in large-scale inspections, multiple drones usually need to be deployed to cooperate, which poses extremely high requirements for the real-time performance, safety and cooperation efficiency of path planning.

[0003] Currently, most of the research on multi-UAV path planning in the existing technology focuses on the global optimal solution in a static environment or local obstacle avoidance strategies. For example, Traditional path planning algorithms such as the A* algorithm or Rapidly-exploring Random Trees (RRT) can find feasible paths under the condition of a given static map, but their core depends on the prior knowledge of the environment, and they have insufficient response capabilities to common dynamic and sudden obstacles (such as temporary construction cranes, floating objects or other aircraft) in transmission line inspection, making it difficult to perform real-time path replanning and having relatively large potential safety hazards. In addition, existing multi-UAV systems often adopt a centralized control architecture or a mode of preset fixed flight trajectories to execute tasks. Although this method is easy to manage, it leads to rigid task allocation and lagging flight strategies. Once the front environment changes or a certain UAV fails, the cooperation efficiency of the entire fleet will be greatly reduced. Especially in the close-range cooperation scenario that requires real-time interaction to avoid collisions, its response speed and decision-making flexibility are difficult to meet the millisecond-level safety requirements.

[0004] Therefore, existing technologies generally have defects such as insufficient adaptability to dynamic environments, slow response of multi-UAV collaborative collision avoidance mechanisms, and lack of ability to quickly model and adjust countermeasures to real-time environmental changes. This directly limits the application efficiency and safety and reliability of UAV clusters in complex and changeable transmission line inspection scenarios. Summary of the Invention

[0005] In order to overcome the limitations of the existing technology, according to one aspect of the present application, a multi-UAV path planning method for power transmission line inspection is provided, which includes: The first drone uploads its current drone state to the MEC server, and the MEC server synchronizes the current drone state of the first drone to other drones in the cluster; Real-time environmental perception is performed through the sensors of each drone to obtain real-time environmental perception results; In response to the real-time environment perception result being that an obstacle is detected, triggering a dynamic risk assessment to update a global risk matrix; and in response to the real-time environment perception result being that there are no accidents, loading and using the global risk matrix of the previous cycle; The MEC server calculates a global demand matrix based on the distance between each UAV and the target point, and synchronizes the global demand matrix to the first UAV; The onboard computing unit of the first UAV determines an optimal action based on the current UAV state, the global risk matrix, and the global demand matrix, and generates a short-term planning path based on the optimal action; The first UAV broadcasts its short-term planned path to the MEC server, wherein the MEC server performs collaborative conflict detection based on the short-term planned paths of each UAV; In response to the MEC server's collaborative conflict detection result indicating that there is no conflict, the first UAV executes the optimal action to fly along the short-term planned path.

[0006] Compared with the existing technology, the main purpose of the multi-UAV path planning method for transmission line inspection provided by this application is to overcome the defects of current UAV path planning technology that strongly relies on static environment models and has insufficient adaptability when facing sudden obstacles, and to solve the technical bottlenecks of traditional multi-UAV systems in collaborative operations, such as delayed collision avoidance response and low overall operating efficiency due to over-centralized control or rigid strategies.

[0007] Specifically, the core concept of this system is to provide a distributed collaborative path planning framework that deeply integrates edge computing with on-board intelligent decision-making of unmanned aerial vehicles (UAVs). Specifically, through a mobile edge computing (MEC) server as the core hub for information interaction, each UAV first uses its own sensors to continuously perceive the surrounding environment in real time, and the perception results are used to dynamically evaluate and update a globally shared risk matrix. At the same time, the MEC server calculates and distributes a global demand matrix that quantifies the task value based on the inspection task objectives and the real-time positions of the UAV cluster. Subsequently, the on-board computing unit of each UAV can independently and in real time fuse its own current state, the global risk matrix, and the global demand matrix, autonomously make decisions on the optimal short-term flight actions and generate a planned path. Most importantly, before the planned path is officially executed, it will first be submitted to the MEC server for unified collaborative conflict detection, and will only be executed after ensuring no collision risk.

[0008] By adopting the above technical concept, the present invention can achieve remarkable technical effects: it endows the UAV cluster with the ability to autonomously perceive and quickly avoid unknown or dynamic risks in the inspection environment, making the path planning no longer a rigid preset flight route, but a real-time evolving intelligent decision-making process. At the same time, based on the collaborative conflict detection mechanism of MEC, it ensures the safety of multi-UAV close-range operations in complex airspace, greatly improving the collaborative efficiency and robustness of cluster operations. Finally, this method can fundamentally improve the safety and reliability of the transmission line inspection task in a changing and complex environment, and achieve a higher level of automated and intelligent inspection. Brief Description of the Drawings

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a flowchart of a multi-UAV path planning method for transmission line inspection according to an embodiment of the present application.

[0011] Figure 2 It is another flowchart of a multi-UAV path planning method for transmission line inspection according to an embodiment of the present application. Detailed Embodiments

[0012] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Instead, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for exemplary purposes only and are not intended to limit the scope of protection of the present disclosure.

[0013] Therefore, in view of the technical deficiencies disclosed in the background art, the present application proposes a multi-UAV path planning method for transmission line inspection. Figure 1 FIG. is a flowchart of a multi-UAV path planning method for transmission line inspection according to an embodiment of the present application. Figure 2 FIG. is another flowchart of a multi-UAV path planning method for transmission line inspection according to an embodiment of the present application.

[0014] As Figure 1 and Figure 2 shown, the multi-UAV path planning method for transmission line inspection according to an embodiment of the present application includes: S110, the first UAV uploads its current UAV state to the MEC server, and the MEC server synchronizes the current UAV state of the first UAV to other UAVs in the cluster; S120, performing real-time environment perception through sensors of each UAV to obtain a real-time environment perception result; S130, in response to the real-time environment perception result detecting an obstacle, triggering a dynamic risk assessment to update the global risk matrix; and in response to the real-time environment perception result having no accidents, loading and using the global risk matrix of the previous cycle; S140, the MEC server calculates a global demand matrix based on the distance between each UAV and the target point, and synchronizes the global demand matrix to the first UAV; S150, the on-board computing unit of the first UAV determines an optimal action based on the current UAV state, the global risk matrix, and the global demand matrix, and generates a short-term planned path based on the optimal action; S160, the first UAV broadcasts its short-term planned path to the MEC server, where the MEC server performs collaborative conflict detection based on the short-term planned paths of each UAV; S170, in response to the collaborative conflict detection result of the MEC server indicating no conflict, the first UAV executes the optimal action to fly along the short-term planned path; and S180, after the first UAV executes the optimal action, the BMS reports the updated remaining energy value and updates it to the current UAV state.

[0015] In step S110, the first drone uploads its current drone state to the MEC server, and the MEC server synchronizes the first drone's current drone state with the other drones in the cluster. In a multi-drone path planning task for power line inspection, the initial stage of system operation includes uploading the drones and their current drone states, which represent their conditions, to the mobile edge computing (MEC) server. Subsequently, after receiving the current drone state of each drone, the MEC server synchronizes it with the other drones in the cluster. Here, the process is illustrated using the first drone as an example. The first drone uploads its current drone state, which represents its own condition, to the mobile edge computing (MEC) server. Subsequently, after receiving this state, the MEC server synchronously distributes the first drone's current drone state to the other drones in the cluster.

[0016] Through this state upload and synchronization mechanism, each drone in the swarm can maintain real-time knowledge of key information such as the location and energy levels of its neighboring drones. This global state sharing is a prerequisite for efficient and safe swarm operations. It fundamentally avoids issues such as flight path conflicts and uncoordinated task allocation that can arise from information occlusion, ensuring that each drone has an accurate understanding of the current swarm layout when making subsequent decisions.

[0017] In this embodiment, the current drone status is a structured data set that specifically includes the drone's location, energy status, risk radius, and identity. The location status is typically the three-dimensional coordinates (x, y, z) acquired by the onboard Global Positioning System (GPS) or Real-Time Differential Kinematics (RTK) module; the energy status is the remaining energy b provided by the battery management system (BMS); the risk radius r is a parameter that dynamically adjusts based on the risk level of the surrounding environment; and the identity is a unique identifier assigned to each drone. These data components together constitute a complete status package that comprehensively describes the drone's current condition.

[0018] That is, in the embodiment of the present application, the state of each drone at time slot t is represented by a six-tuple, where the three-dimensional coordinates (x, y, z) accurately describe the real-time position of the drone in the transmission line inspection space. The risk radius r is a dynamically changing parameter that automatically adjusts according to the risk level of the surrounding environment. The remaining energy b is the core state parameter that directly determines the drone's ability to continue operating. Assume that the system contains N drones, and the state of each drone at time slot t is represented as: .in represents three-dimensional coordinates, represents the risk radius, Indicates the remaining energy.

[0019] The specific execution process of this status interaction includes two actions: uploading and synchronization. The uploading process is initiated by the UAV actively. At the beginning of each decision-making cycle, its on-board computing unit collects data such as position, energy, risk radius, and identity identification from each subsystem, integrates them into a standard data packet, and sends it to the MEC server through a wireless communication module such as 5G / 6G. The synchronization process is dominated by the MEC server. After receiving the status data packet, it will update the cluster status table maintained internally and immediately distribute the information containing the latest status to all other UAVs in the cluster through broadcasting or multicasting, thereby ensuring the low latency and quasi-real-time nature of the status information transmission by virtue of the edge deployment characteristics of the MEC.

[0020] To facilitate better understanding, a running scenario is provided here: Suppose UAV A and UAV B are performing tasks together. At time t0, after UAV A is initialized, it acquires its current UAV status as {ID: "UAV A", position status: (120.1, 30.2, 150.0), energy status: 99%, risk radius: 5 meters}. UAV A immediately uploads this status packet to the MEC server, and the server immediately synchronizes this information to UAV B. Therefore, when UAV B performs its own path planning calculation, it already knows the precise position of UAV A and can handle it as a dynamic avoidance object in subsequent decisions, ensuring the effectiveness of cooperative collision avoidance from the source.

[0021] In step S120, real-time environment perception is performed through the sensors of each UAV to obtain the real-time environment perception result. It should be understood that this step is the prerequisite for realizing dynamic obstacle avoidance and path adjustment and is to perform real-time environment perception operations through the sensors of each UAV. The on-board sensors of the UAV act as the perception terminals of the system, ensuring that the environmental information on which the decision is based can truly and immediately reflect the actual situation of the physical world.

[0022] It should be understood that in complex inspection scenarios such as power transmission lines, there are a large number of unforeseen and dynamically changing factors that may vary over time, such as temporarily appearing construction machinery, flocks of birds or other flying objects. To cope with these emergencies, the drone needs to be able to switch from a static, pre-installed map-based planning mode to an intelligent working mode with dynamic and real-time response capabilities. Through high-frequency real-time environmental perception, the drone can timely capture all environmental changes outside the preset 3D model, providing the most direct and accurate raw data input for subsequent dynamic risk assessment and path replanning. In the embodiment of this application, the real-time environmental perception result is a structured judgment conclusion obtained after processing and comparison, which is mainly divided into two types. One is detecting an obstacle, and the other is without any accident. To obtain this result, the drone will utilize sensors such as on-board cameras and lidar (LiDAR). The camera is used to capture the texture and color information of the environment, while the lidar is used to accurately measure the three-dimensional spatial distance from surrounding objects and generate real-time point cloud data. The two complement each other's advantages and provide strong hardware support for environmental perception.

[0023] This perception process mainly includes two links: environmental scanning and change detection. In environmental scanning, the cameras and lidar carried by the drone will continuously scan the surrounding area, especially the space in front of the flight, generating real-time image streams and point cloud data. In change detection, the on-board computing unit will register and compare the obtained real-time data with a pre-loaded static high-precision 3D digital map containing all known fixed objects (such as iron towers and wires), so as to identify "incremental objects" that cannot be matched. If there are such incremental objects, the perception result is determined to detect an obstacle; otherwise, it is determined to be without any accident.

[0024] This process can be understood through the following scenario: The perception result of a drone at time t1 is without any accident because the point cloud data it scans completely matches the pre-stored iron tower model. At time t2, when it flies near the next iron tower, in the new frame of point cloud data obtained, in addition to the tower body contour, there is also a cluster of point clouds that have never been recorded in the static map, such as a crane that is working. The comparison algorithm immediately identifies this significant difference, thus determining the real-time environmental perception result at this moment as detecting an obstacle, and attaching the position and size information of the obstacle. This result will immediately trigger the subsequent risk assessment and path adjustment processes.

[0025] In step S130, in response to the real-time environment perception result detecting an obstacle, a dynamic risk assessment is triggered to update the global risk matrix; and in response to the real-time environment perception result having no surprises, the global risk matrix of the previous cycle is loaded and used. It should be understood that in the embodiments of the present application, conditional judgment and response according to the real-time situation are based on the real-time environment perception result obtained in the previous link to selectively update the global risk matrix. This step connects the "perception" and "decision-making" links, and its processing logic has clear branches: when the real-time environment perception result detects an obstacle, the system will trigger a dynamic risk assessment to update the global risk matrix; when the result has no surprises, the global risk matrix of the previous cycle can be directly loaded and used.

[0026] The logic of designing this conditional branch lies in efficiently balancing the computational overhead of the system and the real-time nature of decision-making. As a high-frequency cyclic decision-making process, path planning will cause a huge waste of computing resources and then affect the decision response speed if the risks of the entire space are recalculated in each cycle without discrimination. Therefore, only when there is a substantial change in the environment (i.e., an obstacle is detected), the computationally intensive risk assessment process is started. This on-demand computing mechanism can significantly improve the overall operating efficiency of the system while ensuring the highest safety response level.

[0027] The global risk matrix here is a three-dimensional array obtained by discretely grid-dividing the inspection airspace. Each element of this array corresponds to a spatial voxel, and its value represents the risk level of the voxel. The dynamic risk assessment is an event-driven algorithm process, which specifically refers to a set of calculation processes that are started by the system after perceiving a new obstacle to quantify the new risks brought by the obstacle and integrate them into the existing global risk matrix.

[0028] The specific logic of this step depends on the result of the real-time environment perception. If the result has no surprises, the system judges that the existing risk matrix is still valid. At this time, when the on-board computing unit of the drone performs subsequent path planning, it will directly reference the global risk matrix stored in the memory and used in the previous decision-making cycle without any update calculation. On the contrary, if the result detects an obstacle, the system will immediately start a set of dynamic risk assessment processes to update the global risk matrix to ensure that new discovered hazard sources can be taken into account in subsequent decisions.

[0029] When dynamic risk assessment is required, the process is broken down into several steps. First, the position information of the obstacle is obtained based on the perception result. Then, based on this position information, a Gaussian distribution risk model is applied to the obstacle to calculate an independent new risk field that only includes the influence range of this newly emerged obstacle. This model can scientifically quantify the risk intensity of the obstacle and its attenuation in space. Specifically, this application uses the Gaussian distribution to quantify the obstacle risk. For K obstacles, the risk exposure of the UAV i to the obstacle k can be expressed as: where represents the Euclidean distance, the value represents the maximum intensity of the risk, controls the influence range of the risk, and the distance square term ensures the rapid increase of the near-field risk.

[0030] Next, the system will deeply fuse and superimpose this new risk field with the original global risk matrix. To avoid the distortion of the result caused by simple numerical addition, this method uses the operation method of probability multiplication for fusion, and finally generates a brand-new global risk matrix that includes the latest hazard sources. Specifically, its logic is to first convert the risk value into a safety probability (1 - risk value), then multiply the safety probability of the original global risk field at a certain point by the safety probability of the new risk field at this point to obtain a joint safety probability, and finally subtract this joint safety probability from 1 to obtain the fused risk value at this point after update. This process can be expressed as: The global risk matrix realizes the superposition calculation of multi-obstacle risks through probability multiplication operation, which is more in line with the actual physical scenario. The global risk matrix is expressed as: Through this method, the system accurately draws the new risk field onto the original global risk map, generating a brand-new global risk matrix that includes the latest hazard sources and is used for the current decision-making cycle.

[0031] For example, a UAV is flying, and its initial global risk matrix only includes the risks of fixed transmission towers and wires. When its sensor detects an unrecorded engineering crane on the front flight path, it will trigger a dynamic risk assessment. After the MEC server obtains the three-dimensional coordinates of the crane, it generates an independent risk field with a Gaussian distribution for it. Subsequently, the MEC server performs a probability multiplication operation to fuse this "crane risk field" with the old global risk matrix currently used by the system, thus generating a brand-new global risk matrix that includes both fixed risks and newly added dynamic risks, and synchronizes this map to the UAV to ensure that it can actively avoid the crane.

[0032] In the step S140, the MEC server calculates a global demand matrix based on the distances between each UAV and the target points, and synchronizes the global demand matrix to the first UAV. It should be understood that in the method of the present invention, in order to guide the UAV cluster to efficiently execute tasks, the core lies in the MEC server calculating and synchronizing the global demand matrix.

[0033] Specifically, the MEC server calculates and generates a global demand matrix based on the real-time distances between each UAV in the cluster and all the target points to be inspected, and synchronizes it to each UAV in the cluster. The function of this step is equivalent to drawing a dynamic "gravity map" with a clear task value orientation for the UAVs in the three-dimensional space. In this way, it is crucial to transform the abstract inspection task into a quantifiable reward signal that the agent can optimize. Without a clear target traction, even if the UAVs can avoid all risks, their flight trajectories will inevitably be blind and unable to efficiently cover all the key parts that need to be inspected. By constructing a global demand matrix that can reflect the attraction of each inspection target in real time, a clear and positive incentive can be provided for the reinforcement learning decision model of the UAVs, driving them to fly actively and strategically towards those target points with more urgent needs and higher values, thus ensuring the efficiency of the inspection task.

[0034] In the embodiment of the present application, the global demand matrix is a three-dimensional space discretized data structure corresponding to the inspection airspace. In this three-dimensional array, the values of the vast majority of voxels are zero, and only those voxels corresponding to the positions of the target points to be inspected (such as insulator strings, shock-proof hammers, etc.) will be assigned a non-zero positive value, and this value is the demand value. The demand value is a dynamically changing numerical score that quantifies the "benefit" or "value" that a specific UAV can obtain by flying to this target point for inspection at the current moment. Its size is usually related to the inherent priority of the target point and the distance between the UAV and it.

[0035] The generation and distribution process of this demand matrix can be divided into three links: calculation, integration, and synchronization. In the calculation link, the MEC server traverses all the target points to be inspected, and for each target point, calculates its current demand value through an improved Sigmoid demand function based on its inherent task priority and the distance relationship with the current UAV group. This function can well simulate the characteristic that the demand attraction of the target point will increase significantly only when the position of the UAV is close enough to ensure the inspection service quality.

[0036] More specifically, in this step, the MEC server first needs to traverse all the target points preset in the system that need to be inspected. Then, for each target point j in the list, the MEC server will independently calculate its demand value at the current moment t based on the distances between each drone in the current cluster and the target point. This application uses an innovative improved sigmoid demand function to perform this calculation, and its principle and process are as follows: The improved sigmoid demand function is adopted, which has the following advantages compared with the traditional linear model: 1) The parameters can be set differently according to the priorities of different power transmission equipment; 2) The exponential term can better simulate the non-linear relationship between signal strength and demand; 3) The adjustable parameters η and β enable the model to flexibly adapt to various inspection scenarios. The demand function of the end user is designed as: where represents the basic demand level, represents the signal strength, is an adjustable parameter.

[0037] Specifically, represents the signal strength, which in implementation is a function strongly related to the distance from the drone to the target point j. The closer the distance, the higher the signal strength , meaning the better the inspection service quality. And and are two adjustable parameters, where plays the role of the service quality threshold, then controls the steepness of the curve of the demand value changing with the service quality. Through this function, only when the position of the drone is good enough (i.e., ), the demand value of this target point will increase significantly, thus accurately simulating the demand characteristics of effective inspection.

[0038] After calculating the current demand values of all target points, the MEC server performs an integration operation. It will fill the demand value calculated for each target point into the voxel position corresponding to the three-dimensional coordinates of the target point in a three-dimensional array with all initial values being zero. After this operation is completed, this three-dimensional array constitutes the effective global demand matrix within the current decision cycle. The global demand matrix can be expressed as: As shown in the formula, the demand matrix adopts a threshold judgment mechanism: When the signal strength reaches the service radius When it is the case, the service demand is taken into account; otherwise, this point demand is ignored. This design effectively avoids the energy waste of ineffective services. Finally, in the synchronization step, the MEC server immediately distributes this matrix data containing the global task score distribution to all the drones in the cluster via the wireless network.

[0039] For example, assume that an inspection mission includes two target points: insulator A (with high task priority) and dampener B (with low priority). At time t, the MEC server combines the real-time position of the drone and finds that it is closer to insulator A. Thus, according to the demand function, the calculated demand value for point A is as high as 95.8; while it is still far from dampener B, and the calculated demand value is close to 0. Then, the MEC server generates a global demand matrix. In this matrix, only the spatial position corresponding to insulator A is assigned a high value of 95.8. After this "heat map" highlighting the position of point A is synchronized to the drone, its decision-making algorithm can clearly perceive that flying towards point A will obtain huge potential rewards, and thus tends to generate a flight path towards this point.

[0040] In the step S150, the on-board computing unit of the first drone determines the optimal action based on the current drone state, the global risk matrix, and the global demand matrix, and generates a short-term planned path based on the optimal action. That is, the on-board computing unit takes the collected current drone state, global risk matrix, and global demand matrix of itself as inputs to comprehensively determine an optimal action and generate a short-term planned path based on this optimal action. This step constitutes the "decision-making brain" of the individual drone, which can make a trade-off among risk avoidance, task pursuit, and consideration of its own state in real time and output an optimal action instruction.

[0041] This on-board autonomous decision-making ability integrates all the previously collected multi-dimensional situation information, completing the key leap from environmental perception to action decision-making. In the previous steps, the drone obtained three categories of information: its own state, environmental risk, and task value. The goal of this step is to comprehensively balance these three categories of information through the intelligent decision-making model installed on it, and calculate an optimal short-term action strategy that takes into account safety, efficiency, and energy consumption. This endows each drone with true autonomy and is the basis for realizing high-dynamic and intelligent collaborative flight planning.

[0042] It is worth mentioning that in the embodiments of this application, the optimal action generally refers to the action that is considered by the value evaluation model to be able to bring the maximum long-term cumulative reward from the set of all executable discrete actions of the drone in the current state (for example, flying a fixed distance in eight surrounding directions, or hovering in place). The short-term planned path is directly defined by this optimal action, that is, the trajectory between the current position point of the drone and the next position point to be reached after executing this action.

[0043] This process can be implemented on the terminal of the drone through a computational flow including input packing, model inference, and action selection. First, the on-board computing unit of the drone will pack its current state data, the received global risk matrix, and the global demand matrix into a unified format state tensor that meets the input requirements of the subsequent neural network model. This step ensures that all decision-making bases have been taken into account. Then, the on-board computing unit will input this packed state tensor into a locally deployed and well-trained lightweight deep Q-network (DQN) model. As a value evaluation function, this DQN model will quickly infer and calculate the long-term expected return (i.e., Q value) of executing each possible action in the current state. Its output result is a Q-value vector, such as [Q(S, a1), Q(S, a2),..., Q(S, an)], where each item corresponds to the value evaluation of a potential action.

[0044] After obtaining the vector containing the Q values of all optional actions, the on-board computing unit finally determines the optimal action according to the greedy strategy. It will directly select the action with the maximum Q value as the optimal decision in the current state. For example, if the inference result shows that the Q value of the action "fly forward" is the highest, then this action is selected as the optimal action. Immediately, the system can generate a short-term planned path from the current point to the target point based on this action. It should be noted that this path is only in the "planned" state at this stage and has not been executed.

[0045] More specifically, for a more substantial explanation of its process principle, this step models the drone path planning problem as a Markov decision process (MDP) to achieve intelligent decision-making in a dynamic environment. The core of the solution lies in designing a comprehensive state representation model that includes information such as the position coordinates, risk radius, and remaining energy of the drone, and using an improved sigmoid demand function to accurately quantify the service demand of the terminal user (TU). Through the double-network architecture of the deep Q-network (DQN), combined with the experience replay and priority sampling mechanisms, the algorithm can efficiently learn the optimal path strategy, and at the same time achieve information sharing and collision avoidance through the multi-drone cooperation mechanism.

[0046] The state space completely contains the UAV state, environmental requirements, and risk information. It is expressed as: The action space is designed as discrete end-user service selections, which is convenient for algorithm implementation. It is expressed as: The reward function innovatively integrates three key factors, including service demand satisfaction, risk aversion level, and energy consumption cost. It is expressed as: This application adopts a dual-network architecture to effectively solve the overestimation problem in Q-learning. Among them, the main network is responsible for action selection, the target network provides a stable learning target, and the soft update coefficient controls the update speed of parameters. The Q-function update rule is expressed as: The parameters θ^- of the target network are updated regularly: Policy optimization considers two aspects. On the one hand, it is to maximize the long-term reward, and on the other hand, the total energy consumption does not exceed the battery capacity. It can be transformed into an unconstrained optimization problem through the Lagrange multiplier method. Therefore, it can be expressed as: The implementation of its path planning algorithm is as follows: First is the calculation of the cost matrix: The piecewise linear model is used to calculate the cost matrix : Among them, Among them, each training episode contains a complete state-action-reward sequence. The experience replay mechanism improves the learning efficiency by breaking the temporal correlation, increasing the data utilization rate, and stabilizing the training process. The specific process is: First, initialize the cost matrix and the memory matrix ; Then, for each episode: a. Observe the current state ; b. Select an action ; c. Execute the action and obtain the reward and the new state ; d. Store To the memory bank ; e. Sample a minibatch from D to update the Q-network; Finally, update the target network once every C steps.

[0047] Imagine such a scenario: After the on-board computing unit of the first drone inputs its own state it has collected, a risk matrix indicating danger in the right front, and a demand matrix indicating a high-value inspection point directly ahead, its local deep Q-network infers a Q-value vector. This vector is likely to show that the Q-value of "fly to the right" is negative, while the Q-value of "fly forward" is a very high positive value. So, the drone selects "fly forward 1 meter" as the optimal action and generates a short-term planned path from the current position to the position 1 meter ahead, which is prepared for the next-stage collaborative conflict detection.

[0048] In the step S160, the first drone broadcasts its short-term planned path to the MEC server, where the MEC server performs collaborative conflict detection based on the short-term planned paths of each drone.

[0049] That is, after the drone determines the individual optimal short-term planned path, the method of the present invention also designs a key safety verification step to ensure the safety of cluster flight, that is: the drone needs to first broadcast its generated short-term planned path to the MEC server, and then the MEC server performs unified collaborative conflict detection based on all the short-term planned paths collected from each drone in the cluster. This centralized safety arbitration mechanism executed by the central node is the last line of defense to ensure the safety of multi-drone close collaborative operations in complex airspace.

[0050] The necessity of introducing this step lies in solving the potential path conflicts easily caused by the fully distributed decision-making architecture. Although each drone has considered the current positions of other drones during decision-making, since the decision-making process is almost simultaneous and independent in a very short time, it is entirely possible that drone A decides to fly to the right while the adjacent drone B decides to fly to the left, thus creating a fatal intersection point in the future path. Therefore, before each drone officially executes the flight action, the MEC server needs to act as the "God's perspective" to uniformly review the flight intentions of all drones, predict and prevent these possible collisions.

[0051] It should be understood that in the embodiments of the present application, collaborative conflict detection is a predictive geometric calculation process led by the MEC server. Its essence is to simulate whether the physical volumes (including a safety radius) of all drones will overlap in space in the short term in the future if they all fly according to the planned paths submitted by themselves. This conflict detection is not a simple judgment of geometric path intersection, but a four-dimensional spatio-temporal interference check of the flight envelopes of drones that includes the time dimension.

[0052] This process takes the MEC server as the calculation core and includes two links: path collection and conflict prediction. First is path collection. After each drone in the cluster generates its short-term planned path, it will immediately report it to the MEC server. The MEC server aggregates the path plans reported by all drones within the same decision cycle. Then comes the conflict prediction link. The MEC server will traverse all combinations of short-term planned paths. For example, it will check them in pairs and calculate the shortest distance between each pair of drones when they fly along their respective paths.

[0053] If the calculated shortest distance is less than a preset safety threshold (this threshold is greater than the physical size of the drone plus a certain safety margin), then the MEC server determines that there is a collaborative conflict between these two paths; otherwise, if after checking all path combinations, no case where the shortest distance is less than the safety threshold is found, it is determined that the action plan of the entire cluster has no conflict. The MEC server will feedback the global collaborative conflict detection result it obtains to the drone cluster.

[0054] For example, two drones A and B flying towards each other in a narrow passage. At the same moment, A plans a path of moving forward 2 meters, and B also plans a path of moving forward 2 meters. They both report this path to the MEC server. After the server aggregates the information, it conducts a simulation and calculates that the shortest distance between their paths will be 1 meter, which is much less than the preset safety threshold of 5 meters. Therefore, the MEC server will immediately determine that there is an impending collision risk between A and B. Correspondingly, the collaborative conflict detection result will be "there is a conflict", and the MEC will also send avoidance instructions to these two drones. For other drones C far from the conflict area, the result of "there is no conflict" will be returned.

[0055] In step S170, in response to the result of the collaborative conflict detection by the MEC server indicating no conflict, the first drone executes the optimal action to fly along the short-term planned path. As the final closed loop of the entire decision-making cycle, the method of the present invention clearly states that only after receiving and confirming the conflict-free feedback from the MEC server will the drone actually execute the actual flight action. This link can be specifically described as follows: in response to the result of the collaborative conflict detection by the MEC server indicating no conflict, the first drone will execute the optimal action previously determined by calculation to fly along the short-term planned path generated by it. This step sets a necessary safety prerequisite for all physical actions.

[0056] This "review before flight" mechanism implements the decision made after "seeing", "thinking", and "asking" into the "action" in the physical world. Without this link, all previous sensing and calculations would be meaningless. More importantly, by strictly binding the actual flight execution to the detection result of "no conflict", this method constructs a strong safety barrier in terms of mechanism, preventing any rash actions that are not collectively verified and may cause collisions between aircraft, and ensuring the safety of cluster actions at the physical execution level.

[0057] Here, the "result of the collaborative conflict detection indicating no conflict" is manifested at the drone side as a "safe passage" or "execution permission" signal with a logical value of true sent by the MEC server. And "executing the optimal action" is the process in which the on-board computing unit formally transmits a digital instruction such as "fly forward 1 meter" selected by it to the underlying flight control system of the drone. Finally, "flying along the short-term planned path" means that under the drive of the flight controller, the physical fuselage of the drone accurately completes the displacement according to the line segment trajectory.

[0058] Its specific implementation process is a "wait for confirmation and then execute" process. After uploading the planned path, the drone will briefly wait for the result of the collaborative conflict detection returned from the MEC server. Once the communication module receives the "execution permission" instruction indicating "no conflict", its on-board computing unit will immediately formally issue the temporarily stored optimal action (such as "fly forward 1 meter") to the flight controller. After receiving this high-level instruction, the flight controller will calculate a series of precise motor control commands to drive the drone body to complete this section of flight.

[0059] Continuing with the previous example, when the MEC server broadcasts an instruction of "execution permission", only UAV C will receive the "no conflict" result for itself, while UAV A and B will receive the "conflict exists" and adjustment instructions. Therefore, only UAV C will send the planned action to the flight control and execute it. UAV A and B will return to the decision-making step according to the adjustment instructions, regenerate a safe short-term planned path that avoids each other, and then go through the processes of reporting, detection, and execution again until a combination of actions that ensures the safety of all UAVs is found.

[0060] In another embodiment of the present application, multiple inspection UAVs achieve geometric information sharing through the MEC server to ensure spatial consistency; at the same time, share demand information to achieve task collaboration; also share risk information to ensure flight safety. Meanwhile, a cooperative collision avoidance strategy is adopted: the maximum value operation is used for the collision risk to ensure that the individual risk is not underestimated and the system safety is maximized. When other UAVs are detected, update: where is the collision risk threshold.

[0061] In step S180, after the first UAV executes the optimal action, the BMS will report the updated remaining energy value and update it to the current UAV state. After a complete decision-making and action execution, the system will also execute a crucial feedback update link. Specifically, after the first UAV completes the flight corresponding to its optimal action, the on-board battery management system (BMS) will accurately calculate the energy consumption caused by this action and obtain an updated remaining energy value.

[0062] This step is crucial for maintaining the operation safety of the entire system. Energy is the most core constraint resource of the UAV, and any consumption of it will directly affect subsequent mission planning and endurance. If the real energy consumption is not reflected in the system state in a timely manner, the subsequent decision-making model will calculate based on an outdated and inflated battery level value, which is very likely to cause the UAV to make a path plan beyond its return ability, introducing serious safety hazards. Therefore, immediately updating the energy state information after each action is the basis for ensuring the continuous effectiveness of decision-making and flight safety.

[0063] The Battery Management System (BMS) here is a precision electronic module embedded in the drone battery pack, responsible for real-time monitoring of battery data such as voltage and current, and accurately estimating the remaining energy of the battery based on this data. The updated remaining energy value output by this system is the most authoritative source of energy state. Subsequently, the on-board computing unit of the drone will write this latest value obtained from the BMS into and overwrite the energy state field in the current drone state data structure maintained in the memory.

[0064] In a specific embodiment, the battery management system uses the following energy consumption model for energy measurement, and its principle is as follows: Energy Model Fully considers three main energy-consuming links in drone operations: The flight energy consumption is proportional to the square of the moving distance and inversely proportional to the speed. It is expressed as: The hovering energy consumption depends on the communication quality and service requirements, and is expressed as: The computing energy consumption has a quadratic relationship with the complexity of the processing task, and is expressed as: The mechanism of this state update can be illustrated by a running instance. When a flight energy-consuming action is completed, the on-board computing unit of the first drone will query the BMS for the latest state of charge through the internal communication bus. For example, the energy before the drone action is 80%, and the action execution consumes power. After measurement by the BMS, the new state is 79.8%. The on-board computing unit then reads this value of 79.8% and uses it to update the energy item in the "current drone state" it maintains. Thus, at the beginning of the next decision cycle, the drone uploads to the MEC server the state information containing the latest energy value of 79.8%, ensuring that all subsequent decisions of the entire collaborative system are based on the most accurate battery life parameters.

[0065] The above is only a preferred embodiment of the present invention, and it is not intended to limit the present invention in any form. Any simple modification made by those skilled in the art to the technical solution of the present invention by means of equivalent replacement or equivalent transformation within the scope of all technical contents recorded in the technical solution of the present invention falls within the protection scope of the technical solution of the present invention.

Claims

1. A multi - UAV path planning method for transmission line inspection, including: The first UAV uploads its current UAV state to the MEC server, and the MEC server synchronizes the current UAV state of the first UAV to other UAVs in the cluster; Performing real - time environment perception through sensors of each UAV to obtain real - time environment perception results; In response to the real - time environment perception result being that an obstacle is detected, triggering dynamic risk assessment to update the global risk matrix; And in response to the real - time environment perception result being without any accident, loading and using the global risk matrix of the previous cycle; The MEC server calculates the global demand matrix based on the distances between each UAV and the target points, and synchronizes the global demand matrix to the first UAV; The on - board computing unit of the first UAV determines the optimal action based on the current UAV state, the global risk matrix, and the global demand matrix, and generates a short - term planned path based on the optimal action; The first UAV broadcasts its short - term planned path to the MEC server, where the MEC server performs collaborative conflict detection based on the short - term planned paths of each UAV; In response to the collaborative conflict detection result of the MEC server being that there is no conflict, the first UAV executes the optimal action to fly along the short - term planned path.

2. The multi-UAV path planning method for transmission line inspection according to claim 1, wherein The current UAV state includes position state, energy state, risk radius, and identity identification.

3. The multi-UAV path planning method for transmission line inspection according to claim 1, wherein, Performing real - time environment perception through sensors of each UAV to obtain real - time environment perception results, including: Scanning the surrounding environment through the camera and lidar of the UAV to obtain real - time point cloud data and real - time images; Comparing the real - time point cloud data and the real - time images with a pre - loaded static environment model to obtain the real - time environment perception results.

4. The multi-UAV path planning method for transmission line inspection according to claim 3, wherein, In response to the real - time environment perception result being that an obstacle is detected, triggering dynamic risk assessment to update the global risk matrix, including: Obtaining the position information of the obstacle; Applying a Gaussian distribution risk model to the obstacle based on the position information of the obstacle to obtain a new risk field; Fusing and superimposing the new risk field with the original global risk matrix to update the global risk matrix.

5. The multi-UAV path planning method for transmission line inspection according to claim 1, wherein, The MEC server calculates the global demand matrix based on the distances between each UAV and the target points, and synchronizes the global demand matrix to the first UAV, including: The MEC server traverses all target points to be inspected, and calculates its current demand value for each target point based on the distances between each UAV and the target point; Integrating the demand values of all target points to obtain the global demand matrix; and Synchronizing the global demand matrix to the first UAV.

6. The multi-UAV path planning method for power transmission line inspection according to claim 1, wherein The on - board computing unit of the first UAV determines the optimal action based on the current UAV state, the global risk matrix, and the global demand matrix, and generates a short - term planned path based on the optimal action, including: Inputting the current UAV state, the global risk matrix, and the global demand matrix into a locally deployed deep Q - network to obtain the optimal action; and Generate a short-term planned path based on the optimal action.

7. The multi-UAV path planning method for transmission line inspection according to claim 6, wherein, Input the current UAV state, the global risk matrix, and the global demand matrix into the deep Q-network deployed on the on-board computing unit to obtain the optimal action, including: The deep Q-network performs forward propagation inference based on the current UAV state, the global risk matrix, and the global demand matrix to obtain a vector containing multiple Q-values; Select the executable action corresponding to the maximum Q-value from the vector containing multiple Q-values as the optimal action.

8. The multi-UAV path planning method for transmission line inspection according to claim 1, characterized in that The first UAV broadcasts its short-term planned path to the MEC server. Among them, the MEC server performs collaborative conflict detection based on the short-term planned paths of each UAV, including: The MEC server determines whether there is an intersection in the spatial position between the short-term planned path of another UAV and the first UAV and the distance is less than the safety threshold.

9. The multi-UAV path planning method for transmission line inspection according to claim 1, wherein The multi-UAV path planning method for power transmission line inspection further includes the step of: After the first UAV executes the optimal action, the BMS reports the updated remaining energy value and updates it to the current UAV state.

Images