Distribution line inspection route planning method based on cooperation of multiple unmanned aerial vehicles

Abstract

The invention provides a distribution line inspection route planning method based on cooperation of multiple unmanned aerial vehicles. The method is applied to the technical field of distribution line inspection, and comprises the steps that a graph theory model representing the operation environment of the unmanned aerial vehicle is constructed, nodes of the graph theory model represent key positions of the unmanned aerial vehicle, edges represent reachable paths of the unmanned aerial vehicle, and weights of the edges are dynamically adjusted according to environment sensing data; distributing a distribution line to be inspected to a plurality of unmanned aerial vehicles by adopting hierarchical planning, and generating an initial inspection route of each unmanned aerial vehicle; performing multi-factor collaborative optimization on the initial inspection route according to the position, cruising ability, task load, battery state and environmental condition of the unmanned aerial vehicle to obtain an optimized inspection route; and controlling the inner layer of the unmanned aerial vehicle based on an improved Markov decision so as to control the unmanned aerial vehicle to stably fly according to the optimized inspection route. According to the invention, route planning of the unmanned aerial vehicle in a complex environment is realized, and the intelligent level of routing inspection is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of distribution line inspection, and in particular to a distribution line inspection route planning method based on multi-UAV collaboration. Background Art

[0002] The distribution network is a vital public infrastructure, directly connecting millions of households and industries. As a crucial component of power transmission, distribution lines are crucial for ensuring the safe and stable operation of the network. However, traditional distribution line inspection methods rely primarily on manual inspections, which present significant challenges, particularly given the long lengths of distribution lines, complex terrain, and numerous power towers. In recent years, the widespread use of drones in power line inspections has significantly improved the quality and efficiency of inspections. Autonomous drone inspections in the power transmission sector are relatively mature, with two common approaches: laser point cloud scanning and manual point-of-interest annotation, and manual point planning. These methods have significantly improved the efficiency and quality of equipment inspections. However, the application of this technology in the distribution sector is still in its infancy. While drones have been used for distribution line inspections, their functionality is relatively limited, with limited historical data management and route import capabilities. The lack of unified collaborative planning for the coordinated use of multiple drones leads to uneven distribution of inspection tasks and repeated inspections, resulting in a waste of time and resources.

[0003] Therefore, it is necessary to provide a distribution line inspection route planning method based on multi-UAV collaboration to improve the efficiency and reliability of distribution line inspection under multi-UAV collaboration. Summary of the Invention

[0004] The present disclosure provides a distribution line inspection route planning method based on multi-UAV collaboration, which aims to solve the problems of uneven inspection task distribution, multiple repeated inspection areas, and high inspection time cost when multiple UAVs jointly carry out distribution line inspections.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The present invention provides a distribution line inspection route planning method based on multi-UAV collaboration, comprising:

[0007] Constructing a graph model representing the drone's operating environment, where nodes represent key locations of the drone, edges represent reachable paths for the drone, and edge weights are dynamically adjusted based on environmental perception data;

[0008] A hierarchical planning approach is used to allocate the power distribution lines to be inspected to multiple drones, generating the initial inspection routes for each drone.

[0009] Perform multi-factor collaborative optimization on the initial inspection route based on the UAV's position, endurance, mission load, battery status, and environmental conditions to obtain an optimized inspection route;

[0010] The inner layer of the UAV is controlled based on the improved Markov decision making method to control the UAV to fly stably according to the optimized inspection route.

[0011] In some preferred embodiments, the graphical model G=(V, E) is defined as follows:

[0012] V={v1,v2,...,v nn}

[0013] E={e ij |v i , v j ∈V, and v i With v j reachable}

[0014] Among them, G: graph theory model, representing the operating environment of the UAV; V: node set, representing the key positions of the UAV, nn represents the total number of nodes; E: edge set, representing the connection relationship between nodes, that is, representing the reachable path, e ij : From node v i to v j edge;

[0015] In a dynamic unknown environment, the edge weight w ij Dynamically adjust based on environmental perception data:

[0016] w ij =d ij +αr ij +βc ij

[0017] Among them, w ij :Edge e ij The weight of node v i to v j The comprehensive cost; d ij : From node v i to v j The Euclidean distance of r ij : From node v i to v j The risk value is the probability that there may be an obstacle; c ij : From node v i to v j Energy consumption cost; α, β: weight coefficients, used to balance the impact of different costs.

[0018] In some preferred embodiments, wherein the power distribution lines are assigned to multiple UAVs by using hierarchical planning, generating the initial inspection route for each UAV comprises:

[0019] generating the initial inspection route for each UAV; wherein the initial task assignment objective function Z is:

[0020]

[0021] wherein N: total number of UAVs; M: total number of task points; cost nm : distance or cost of UAV n to task point m; x nm : task assignment variable, if UAV n is assigned to task point m, then x nm = 1, otherwise 0.

[0022] In some preferred embodiments, wherein the initial inspection route is optimized by considering the location, endurance, task load, battery status and environmental conditions of the UAVs, to obtain an optimized inspection route, comprises:

[0023] wherein the path planning objective of the UAV in the inspection environment is to find the optimal path from the start point to the end point:

[0024]

[0025] wherein θ(Path): total cost of path Path; Path: a complete path from the start point to the end point, composed of a series of edges e ij ; w ij : weight of edge e ij , representing the comprehensive cost from node v i to v j .

[0026] In some preferred embodiments, wherein the initial inspection route is optimized by considering the location, endurance, task load, battery status and environmental conditions of the UAVs, to obtain an optimized inspection route, further comprises:

[0027] by iteratively optimizing the initial inspection route, to ensure that the inspection efficiency is maximized.

[0028] In some preferred embodiments, wherein the graph model G = (V, E) is used to update the weights of nodes and edges, comprises:

[0029] Node state update:

[0030] V i (t+1) = τ(v i (t), H i )

[0031] where v i (t): state of node v i at time t; H i : environmental information sensed by sensors or vision; τ(·): node state update function for dynamically adjusting node attributes according to sensed information;

[0032] Dynamic edge weight update:

[0033] w ij (t+1) = μ(w ij (t), o ij , h ij )

[0034] where w ij (t): weight of edge e ij at time t; o ij : new obstacle sensing data; h ij : constraint adjustment due to environmental changes; μ(·): weight update function.

[0035] In some preferred embodiments, the method comprises:

[0036] Drones share location information through a wireless network to achieve information intercommunication and collaborative work;

[0037] When a drone n is out of power or fails, an automatic backup task allocation strategy is triggered to allocate its unfinished tasks to other suitable drones, and the flight path is re-planned to ensure the completion of the total task amount.

[0038] In some preferred embodiments, the inner layer of the drone is controlled by combining the improved Markov decision to control the stable flight of the drone, comprising:

[0039] The classic Markov decision process is extended to multiple drones, and each drone selects an action according to the global state, the environment transitions to a new state according to the joint action, and each drone obtains a corresponding reward, and the goal is to maximize the cumulative reward, comprising:

[0040] The state space, action space, and reward function are defined by modeling the multi-drone system using this framework.

[0041] In some preferred embodiments, the inner layer of the drone is controlled by combining the improved Markov decision to control the stable flight of the drone, further comprising:

[0042] A state transition model of the environment is established to represent how the environment transitions to a new state given the global state and joint action;

[0043] Estimate the local value function and Q-value function for each drone to evaluate the long-term benefits and action merits in different states;

[0044] According to the current state of the drone and the task progress, dynamically adjust the strategy, reward function and state transition model to adapt to the dynamically changing environment and task requirements.

[0045] In some preferred embodiments,

[0046] Each drone selects an action based on its strategy according to the current global state;

[0047] The environment transitions to a new global state according to the joint action of all drones;

[0048] The drone obtains an immediate reward according to the new state and joint action, and updates its cumulative reward;

[0049] Use a greedy selection strategy or a value-based strategy update method to iteratively optimize the strategy of each drone; the strategy update considers the mutual influence between drones to ensure the synergy and efficiency of the overall system.

[0050] Compared with the prior art, the present disclosure has the following beneficial effects:

[0051] (1) The present application realizes the autonomous flight and route planning of the drone in a complex environment by constructing a graph theory model representing the operating environment of the drone and controlling the inner layer of the drone based on an improved Markov decision process, improves the intelligent level of the inspection, greatly reduces the dependence on professional pilots, reduces the work burden of the inspection personnel, and improves the inspection efficiency and safety.

[0052] (2) The present application uses a hierarchical planning method to distribute power distribution lines to multiple drones, realizes multi-drone cooperative operation, not only improves the inspection efficiency, but also enhances the comprehensiveness and accuracy of the inspection through information sharing and cooperative operation between drones.

[0053] (3) The edge weight in the graph theory model of the present application is dynamically adjusted according to the environmental perception data, which can adapt to environmental changes in real time. According to the position, endurance, task load, battery status and environmental conditions of the drone, the route length and task allocation are optimized in multiple factors to ensure the efficient completion of the inspection task.

[0054] (4) When the endurance of the drone is insufficient or a fault occurs, the present application can automatically trigger a backup task allocation strategy to allocate the unfinished task of the drone to other suitable drones and re-plan the route to ensure the completion of the total task amount. This emergency handling capability greatly improves the reliability of the drone inspection.

[0055] In summary, the application has achieved remarkable beneficial effects in improving the intelligent level of distribution network unmanned aerial vehicle inspection, realizing multi-unmanned aerial vehicle cooperative operation, enhancing the adaptability of the inspection environment, improving the emergency handling capability and reducing the dependence on manual operation, and provides a solution for multi-unmanned aerial vehicle full-process autonomous and intelligent inspection of distribution line.

[0056] It should be understood that the content described in the summary section is not intended to limit the key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0057] The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by describing in detail the following embodiments with reference to the attached drawings. The drawings are intended to better understand the present disclosure and do not limit the present disclosure. In the drawings, the same or similar reference numerals refer to the same or similar elements, and

[0058] Figure 1 is a flowchart of a power distribution line inspection route planning method based on multi-unmanned aerial vehicle cooperation provided by an embodiment of the present disclosure;

[0059] Figure 2 is a data flowchart of a power distribution line inspection route planning method based on multi-unmanned aerial vehicle cooperation provided by an embodiment of the present disclosure;

[0060] Figure 3 shows a block diagram of an exemplary electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0061] To make the purposes, technical solutions, and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present disclosure.

[0062] To make the purposes, technical solutions, and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present disclosure.

[0063] Figure 1 is a flowchart of a power distribution line inspection route planning method based on multi-unmanned aerial vehicle cooperation provided by an embodiment of the present disclosure, Figure 2 is a data flowchart of a power distribution line inspection route planning method based on multi-unmanned aerial vehicle cooperation provided by an embodiment of the present disclosure. Referring to Figure 1 and Figure 2As shown, the present invention provides a distribution line inspection route planning method 100 based on multi-UAV collaboration, comprising:

[0064] S110: Constructing a graph theory model representing the operating environment of the UAV, wherein the nodes of the graph theory model represent key locations of the UAV, the edges represent reachable paths of the UAV, and the weights of the edges are dynamically adjusted according to the environmental perception data;

[0065] This step S110 uses graph theory to represent the operating environment of the drone, constructs an environmental model through nodes and edges, and performs modeling in an environment where the obstacle positions are dynamically unknown and the degree of knowledge of the global environment is unknown.

[0066] Construction and dynamic adjustment of graph theory models:

[0067] (1) Definition of nodes and edges

[0068] Node determination: Key location nodes are generated by predefined distribution line tower coordinates and drone take-off and landing points, and obstacle boundary points are dynamically added in combination with sensor data. Node set V = {v1, v2, ..., v nn}, each node v i Contains three-dimensional coordinates (x i ,y i , z i ), where z i Determined by terrain elevation data from lidar scans.

[0069] Edge generation: If there is no obstacle between the two nodes (determined by visual inspection), then an edge e is generated. ij Obstacle detection uses the YOLOv5 algorithm to process drone camera data in real time and identify the outlines of obstacles such as trees and buildings. Specifically: The definition of the graph model G = (V, E) is as follows:

[0070] V={v1,v2,...,v nn}

[0071] E={e ij |v i , v j ∈V, and v i With v j reachable}

[0072] Where G: graph theory model, representing the UAV operating environment;

[0073] V: node set, representing the key locations of the UAV (such as mission points, boundary points of obstacles, or waypoints), nn represents the total number of nodes;

[0074] E: edge set, representing the connection relationship between nodes, such as reachable paths;

[0075] e ij : edge from node v i to v j , usually contains a weight (such as distance, energy consumption or risk cost).

[0076] In a dynamic unknown environment, the weight w ij of the edge needs to be dynamically adjusted according to the data perceived by the environment:

[0077] w ij = d ij + a r ij + b c ij

[0078] where w ij : weight of edge e ij , represents the comprehensive cost from node v i to v j ;

[0079] d ij : Euclidean distance (spatial distance) from node v i to v j ;

[0080] r ij : risk value from node v i to v j , i.e. the probability that there may be obstacles;

[0081] c ij : energy consumption cost from node v i to v j ;

[0082] a, b: weight coefficients for balancing the influence of different costs.

[0083] where the risk value r ij : based on historical inspection data and real-time sensor fusion calculation. The formula is:

[0084]

[0085] where q is an empirical coefficient, for example, according to practical experience, the value is 0.7;

[0086] Energy consumption cost c ij : considering the headwind resistance and load, determined by the following formula:

[0087]

[0088] where V 巡航 = 10 m / s is the cruising speed, P 基础= 200W, K = 0.5 is the energy consumption coefficient, w 负载 is the task device power, C d is the wind resistance coefficient (dimensionless), p is the air density (kg / m 3 ), A is the windward area (m2), v 风 is the wind speed (m / s).

[0089] S120: The power distribution line to be inspected is distributed to multiple unmanned aerial vehicles by adopting hierarchical planning, and initial inspection routes of the unmanned aerial vehicles are generated.

[0090] In the application, the unmanned aerial vehicle control is divided into two layers, namely, an inner layer and an outer layer. The outer layer involved in S120 and step S130 is responsible for decision-making and path planning, and the inner layer involved in step S140 is responsible for stable flight (to be described later). The two layers are separated to avoid conflicts of computing resources and ensure flight safety.

[0091] The path planning of the outer layer involved in step S120 needs to find a collision-free path in a complex environment, which involves problem space construction and search. The multi-unmanned aerial vehicle system faces challenges such as limited resources (such as battery life), environmental perception (obstacle perception), and combination problems, which are different from single-unmanned aerial vehicle path planning. The unmanned aerial vehicle uses auxiliary sensors (such as laser and visual sensors), visual odometry, and simultaneous localization and mapping technology to generate environmental knowledge to enhance environmental perception ability.

[0092] The power distribution line to be inspected is distributed to multiple unmanned aerial vehicles by adopting hierarchical planning, ensuring comprehensive line coverage and balanced tasks, generating initial inspection routes of the unmanned aerial vehicles, and making the initial routes reasonable. The initial task allocation objective function Z is:

[0093]

[0094] Wherein, N: the total number of unmanned aerial vehicles;

[0095] M: the total number of task points;

[0096] cost nm : the distance or cost of unmanned aerial vehicle n to task point m;

[0097] x nm : task allocation variable, if unmanned aerial vehicle n is allocated to task point m, then x nm = 1, otherwise 0.

[0098] Task allocation constraint condition:

[0099] Full coverage constraint: each task point m must be accessed by at least one unmanned aerial vehicle:

[0100]

[0101] Endurance constraint: the total flight distance of drone n should not exceed its maximum endurance L n :

[0102]

[0103] Solution algorithm: Hungarian algorithm is used to solve the above integer programming problem, and Gurobi optimizer is used to realize it.

[0104] Initial inspection route generation:

[0105] Based on A* algorithm to search the optimal path in three-dimensional grid map, the grid resolution is 1m x 1m x 1m, and the cost function is:

[0106] f(n) = g(n) + h(n) + θ · height penalty

[0107] Where g(n) is the actual cost, h(n) is the heuristic function (Euclidean distance), and θ = 0.2 is the height weight to prevent flying too low.

[0108] The design principle of height penalty is:

[0109] Safe height range: Set the minimum safe height h min of the drone (such as 15 meters from the ground) and the maximum allowed height h max (such as 120 meters limited by regulations).

[0110] Ideal cruising altitude: According to the inspection requirements (such as the focusing distance of the camera), define the best cruising altitude h ideal (such as 50 meters).

[0111] Penalty logic: Punish the flight that deviates from the ideal height, guide the route to be as close to h ideal as possible.

[0112] The height penalty is quantified by a piecewise function, the formula is:

[0113]

[0114] Where h n : the current altitude of the drone n flying (unit: meters). k1, k2: penalty coefficients, usually k1 > k2, because the risk of low altitude is higher (such as trees, buildings), it needs stronger constraints. In the embodiment of the invention, for example, take k1 = 0.5, k2 = 0.3, which can be optimized through experiments.

[0115] Optionally, in complex terrain (such as valleys), k1 can be temporarily increased to strengthen low-altitude avoidance. Task requirement adjustment: if high-definition shooting is required, h ideal can be reduced to 30 meters, and k2 can be increased to suppress high-altitude flight.

[0116] S130: Multi-factor collaborative optimization of the initial inspection route according to the position, endurance, task load, battery status, and environmental conditions of the UAV, to obtain an optimized inspection route, including:

[0117] S131: Multi-factor collaborative optimization and dynamic adjustment

[0118] Multi-factor collaborative optimization of the initial inspection route according to the position, endurance, task load, battery status, and environmental conditions of the UAV, to obtain an optimized inspection route, including:

[0119] When the power of a certain UAV is lower than a certain threshold, adjust its route to reduce the flight distance, and reasonably distribute part of its tasks to other UAVs with sufficient power;

[0120] Wherein, the goal of path planning of the UAV in the inspection environment is to find the optimal path from the starting point to the end point:

[0121]

[0122] Wherein, θ(Path): the total cost of the path Path;

[0123] Path: a complete path from the starting point to the end point, composed of a series of edges e ij ;

[0124] w ij : the weight of the edge e ij , representing the comprehensive cost from node v i to v j .

[0125] Through iterative optimization of the initial inspection route planning, the inspection efficiency is maximized, and in each iteration, the advantages and disadvantages of the route are re-evaluated according to the latest environmental information and UAV state information, and the route planning strategy is adjusted.

[0126] In a specific embodiment, when the power is low, the following task redistribution is performed: when the power of the UAV is lower than 20%, the following steps are triggered: task stripping: add the unfinished task points to the global task pool. Allocation nearby: calculate the distance of other UAVs to these task points, and allocate them in order of proximity until the remaining endurance is sufficient for return. Path re-planning: use the dynamic A* algorithm to update the route for the UAV receiving the task.

[0127] In a specific embodiment, when the environment mutates, the following processing process is performed: if the wind speed is detected to be greater than 15 m / s, all UAVs switch to low-altitude mode (height < 50 meters), and use the RRT* algorithm to quickly generate a wind-avoiding path.

[0128] To adapt to the inspection environment, the graph model G=(V, E) is used to update the weights of nodes and edges, including:

[0129] (1) Node state update:

[0130] V i (t+1) = τ(v i (t), H i )

[0131] where v i (t): the state of node v i at time t; H i : environmental information sensed by sensors or vision; τ(·): node state update function, used to adjust node attributes according to sensed information, for example, when a new obstacle is detected by a laser radar, a new node is inserted every 5 meters at its boundary, and nodes that are cleared of obstacles are deleted.

[0132] (2) Dynamic update of edge weights:

[0133] w ij (t+1) = μ(w ij (t), o ij , h ij )

[0134] where w ij (t): the weight of edge e ij at time t; o ij : new obstacle sensing data; h ij : constraint adjustment caused by environmental changes (such as sudden wind speed or no-fly zone); μ(·): weight update function, specifically,

[0135] S132: Emergency handling and backup strategy

[0136] Drones share location information through wireless networks to achieve information exchange and collaborative work;

[0137] When a drone n is out of power or fails, the backup task allocation strategy will be automatically triggered, assigning its unfinished tasks to other suitable drones and re-planning the route to ensure the completion of the total task amount.

[0138] S140: Control the inner layer of the drone based on the improved Markov decision to control the drone to fly stably according to the optimized inspection route.

[0139] Specifically, step S140 includes:

[0140] The classic Markov decision process is extended to multiple drones. The drones select actions based on the global state, and the environment transfers state based on the joint action. Each drone receives a corresponding reward. The goal is to maximize the cumulative reward, including:

[0141] This framework is used to model the multi-UAV system, define the state space (including global and local states), action space (discrete and the same for each UAV), and reward function (encourage exploration and full coverage, and penalize overlap and collision). The specific process includes the following:

[0142] a. System Modeling: The multi-UAV system is modeled as an extended Markov decision process, where each UAV acts as a decision-maker and selects an action based on the global state. The environment transitions to a new state based on the combined actions of all UAVs, and the UAVs receive rewards.

[0143] b. State space definition:

[0144] Global state space s: contains global information such as the position, speed, battery level, mission status, obstacle location, and environmental conditions of all drones;

[0145] s=(p1,V1,b1,r1...,p N , V N , b N , r N , o, e)

[0146] Among them, p n :The position coordinates of drone n (x n ,y n , z n );

[0147] v n : The velocity vector of UAV n, including (v x，n , v y，n , v z，n );r n : The current mission status of drone n (e.g., waiting for inspection, inspecting, completed);

[0148] o: obstacle position set;

[0149] e: Environmental conditions (such as wind speed, visibility). Local state space: For each drone, it includes its own position, speed, battery status, current mission point, and nearby obstacle information;

[0150] s n =(p n , v n , b n , r n , o n )

[0151] where b n : battery status of drone n, b n ∈ [0, 100%];

[0152] o n : obstacle information in the vicinity of drone n.

[0153] c. Action space definition:

[0154] Define a discrete action space for each drone, including but not limited to moving direction (up, down, left, right), speed adjustment, task switching, etc., and all drones have the same action space;

[0155] a n ∈ {up, down, left turn, right turn, constant speed forward, acceleration, deceleration, hover, task switching}

[0156] All drones have the same action space, which is convenient for unified processing and policy updating.

[0157] d. Reward function design:

[0158] Design a joint reward function R(s, a) to encourage drones to explore unpatrolled areas, achieve full coverage inspection, while punishing collisions and task overlap between drones;

[0159] The joint reward function considers the global state and evaluates the action of each drone, rewarding exploration of new areas and full coverage, and imposing negative rewards for collisions and task overlap; that is, the reward function R(s, a) encourages the following goals:

[0160] Exploration: reward drones for accessing unpatrolled areas.

[0161] Full coverage: reward for achieving maximum coverage of the inspection area.

[0162] Collision avoidance: punish drones for collisions.

[0163] Avoid task overlap: punish multiple drones for covering the same area.

[0164] For example: coverage reward: +10 for each new task point visited. Collision penalty: -20 if the distance to other drones is <5 meters. Repeat penalty: -5 / visit if entering an area already patrolled by other drones.

[0165]

[0166] R n (s, a n , a -n ) = Y1B n - γ2On - γ3Risk n

[0167] where R(s, a): joint reward function, evaluates the overall benefit of global action;

[0168] R n (s, a n , a -n ): local reward function of UAV n, depends on global state s, its own action a n and joint actions of other UAVs a -n (except n);

[0169] B n : coverage gain of UAV n (e.g. number of new covered task points);

[0170] O n : task repetition cost (e.g. number of task points already covered by other UAVs);

[0171] Risk n : collision risk cost;

[0172] γ1, γ2, γ3: weight factors, e.g. take γ1 = 1.0, γ2 = -0.5, γ3 = -1.0, to balance each reward objective.

[0173] The strategy of each UAV n is:

[0174] π n (a n | s) = P(a n | s)

[0175] π n (a n | s): policy probability distribution of UAV n to choose action a n in state s.

[0176] The joint value function represents the long-term benefit of all UAVs:

[0177]

[0178] where V(s): global value function, represents the expected cumulative reward of all future time steps starting from state s;

[0179] Expectation operator, used to calculate the expected value of a random variable;

[0180] δ∈[0, 1]: discount factor, measures the importance of future rewards, the closer to 1, the more important future rewards; the closer to 0, the system pays more attention to current rewards;

[0181] R(s n , a n，t ) : immediate reward function, obtained at state s n and action a t at time step t;

[0182] t : time step, used to accumulate rewards over multiple time steps;

[0183] s0: initial state.

[0184] The local value function of the UAV is:

[0185]

[0186] where V n，t (s) : local value function of UAV n, representing the expected cumulative reward of UAV n in future time steps by its actions a t starting from state s;

[0187] δ : discount factor, consistent with the definition in global value function;

[0188] R n，t (s n , a n ) : local immediate reward obtained by UAV n at state s n and action a N at time step t.

[0189] e. Decision process:

[0190] Each UAV selects an action based on the current global state according to its policy; the environment transitions to a new global state according to the joint action of all UAVs; the UAVs obtain the immediate reward according to the new state and joint action and update their cumulative rewards. Specifically, it includes:

[0191] (1) State perception:

[0192] Each UAV obtains the global state s through sensors and communication networks.

[0193] (2) Action selection:

[0194] According to the current policy π N (a N | s), each UAV selects an action a n .

[0195] The policy can be a greedy selection based on Q value or other optimization policy.

[0196] (3) State transition:

[0197] The state transition model is:

[0198] P(s' | s, a1, a2,..., an) : the transition probability from the current state s to the next state s', depending on the joint action of all UAVs. N

[0199] P(s' | s, a) : the conditional probability of transitioning from the current state s to the next state s', depending on the joint action of all UAVs.

[0200] s : the current global state, representing the environment information of the multi-UAV system, such as position, velocity, task status, obstacle position, and environmental conditions, etc.

[0201] Joint action set a = (a1, a2,..., an) : contains the moving direction (such as up, down, left, right), speed adjustment, task switching, etc. of each UAV. N

[0202] The environment transitions to a new state s' according to the joint action a = (a1, a2,..., an). N

[0203] (4) Reward acquisition:

[0204] Each UAV obtains an immediate reward R n (s, a n , a-n) according to the new state s' and the joint action a.

[0205] (5) Cumulative reward update:

[0206] Update the cumulative reward of the UAV, which is used to evaluate the long-term performance.

[0207] f. Policy update and optimization:

[0208] (1) Q-value update:

[0209] Each UAV updates the Q-value based on the joint action, for example, using the Q-learning algorithm to update the Q-value:

[0210]

[0211] Q n (s, a) : the Q-value of UAV n taking joint action a (including its own action and the actions of other UAVs) in state s; Q n (s, a) = 0, all Q-values are set to 0 initially, and then updated through iterative learning.

[0212] R n (s, a n , a-n) : the immediate reward function, representing the reward obtained by UAV n taking its own action a n and other UAVs taking actions a-n in state s; ​​​

[0213] delta: discount factor, for example, select delta = 0.9, measure the importance of future rewards;

[0214] P(s'|s, a): probability of transferring from the current state s to the next state s' under joint action a;

[0215] V n (s'): local value function of the next state s', used to measure the expected future cumulative reward of UAV n starting from s'.

[0216] Q value update formula:

[0217]

[0218] where learning rate alpha' = 0.1, discount factor gamma' = 0.9, and epsilon = 0.2 in the epsilon-greedy strategy.

[0219] (2) Strategy update:

[0220] Use greedy selection strategy or other value-based strategy update methods such as Q-learning to iteratively optimize the strategy of each UAV; the strategy update considers the mutual influence between UAVs to ensure the collaboration and efficiency of the overall system. The following takes the greedy selection strategy as an example:

[0221]

[0222] pi n (a n |s): the strategy of UAV n selecting action a n in state s, following the greedy selection.

[0223] Find the action a n that maximizes Q n (s, a -n , a n ).

[0224] Q n (s, a n , a -n ): Q value of UAV n under joint action (a n , a -n ), used to measure the pros and cons of actions.

[0225] (3) Dynamic adjustment mechanism:

[0226] According to the current state of the UAV (such as battery level, task progress) and environmental changes, dynamically adjust the strategy, reward function and state transition model. For example, when a UAV has low battery level, adjust its strategy to reduce flight distance and redistribute tasks.

[0227] In some preferred embodiments, wherein the inner layer control of the UAVs by combining improved Markov decision and reinforcement learning to control the UAVs to fly stably further comprises:

[0228] establishing a state transition model of the environment to represent how the environment transitions to a new state given a global state and joint action;

[0229] estimating a local value function and a Q-value function for each UAV to evaluate long-term rewards and action merits in different states;

[0230] dynamically adjusting the policy, reward function, and state transition model according to the current state of the UAVs and task progress to adapt to the dynamically changing environment and task requirements.

[0231] In some preferred embodiments, wherein,

[0232] each UAV selects an action according to its policy based on the current global state;

[0233] the environment transitions to a new global state according to the joint action of all UAVs;

[0234] the UAVs obtain an immediate reward according to the new state and joint action and update their cumulative rewards;

[0235] adopting a greedy selection policy or a value-based policy update method to iteratively optimize the policy of each UAV; the policy update considers the mutual influence between UAVs to ensure the synergy and efficiency of the overall system.

[0236] According to the above embodiments of the present disclosure, based on the graph theory-based environment representation and modeling, combined with environment perception and data updating capabilities, efficient path planning and adjustment are achieved in the case of dynamic unknown obstacles and insufficient global environment knowledge, ensuring the safety of UAV operations in complex environments. For the outer route planning of multiple UAVs, through the combination of hierarchical planning and multi-factor collaborative optimization, the endurance capability, task load, battery status, and environmental conditions of the UAVs are considered, emergency handling and backup strategies are set, and the strategy is dynamically adjusted to achieve task balance. For the inner control method of multiple UAVs, through the improved Markov decision, the UAVs can make dynamic decisions according to the global state, effectively avoiding collisions and task overlaps between UAVs while achieving the inspection goal (such as exploring un-inspected areas, full-coverage inspection, etc.). The present disclosure has achieved remarkable results in improving the intelligent level of distribution network UAV inspection, realizing multi-UAV collaborative operation, enhancing environmental adaptability of inspection, improving emergency handling capability, and reducing manual dependence, etc., providing a solution for autonomous and intelligent inspection of distribution network lines.

[0237] Figure 3 A schematic block diagram of an electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present disclosure described and / or claimed in this document.

[0238] The electronic device 300 includes a computing unit 301 that can perform various appropriate actions and processes in accordance with a computer program stored in a ROM 302 or a computer program loaded from the storage unit 308 into a RAM 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 can also be stored. The computing unit 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. An I / O interface 305 is also connected to the bus 304.

[0239] A plurality of components in the electronic device 300 are connected to the I / O interface 305, including an input unit 306, such as a keyboard, a mouse, and the like, an output unit 307, such as various types of displays, speakers, and the like, a storage unit 308, such as a magnetic disk, an optical disk, and the like, and a communication unit 309, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 309 allows the electronic device 300 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0240] The computing unit 301 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 301 performs various methods and processes described above, such as the power distribution line inspection route planning method based on multi-UAV coordination. For example, in some embodiments, the power distribution line inspection route planning method based on multi-UAV coordination can be implemented as a computer software program, which is tangibly embodied in a machine-readable medium, such as the storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 300 via the ROM 302 and / or the communication unit 309. When the computer program is loaded onto the RAM 303 and executed by the computing unit 301, one or more steps of the power distribution line inspection route planning method based on multi-UAV coordination described above can be performed. Alternatively, in other embodiments, the computing unit 301 can be configured to perform the power distribution line inspection route planning method based on multi-UAV coordination by any other appropriate means, such as by means of firmware.

[0241] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0242] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0243] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0244] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0245] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0246] The computer system can include clients and servers. This relationship can be. The servers are typically remote from the clients with the interactions between them occurring over a communication network. The relationship between a client and a server is one of client-server. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0247] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present disclosure. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technology disclosed in the present disclosure are achieved, and the present disclosure is not limited herein.

[0248] The specific embodiments described above are not intended to be limiting. One of skill in the art will understand that various modifications, combinations, sub-combinations, and alternatives can be made to the specific embodiments described above without departing from the scope of the disclosure. Any modifications, combinations, sub-combinations, and alternatives falling within the spirit and principles of the disclosure should be considered.

Claims

1. A distribution line inspection route planning method based on multi-UAV collaboration, characterized in that: include: Constructing a graph model representing the drone's operating environment, where nodes represent key locations of the drone, edges represent reachable paths for the drone, and edge weights are dynamically adjusted based on environmental perception data; A hierarchical planning approach is used to allocate the power distribution lines to be inspected to multiple drones, generating the initial inspection routes for each drone. Perform multi-factor collaborative optimization on the initial inspection route based on the UAV's position, endurance, mission load, battery status, and environmental conditions to obtain an optimized inspection route; The inner layer of the UAV is controlled based on the improved Markov decision making method to control the UAV to fly stably according to the optimized inspection route.

2. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 1 is characterized in that: in, The graphical model G=(V, E) is defined as follows: V={v1,v2,...,v nn } E={e ij |v i , v j ∈V, and v i With v j reachable} Among them, G: graph model, representing the operating environment of the UAV; V: node set, representing the key positions of the UAV, nn represents the total number of nodes; E: edge set, representing the connection relationship between nodes, that is, representing the reachable path, e ij : From node v i to v j edge; In a dynamic unknown environment, the edge weight w ij Dynamically adjust based on environmental perception data: w ij =d ij +αr ij +βc ij Among them, w ij :Edge e ij The weight of node v i to v j The comprehensive cost; d ij : From node v i to v j The Euclidean distance of r ij : From node v i to v j The risk value is the probability that there may be an obstacle; c ij : From node v i to v j Energy consumption cost; α, β: weight coefficients, used to balance the impact of different costs.

3. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 2 is characterized in that: in, The hierarchical planning is used to allocate the power distribution lines to be inspected to multiple drones, and the initial inspection routes of each drone are generated, including: The initial task allocation objective function Z is: Where N is the total number of drones; M is the total number of mission points; cost nm : The distance or cost from drone n to mission point m; x nm : Task assignment variable. If drone n is assigned to task point m, then x nm =1, otherwise 0.

4. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 3 is characterized in that: in, The initial inspection route is optimized by multi-factor collaboration based on the position, endurance, mission load, battery status and environmental conditions of the UAV to obtain an optimized inspection route. include: Among them, the path planning goal of the drone in the inspection environment is to find the optimal path from the starting point to the end point: Among them, θ(Path): the total cost of the path Path; Path: a complete path from the starting point to the end point, consisting of a series of edges e ij Composition; w ij :Edge e ij The weight of node v i to v j the comprehensive cost.

5. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 4 is characterized in that: in, The multi-factor coordinated optimization of the initial inspection route based on the position, endurance, mission load, battery status and environmental conditions of the UAV to obtain an optimized inspection route also includes: By iteratively optimizing the initial inspection route, the inspection efficiency is maximized.

6. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 5 is characterized in that: in, The graph model G = (V, E) is used to update the weights of nodes and edges including: Node status update: v i (t+1)=τ(v i (t),H i ) Among them, v i (t): node v at time t i Status; H i : Environmental information perceived by sensors or vision; τ(·): Node state update function, used to dynamically adjust node attributes based on perceived information; Dynamic update of edge weights: w ij (t+1)=μ(w ij (t),o ij ,h ij ) Among them, w ij (t): edge e at time t ij The weight of ij : New obstacle perception data; h ij : Constraint adjustment caused by environmental changes; μ(·): weight update function.

7. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 5 is characterized in that: in, The method comprises: Drones share location information through wireless networks to achieve information exchange and collaborative operations; When the drone n has insufficient battery life or fails, the backup task allocation strategy is automatically triggered to assign its unfinished tasks to other suitable drones and replan the route to ensure the completion of the total task volume.

8. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 1 is characterized in that: in, The improved Markov decision-making method is used to control the inner layer of the UAV to control the UAV to fly stably, including: The classic Markov decision process is extended to multiple drones. The drones select actions based on the global state, and the environment transfers state based on the joint action. Each drone receives a corresponding reward. The goal is to maximize the cumulative reward, including: This framework is used to model multi-UAV systems and define the state space, action space, and reward function.

9. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 8, characterized in that: in, The control of the inner layer of the UAV based on the improved Markov decision to control the stable flight of the UAV also includes: Establish a state transition model for the environment to represent how the environment transitions to a new state given a global state and joint actions; Estimate the local value function and Q-value function for each drone to evaluate the long-term benefits and actions in different states; According to the current state of the UAV and the progress of the mission, the strategy, reward function and state transition model are dynamically adjusted to adapt to the dynamically changing environment and mission requirements.

10. The method for planning a distribution line inspection route based on multi-UAV collaboration according to claim 8, characterized in that: in, Each drone selects an action according to its strategy based on the current global state; The environment moves to a new global state based on the joint actions of all drones; The drone receives an immediate reward based on the new state and joint action, and updates its cumulative reward; A greedy selection strategy or a value-based policy update method is used to iteratively optimize the strategy of each UAV; the strategy update takes into account the mutual influence between UAVs to ensure the coordination and efficiency of the overall system.

Images