Unmanned aerial vehicle adaptive inspection path planning and task scheduling method and system for new energy infrastructure

Abstract

The application discloses a kind of unmanned aerial vehicle adaptive inspection path planning and task scheduling method and system for new energy infrastructure, comprising: fusion multi-source perception data constructs dynamic environment model, generates bottom layer grid map;With bottom layer grid map as foundation constructs heterogeneous double-layer map, establishes the bidirectional mapping and consistency maintenance mechanism between double layers, and is associated with time window resource occupation attribute for upper node;Based on upper node map, four-dimension multi-objective optimization evaluation model is constructed, and the weight coefficient is dynamically adjusted after quantitative evaluation of system comprehensive energy efficiency, task execution efficiency, cluster load balancing and task priority;With double-layer map as input, optimization objective function and weight coefficient as criterion generate inspection path and task scheduling scheme;Combined with time window resource occupation attribute, execute cooperative scheduling mechanism, realize multi-unmanned aerial vehicle cooperative operation and respond to abnormal events.The application improves the adaptive ability and operation efficiency of unmanned aerial vehicle inspection under new energy infrastructure scene.

Description

Technical Field

[0001] This invention belongs to the field of UAV path planning and task scheduling technology, specifically relating to a method and system for adaptive inspection path planning and task scheduling of UAVs for new energy infrastructure. Background Technology

[0002] With the rapid development of new energy infrastructure, wind farms, photovoltaic power stations, and power transmission and transformation facilities are widely distributed in complex environments such as mountains, hills, and deserts. Traditional manual inspection methods suffer from low efficiency, high safety risks, and limited coverage, making it difficult to meet the operation and maintenance needs of large-scale new energy facilities. Drones, with their high flexibility, strong mobility, and ability to carry multiple sensors, have become a core tool for intelligent operation and maintenance of new energy infrastructure.

[0003] However, drone inspections in new energy infrastructure scenarios face the following technical challenges:

[0004] First, the uncertainty of complex environments affects flight safety. New energy facilities are mostly located in areas with large terrain undulations and variable weather conditions. At the same time, the strong electromagnetic interference generated by power transmission and transformation equipment can interfere with UAV navigation signals. Traditional path planning methods based on static maps are difficult to adapt to dynamic environmental changes, resulting in safety hazards such as collisions and loss of control.

[0005] Second, balancing path efficiency and task coverage is difficult under multi-objective constraints. UAV inspection needs to simultaneously meet the requirements of energy economy under limited endurance, comprehensive coverage of all equipment locations, and obstacle avoidance safety in multi-UAV collaborative operations. There are inherent contradictions among these three aspects, and existing methods are unable to achieve an effective trade-off among multiple objectives.

[0006] Third, it lacks adaptive scheduling capabilities under dynamic task requirements. During the inspection process, new tasks such as defect retesting and emergency repairs may occur, or abnormal events such as drone malfunctions and sudden weather changes may be encountered. Traditional fixed path planning and static task scheduling methods are slow to respond and cannot achieve dynamic task reallocation and real-time path adjustment.

[0007] To address the aforementioned issues, existing research has proposed using grid-based or topological graph methods for path planning and incorporating intelligent algorithms such as genetic algorithms and particle swarm optimization for task scheduling. However, existing methods are mostly designed for single objectives or simple environments, making it difficult to simultaneously solve dynamic modeling, multi-objective optimization, and cooperative scheduling problems in complex environments. Some studies have used the NSGA-III algorithm for multi-objective optimization, but it still has shortcomings in the adaptive capabilities of genetic operators, the environmental representation capabilities of map structures, and the dynamic response capabilities of scheduling mechanisms.

[0008] Therefore, how to achieve path planning and task scheduling for drone inspections in complex new energy infrastructure scenarios, while taking into account environmental adaptability, multi-objective trade-offs and dynamic response capabilities, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0009] This invention aims to solve the following technical problems existing in the current technology for drone inspection in new energy infrastructure scenarios: insufficient flight path safety due to dynamic changes in complex environments; difficulty in balancing path efficiency and task coverage under multi-objective constraints; lack of adaptive scheduling capability under dynamic task requirements, making it impossible to achieve dynamic task redistribution and real-time path adjustment.

[0010] In a first aspect, embodiments of this application provide a method for adaptive inspection path planning and task scheduling of unmanned aerial vehicles (UAVs) for new energy infrastructure, the method comprising:

[0011] A dynamic environment model is constructed by integrating multi-source sensing data to obtain terrain elevation, obstacle distribution, meteorological and electromagnetic interference data. The flight area is divided using the grid method to generate a bottom-level grid map.

[0012] A heterogeneous two-layer map is constructed based on the underlying raster map. The bottom layer is a raster map and the upper layer is a node map. A bidirectional mapping and consistency maintenance mechanism is established between the two layers, and a time window resource occupancy attribute is associated with each upper-layer node.

[0013] A four-dimensional, multi-objective optimization evaluation model is constructed based on the upper-level node map to quantitatively evaluate the system's overall energy efficiency, task execution efficiency, cluster load balancing, and task priority, and dynamically adjust the weight coefficients of each dimension.

[0014] Using the underlying grid map and the upper node map as input, and the multi-objective optimization objective function and weight coefficients as criteria, an optimization algorithm is used to generate the inspection path and task scheduling scheme for the UAV.

[0015] Based on the task scheduling scheme and combined with the resource occupancy attributes of the time window, a collaborative scheduling mechanism is executed to achieve collaborative operation of multiple UAVs and respond to abnormal task events through dynamic adjustment.

[0016] Secondly, embodiments of this application provide an adaptive inspection path planning and task scheduling system for unmanned aerial vehicles (UAVs) for new energy infrastructure, applied to the method described in the first aspect, the system comprising:

[0017] The environmental modeling module is used to integrate multi-source sensing data to build a dynamic environment model, acquire terrain elevation, obstacle distribution, meteorological and electromagnetic interference data, divide the flight area using the grid method, and generate a bottom-level grid map.

[0018] The map building module is used to build a heterogeneous two-layer map based on the underlying raster map. The underlying layer is a raster map and the upper layer is a node map. It establishes a bidirectional mapping and consistency maintenance mechanism between the two layers and associates time window resource occupancy attributes with each upper-layer node.

[0019] The optimization evaluation module is used to build a four-dimensional multi-objective optimization evaluation model based on the upper-level node map, to quantitatively evaluate the system's overall energy efficiency, task execution efficiency, cluster load balancing and task priority, and to dynamically adjust the weight coefficients of each dimension.

[0020] The path planning module is used to generate inspection paths and task scheduling schemes for UAVs by taking the underlying grid map and the upper node map as inputs, the multi-objective optimization objective function and weight coefficients as criteria, and using optimization algorithms.

[0021] The collaborative scheduling module is used to execute a collaborative scheduling mechanism based on the task scheduling scheme and the resource occupancy attributes of the time window, so as to realize the collaborative operation of multiple UAVs and respond to abnormal task events through dynamic adjustment.

[0022] Thirdly, embodiments of this application provide an electronic device, including:

[0023] processor;

[0024] Memory used to store processor-executable instructions;

[0025] The processor is configured to implement the UAV adaptive inspection path planning and task scheduling method for new energy infrastructure as described in the first aspect when executing the instructions.

[0026] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that instructs a device to execute the UAV adaptive inspection path planning and task scheduling method for new energy infrastructure as described in the first aspect.

[0027] Compared with existing technologies, this invention has the following advantages: By fusing multi-source perception data from airborne sensors and ground equipment to construct a dynamic environment model and adopting a heterogeneous two-layer map structure, it achieves refined representation and real-time updates of terrain, obstacles, meteorological, and electromagnetic interference information in complex environments, effectively improving the environmental adaptability and flight safety of UAV path planning. By constructing a four-dimensional multi-objective optimization evaluation model that includes system comprehensive energy efficiency, task execution efficiency, cluster load balancing, and task priority, and combining it with a dynamic weight adjustment strategy, it achieves joint optimization of path and task schemes under multiple constraints, balancing energy consumption economy, comprehensive coverage, and operational efficiency. By establishing a bidirectional mapping and consistency maintenance mechanism between the bottom-level grid and upper-level nodes, it associates the resource occupancy attributes of upper nodes with time windows, and combines a collaborative scheduling mechanism to realize the allocation of time window resources and conflict avoidance for multiple UAVs. It also has the ability to handle abnormal events such as new task response, fault task migration, and meteorological and electromagnetic adaptive adjustment, significantly improving the robustness and dynamic response capability of the scheduling system. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of a method for adaptive inspection path planning and task scheduling of unmanned aerial vehicles (UAVs) for new energy infrastructure, provided as an embodiment of this application.

[0029] Figure 2 This is an architecture diagram of an UAV adaptive inspection path planning and task scheduling system for new energy infrastructure provided in one embodiment of this application.

[0030] Figure 3 A schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0032] It should be noted that in the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0033] Based on the embodiments described in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] Example 1

[0035] Figure 1 This is a schematic flowchart illustrating an adaptive inspection path planning and task scheduling method for unmanned aerial vehicles (UAVs) for new energy infrastructure, provided as an embodiment of this application. Figure 1 As shown, an adaptive inspection path planning and task scheduling method for UAVs for new energy infrastructure includes:

[0036] S1. A dynamic environment model is constructed by fusing multi-source sensing data, acquiring terrain elevation, obstacle distribution, meteorological, and electromagnetic interference data. The flight area is divided using a grid method to generate a base-level grid map. This step is responsible for collecting and processing basic environmental data for the UAV's flight area. Through onboard sensors working in conjunction with ground equipment, information such as terrain undulations, obstacle locations, wind speed and direction changes, and the intensity of electromagnetic interference generated by power transmission and transformation equipment is acquired. The processed data is then filled into each cell of the base-level grid map, providing an environmental base map for subsequent path planning.

[0037] Specifically, in this embodiment, the process of fusing multi-source sensing data to construct a dynamic environment model includes:

[0038] S1.1 Terrain elevation and obstacle distribution are collected using airborne lidar and visual sensors, while meteorological parameters and electromagnetic interference intensity data around power transmission and transformation equipment are collected using ground-based weather stations and electromagnetic monitoring equipment. In this embodiment, the lidar and visual sensors mounted on the UAV collect terrain elevation data and obstacle distribution information of the flight area. The lidar measures ground distance by emitting laser beams, accurately acquiring terrain undulations and building outlines; the visual sensors use image recognition to assist in identifying obstacle types such as trees and poles. Simultaneously, weather stations and electromagnetic monitoring equipment are deployed on the ground. The weather stations measure wind speed, wind direction, and visibility, while the electromagnetic monitoring equipment collects electromagnetic interference intensity data generated by the power transmission and transformation equipment during operation, providing a basis for subsequent risk area delineation.

[0039] The electromagnetic monitoring equipment's data acquisition frequency is set to 10Hz, meaning electromagnetic interference intensity data is collected every 0.1 seconds. Data from ground-based fixed monitoring equipment is transmitted to the control center via a wired network, with a data delay of no more than 50 milliseconds. Data from airborne electromagnetic sensors is transmitted back via a UAV data link; in environments with strong electromagnetic interference, the data update interval may be extended to 0.5 to 1 second. When no data update is received from a sensor for 3 consecutive seconds, the confidence decay factor for that sensor is activated, and its fusion weight decreases exponentially according to the aforementioned formula.

[0040] S1.2. Multi-source data are spatiotemporally fused to generate a comprehensive environmental dataset encompassing both static and dynamic elements. In this embodiment, the multi-source data from UAVs and ground equipment are aligned and fused in both time and space dimensions. Due to differences in data acquisition frequencies and coordinate systems among different sensors, time alignment is first achieved through timestamp synchronization and data interpolation. Then, coordinate transformation unifies the airborne and ground data into the same spatial coordinate system. The resulting comprehensive environmental dataset includes both static environmental elements such as terrain and obstacles, and dynamic environmental elements that change over time, such as wind speed and electromagnetic interference, providing a data foundation for real-time updates of the environmental model.

[0041] In this embodiment, to address the issue of significant differences in measurement reliability among different sensors in complex environments, a dynamic weighted fusion algorithm based on confidence decay factor is proposed. The fusion weight depends not only on the current measurement accuracy but also on the historical confidence decay trend of the sensors, as detailed below:

[0042] ,

[0043] ,

[0044] in, Indicates the first Each sensor at time Dynamic fusion weights; For the first Each sensor at time The measurement variance is calculated from the noise level fed back by the sensor in real time; This is the information entropy gain term; the higher the measurement accuracy and the smaller the variance, the larger the value of this term. For the first The confidence decay coefficient of each sensor, ranging from 0.01 to 0.1, reflects the rate at which the sensor's measurement reliability decreases over time in harsh environments (such as strong electromagnetic interference or heavy fog). For the first The time interval between the last successful data acquisition by each sensor; The confidence decay factor is the exponential decay of the weight of the sensor when it has not been updated for a long time. These are the estimated environmental parameters after fusion (including terrain elevation, obstacle distance, wind speed, and electromagnetic interference intensity). For the first The original measurements from each sensor. The innovation of this formula lies in the fact that it introduces a sensor confidence attenuation mechanism into multi-source fusion for the first time, solving the data gap problem caused by GPS signal loss under electromagnetic interference. The total number of sensors participating in the fusion For sensor indexing, for all The molecules of each sensor are summed to ensure that the sum of all weights is 1, thus achieving normalization.

[0045] S1.3. Based on the fused data, the environmental area is divided into a flightable area, an obstacle area, and a risk area containing areas with strong electromagnetic interference and frequent gusts using the raster method. The division results are used as the initial attributes of the bottom raster map to provide a basis for risk avoidance in path planning.

[0046] In this embodiment, based on the fused comprehensive environmental data, a grid method is used to divide the entire flight area into three types. The grid method discretizes a continuous space into grid cells of the same size, with each grid independently labeled with its attributes. The flyable area refers to an area without obstacles and with suitable environmental conditions for flight; the obstacle area refers to an area with physical obstructions such as mountains or buildings, where drones are prohibited from entering; the risk area refers to an area without physical obstacles but with safety hazards such as strong electromagnetic interference or frequent gusts of wind. The division results are used as the initial attributes of each grid cell in the underlying grid map. During subsequent path planning, the algorithm automatically avoids obstacle and risk areas based on the grid attributes, thereby ensuring flight safety.

[0047] S2. Construct a heterogeneous two-layer map based on the underlying raster map. The bottom layer is a raster map, and the upper layer is a node map. Establish a bidirectional mapping and consistency maintenance mechanism between the two layers, and associate a time window resource occupancy attribute with each upper-layer node. This step builds a two-layer structure based on the underlying raster map: the bottom layer retains detailed raster information to describe specific terrain and obstacles, while the upper layer abstracts key nodes (such as wind turbine locations, towers, intersections, etc.) for task organization. A mutual mapping relationship is established between the two layers; changes in the bottom layer environment will be synchronized to the upper layer, and task allocation in the upper layer will also be reflected in the cost calculation of the bottom layer. At the same time, a time window resource is bound to each upper-layer node, recording which drone occupies the node during which time period, laying the foundation for multi-drone collaboration.

[0048] Specifically, in this embodiment, the method for constructing the heterogeneous two-layer map includes:

[0049] S2.1. Based on the underlying raster map, retaining its detailed spatial representation capabilities as the underlying layer. In this embodiment, the underlying raster map generated in the previous stage is used as the base layer of the two-layer map. The raster map discretizes continuous space into grid cells, each of which records detailed information such as terrain type, risk level, and electromagnetic interference intensity at that location, thus possessing a detailed spatial representation capability of the environment. The underlying layer retains this capability, providing detailed terrain and obstacle data support for subsequent path planning.

[0050] S2.2 Build a node map for task topology modeling on the bottom layer as the upper layer, where the nodes are mapped to key equipment locations and intersections in the new energy infrastructure scenario.

[0051] In this embodiment, a node map is constructed as the upper-level structure above the underlying raster map. The node map is an abstract network model that abstracts key locations in the environment as point-like nodes. These nodes map to actual key locations in new energy infrastructure scenarios, such as the foundation sites of wind turbines, concentrated areas of photovoltaic panels, transmission line towers, insulator strings, and intersections in road networks. The upper-level node map is used for task topology modeling, that is, describing the connection relationships and passage order between task points, facilitating subsequent task allocation and path optimization.

[0052] S2.3 Establish a bidirectional mapping relationship between the bottom-level grid cells and the upper-level nodes, so that changes in the bottom-level environmental attributes can be transmitted to the upper-level nodes, and changes in the upper-level task allocation can also be mapped to the bottom-level update passage cost.

[0053] In this embodiment, a bidirectional mapping relationship is established between the bottom-level grid cells and the upper-level nodes. Bidirectional mapping means that one or more grid cells at the bottom level can correspond to a node at the upper level, and conversely, a node at the upper level can correspond to a grid area at the bottom level. This mapping mechanism allows changes in the bottom-level environmental attributes to be propagated upwards to associated nodes. For example, when a grid detects a new obstacle or increased electromagnetic interference, its corresponding upper-level node will automatically update its risk status. Simultaneously, upper-level task allocation or node state changes can also be mapped downwards to the bottom level, updating the passage cost of the relevant grids. The passage cost refers to the energy or time cost required for a drone to pass through a grid. Through this bidirectional mapping, the bottom and upper levels always maintain data synchronization.

[0054] In this embodiment, the mapping between the bottom-level grid and the upper-level nodes draws on the concept of field theory in physics, treating each node as a gravitational source, and the grid membership degree is determined by the superposition of field strengths:

[0055] ,

[0056] ,

[0057] in, For upper-level nodes In the bottom grid The potential energy generated at that location; The node charge is proportional to the task priority and node type (wind turbine, tower, intersection); This represents the two-dimensional coordinate vectors of the grid center and the node region; The distance is Euclidean. This is the height attenuation coefficient, with a value ranging from 0.1 to 0.8; The elevation difference between the grid and the node; For reference, the altitude difference (the average of the maximum altitude difference in the region); For grid Belonging to nodes The probability of this is determined. For the first time, field theory is introduced into map mapping, solving the problem of inaccurate classification of traditional nearest neighbor mapping in terrain-undulating areas; the potential energy function considers both planar distance and altitude differences, making the node boundary division in mountainous environments more consistent with the actual travel logic of UAVs.

[0058] S2.4 Simultaneously, associate each upper-level node with a time window resource occupancy attribute, record the state of the node being occupied in different time periods, and provide a basis for conflict avoidance for multi-UAV collaborative operations.

[0059] In this embodiment, a time window resource occupancy attribute is associated with each upper-level node. A time window is a mechanism for recording the time period during which resources are occupied. Each node's time window records which drone occupied the node during which time period. When multiple drones need to pass through the same node sequentially, the system can clearly determine the node's occupancy status for each time period through the time window records. During subsequent collaborative operations, the system will perform conflict checks based on the estimated arrival time of each drone at the node, combined with the node's time window status. If a time period is already occupied, subsequent drones need to wait or adjust their paths, thereby achieving orderly passage between multiple drones and avoiding node resource conflicts.

[0060] The time window resource occupancy attributes are stored using a linked list structure. Each node maintains a linked list of time windows arranged in ascending order of start time. Each time window element contains the following fields: a unique identifier for the occupying drone, estimated arrival time, estimated departure time, and task priority level. When a new drone requests node resources, the system traverses the linked list to find the first free interval that can accommodate the drone's estimated occupancy duration. If no such interval exists, the system returns a failure and triggers a conflict resolution process.

[0061] Specifically, in this embodiment, the bidirectional mapping and consistency maintenance mechanism between the bottom-level grid cells and the upper-level nodes includes:

[0062] S2.5 Establish a forward mapping from the bottom-level grid cell set to the upper-level node set, and a reverse mapping from the upper-level node to the bottom-level grid cell set, so that changes in the environmental attributes of any grid cell can be propagated upwards to its associated nodes, and changes in the task assignment of any node can be mapped downwards to the grid cells it covers.

[0063] In this embodiment, a bidirectional mapping relationship is established between the bottom-level grid and the upper-level nodes. Forward mapping refers to the correspondence from the bottom-level grid cell set to the upper-level node set, i.e., determining which grid cells belong to which upper-level node; reverse mapping refers to the mapping from the upper-level node to the bottom-level grid cell set it covers, i.e., determining which grid areas a node corresponds to. Through these two mappings, when the environmental attributes of a bottom-level grid change, such as detecting a new obstacle or increased electromagnetic interference, this change is automatically propagated upwards to its associated nodes, updating the node's risk status. Conversely, when the task allocation of an upper-level node changes, such as a task being reassigned to another drone, this change is automatically mapped downwards to all grid cells covered by that node, updating the passage cost of these grids, i.e., the energy consumption or time cost required for a drone to pass through that area. This bidirectional transmission mechanism ensures that the bottom-level environmental information and the upper-level task information remain synchronized.

[0064] S2.6. Set up an incremental update mechanism. When the changes in the attributes of the underlying raster exceed the preset threshold, the recalculation of the attributes of the upper-level nodes will be automatically triggered. When the upper-level task allocation changes, it will be automatically mapped to the access cost and risk level of the underlying raster to update the relevant raster, thereby maintaining data consistency between the underlying and upper layers.

[0065] In this embodiment, an incremental update mechanism is implemented to maintain data consistency between the lower and upper layers. Incremental updates refer to partial updates only on the changed parts, rather than recalculating all data with every change, thereby significantly reducing computational overhead. Specifically, when the environmental attributes of the lower-level raster change beyond a preset threshold, the system automatically triggers the recalculation of the attributes of its associated upper-level nodes, such as reassessing the node's risk level or passage difficulty; when the upper-level task allocation changes, the system automatically maps it to the lower layer, updating the passage cost and risk level of the relevant rasters. Through this incremental update mechanism, the data of the lower-level raster map and the upper-level node map remain consistent, while avoiding frequent full calculations and improving system operating efficiency.

[0066] An update is triggered when the change in the environmental attributes of the underlying raster exceeds a preset threshold. The threshold ranges from 0.05 to 0.15, indicating that a recalculation is triggered when the relative change rate of the attribute exceeds 5% to 15%. A larger value (0.12~0.15) is used in scenarios with stable weather conditions to reduce unnecessary updates; a smaller value (0.05~0.08) is used in scenarios with variable weather or dynamic obstacles to improve update sensitivity.

[0067] S3. Based on the upper-level node map, construct a four-dimensional multi-objective optimization evaluation model to quantitatively assess the system's overall energy efficiency, task execution efficiency, cluster load balancing, and task priority, and dynamically adjust the weight coefficients of each dimension. This step is responsible for evaluating the quality of a task allocation scheme. Scoring is based on four angles: total power consumption of all drones, when the slowest drone finishes its flight, whether the workload among drones is balanced, and whether urgent tasks are prioritized. The scores for these four angles are not fixed but dynamically adjusted according to the optimization process to ensure that the focus changes at different stages.

[0068] Specifically, in this embodiment, the method for constructing the four-dimensional multi-objective optimization evaluation model includes:

[0069] S3.1 Based on the task topology represented by the upper-level node map, quantitative evaluation indicators are established from four dimensions: overall system energy efficiency, task execution efficiency, cluster load balancing, and task priority.

[0070] In this embodiment, based on the task topology described by the upper-layer node map, namely the connection relationships and distribution locations between each inspection task point, quantitative evaluation indicators are established from four dimensions. The system comprehensive energy efficiency dimension measures the total energy consumed by the entire drone fleet to complete the inspection task; the task execution efficiency dimension measures the speed at which all tasks are completed, specifically focusing on the time taken by the slowest drone to complete the task; the cluster load balancing dimension measures whether the task distribution among drones is even, avoiding situations where some drones are overloaded while others are idle; and the task priority dimension measures whether tasks of different urgency levels are prioritized. These four dimensions comprehensively evaluate the merits of a task scheduling scheme from the perspectives of energy consumption, time, balance, and urgency.

[0071] S3.2, where the overall system energy efficiency reflects the total energy consumption of all UAVs during the inspection process, the task execution efficiency reflects the longest time required for all UAVs to complete all tasks, the cluster load balancing reflects the degree of difference in task allocation among UAVs, and the task priority reflects the completion satisfaction of various inspection tasks according to their urgency.

[0072] In this embodiment, the specific meanings of the four dimensions are clarified. The system's overall energy efficiency is calculated based on the total energy consumption of all drones during flight, including energy consumption during constant speed flight, acceleration / deceleration, hovering, and turning. Lower total energy consumption indicates a more energy-efficient solution. Task execution efficiency is measured by the drone with the longest task completion time, as the completion time of the entire inspection task depends on the slowest drone; a shorter time indicates higher efficiency. Cluster load balancing calculates the difference in task load among drones, which can be flight distance or the number of task points. Smaller differences indicate a more balanced load, which is beneficial for overall resource utilization. Task priority categorizes inspection tasks into different levels based on urgency, with emergency repair tasks having the highest priority, followed by defect review tasks, and routine inspection tasks having the lowest priority. Completion satisfaction reflects whether high-priority tasks are executed within the specified time.

[0073] In this embodiment, load balancing uses Kullback-Leibler divergence to measure the difference between the actual load distribution and the UAV capability distribution:

[0074] , , ,

[0075] in, For load balancing metrics, the lower the value, the better; For the first The proportion of the actual task load undertaken by the drone; For the first The workload of a drone (a weighted sum of flight distance and mission points). Ideal load distribution ratio based on battery life; For the first The drone's endurance (remaining battery power or maximum flight time). Let be the total number of drones. Unlike traditional methods that aim for absolute average load balancing, this formula allocates tasks based on the differences in the endurance of each drone, with drones having longer endurance undertaking more tasks. When = hour =0, meaning that allocation based on ability is considered the most balanced state, which is more in line with the physical constraints of actual operation scenarios.

[0076] It should be noted that when a drone's battery life reaches zero (i.e., it runs out of power and cannot perform a mission), the corresponding... When the value is 0, the drone does not participate in task allocation, and its corresponding term is set to 0 in the KL divergence calculation; when =0 and When = 0, we define 0·ln(0 / 0) = 0.

[0077] S3.3. The quantitative indicators of the above four dimensions are weighted and integrated to form a multi-objective optimization evaluation model.

[0078] In this embodiment, the quantitative indicators of the four dimensions are assigned different weights according to their respective importance and then weighted and fused to form a unified multi-objective optimization evaluation model. Weighted fusion refers to multiplying the indicator value of each dimension by its corresponding weight coefficient and then summing the results to obtain a comprehensive score. Since the four dimensions have different dimensions and numerical ranges, they need to be normalized first to map them to the same numerical range. The weight coefficients are not fixed but dynamically adjusted during the optimization process based on the improvement of each dimension. Through this weighted and fused evaluation model, the system can provide a comprehensive score for any task scheduling scheme, comparing the merits of different schemes and thus guiding the optimization algorithm to search for the optimal solution.

[0079] Specifically, in this embodiment, the method for dynamically adjusting the weight coefficients of each dimension includes:

[0080] S3.4 Construct a weight vector consisting of system comprehensive energy efficiency weight, task execution efficiency weight, cluster load balancing weight, and task priority weight.

[0081] In this embodiment, the weight coefficients of the four dimensions are combined into a weight vector. The weight vector is an array containing four values, each corresponding to the importance of overall system energy efficiency, task execution efficiency, cluster load balancing, and task priority, respectively. All weight coefficients range from 0 to 1, and the sum of the four weights equals 1. By adjusting the values ​​of each component in the weight vector, the optimization algorithm can change the emphasis placed on different objectives. For example, when the weight of overall system energy efficiency is high, the algorithm will prioritize searching for energy-saving solutions; when the weight of task priority is high, the algorithm will be more inclined to prioritize urgent tasks. The initial values ​​of the weight vector can be preset according to actual business needs.

[0082] S3.5 During the optimization iteration process, the distribution of the weight vector is dynamically adjusted according to the improvement of the objective function of each dimension. Dimensions that improve faster are given lower weights to avoid premature convergence, while dimensions that improve slower are given higher weights to strengthen the search.

[0083] In this embodiment, during the iterative search of the optimization algorithm, the distribution of the weight vector is dynamically adjusted based on the improvement rate of the objective function values ​​for each dimension. The improvement rate of the objective function reflects the optimization progress of a certain dimension. Rapid improvement means that the dimension is already close to a better level, and further increasing the weight is not meaningful; on the contrary, it may cause the algorithm to fall into a local optimum prematurely, so a lower weight is assigned. Slow improvement means that the dimension still has a large optimization space, and the search effort in this direction needs to be strengthened, so a higher weight is assigned. Through this dynamic adjustment strategy of suppressing the strong and supporting the weak, the algorithm can continuously explore target dimensions that have not yet been fully optimized, avoiding the premature dominance of one target in the search direction at the expense of other targets, thereby achieving balanced optimization among multiple targets.

[0084] In this embodiment, the weight adjustment adopts a metaheuristic algorithm based on preference entropy, and the weights are updated exponentially according to the target improvement energy:

[0085] ,

[0086] ,

[0087] in, For the first The generation The weights of each objective (energy consumption / time / balance / priority), For the first The improvement energy of each objective is normalized to 0~1; For the first The generation The optimal value for each objective; The values ​​are extremely small positive numbers to prevent division by zero. The core innovation is assigning lower weights to dimensions that improve faster, which is the opposite of traditional methods that increase the weight of the fastest-improving objective. The design philosophy of this formula is to suppress the strong and support the weak, preventing the algorithm from prematurely locking onto easily optimized objectives at the expense of other objectives, thus achieving a balanced expansion of the Pareto front. Among these, improving energy... The larger the value, the slower the improvement of the target (or the greater the potential for improvement), corresponding to the exponential factor in the formula. This allows it to receive a larger weight increment.

[0088] S3.6 Simultaneously, adjust the step size of weight updates according to the degree of environmental change, so that the weights can adapt to the search needs of different optimization stages, thereby achieving a balance between global exploration and local development.

[0089] In this embodiment, the step size for weight updates is adjusted according to the degree of environmental change. The weight update step size refers to the magnitude of the change in weights each time they are adjusted; a larger step size results in faster weight changes, while a smaller step size results in smoother weight changes. When significant environmental changes occur, such as the detection of a new high-priority task, sudden changes in weather conditions, or significant changes in electromagnetic interference intensity, the system uses a larger step size to quickly adjust the weights, enabling the algorithm to adapt rapidly to the new environmental conditions. When the environment is relatively stable, a smaller step size is used for fine-tuning, avoiding drastic weight fluctuations that could affect the algorithm's convergence stability. Through this adaptive step size mechanism, the weight vector can adapt to the search needs of different optimization stages. In the early stages of the algorithm's operation, it focuses on global exploration to cover a wider solution space, while in the later stages, it focuses on local development to refine the search for the optimal solution, thereby achieving a dynamic balance between global exploration and local development.

[0090] Specifically, the initial step size is set to 0.1. When the number of new tasks exceeds 20% of the total number of current tasks, or the change in meteorological parameters exceeds the preset threshold, or the change in electromagnetic interference intensity exceeds 30%, it is determined that the environment has changed significantly. At this time, the step size is adjusted to 1.5 to 2 times the initial step size; otherwise, the step size is reduced generation by generation with a decay coefficient of 0.95 until the minimum step size of 0.02 is reached.

[0091] S4. Using the underlying grid map and the upper-level node map as input, and the multi-objective optimization objective function and weight coefficients as criteria, an optimization algorithm is used to generate the UAV's inspection path and task scheduling scheme. This step takes the map data, optimization objective function, and weight coefficients produced in the previous step as input, and calls an intelligent optimization algorithm to search for the optimal solution. The algorithm will try different task allocation methods and flight sequences, iteratively compare the advantages and disadvantages of each scheme, and finally output a set of inspection path and task allocation schemes that take into account energy consumption, time, balance, and priority.

[0092] Specifically, in this embodiment, the step of generating the UAV inspection path and task scheduling scheme using an optimization algorithm includes:

[0093] S4.1. An improved NSGA-Ⅲ algorithm is adopted as the optimization engine. The fine spatial environment information provided by the bottom-level grid map and the task topology provided by the upper-level node map are used as the input of the search space. The four-dimensional multi-objective optimization objective function and dynamic weight coefficients are used as the fitness evaluation criteria.

[0094] In this embodiment, an improved NSGA-III algorithm is used as the optimization engine. NSGA-III is an intelligent algorithm commonly used to solve multi-objective optimization problems. Its core feature is maintaining the diversity of the solution set through non-dominated sorting and reference point mechanisms. The algorithm's search space consists of two parts: first, the detailed spatial environment information provided by the bottom-level grid map, i.e., the terrain, obstacles, and risk attributes of each grid, used to constrain the feasibility of the flight path; second, the task topology provided by the upper-level node map, i.e., the connection relationships between each task node, used to plan the task execution order. The algorithm uses a four-dimensional multi-objective optimization objective function and dynamic weight coefficients as fitness evaluation criteria. That is, the comprehensive score of each candidate solution is determined by the weighted fusion result of four dimensions: energy consumption, time, balance, and priority. The higher the score, the better the solution.

[0095] S4.2 The algorithm adopts a two-layer hybrid coding method, where the first layer of coding represents the allocation relationship between the task and the UAV, and the second layer of coding represents the task execution order within each UAV.

[0096] In this embodiment, the algorithm employs a two-layer hybrid encoding method to represent a complete task scheduling scheme. The first layer is integer encoding, used to characterize the allocation relationship between tasks and drones. Each task corresponds to an integer value, representing the drone number responsible for executing that task. This encoding method clearly expresses which tasks are assigned to which drone. The second layer is permutation encoding, used to characterize the task execution order within each drone. For multiple tasks assigned to the same drone, they are arranged in a certain order, and the drone will visit each task point sequentially according to this arrangement. The two-layer hybrid encoding clearly separates the two decision problems of which drone executes which tasks and in what order each drone executes tasks, facilitating subsequent genetic operations and scheme evaluation.

[0097] S4.3 During the evolution process, the crossover probability and mutation probability are adaptively adjusted according to the population diversity status, and the task sequence is segmented and crossover is performed according to the region or equipment type to which the task belongs.

[0098] In this embodiment, during the algorithm's evolutionary iteration, the crossover and mutation probabilities are adaptively adjusted based on the population diversity status. Population diversity refers to the degree of difference between individuals in the current population; greater difference indicates a wider search range, while smaller difference indicates that individuals tend to be homogeneous. When population diversity is high, the algorithm appropriately reduces the crossover and mutation probabilities to retain superior individuals; when population diversity is low, the algorithm increases the crossover and mutation probabilities to introduce new search directions and avoid getting trapped in local optima. Simultaneously, the algorithm performs segmented crossover operations on the task sequence based on the task's geographical region or device type. That is, the task sequence is divided into multiple segments according to geographical region or device category, and crossover operations are performed within each segment, rather than performing a full crossover on the entire sequence. Segmented crossover can preserve superior segments of the task sequence from different regions, improving the quality of the generated offspring solutions.

[0099] Specifically, the segmentation criteria include: dividing task points into different spatial clusters based on geographical coordinates, with each cluster corresponding to an inspection sub-region, and the cluster radius set to 200 to 500 meters based on the coverage range of a single drone patrol; dividing tasks into wind turbine inspection sections, tower inspection sections, insulator inspection sections, and line corridor inspection sections based on equipment type, and isolating the task sequences corresponding to different equipment types from each other during cross-operations to preserve the professional operation sequences of each type of task.

[0100] In this embodiment, drawing on the clonal selection mechanism of the immune system, antibody concentration and a memory cell bank are introduced to regulate the probability of crossover mutation:

[0101] ,

[0102] ,

[0103] in, This represents the antibody concentration in the population; the more similar the individuals, the higher the concentration. The Hamming distance between the two volume codes (calculated after the hybrid coding is unified); Population size; For the first The probability of mutation in a generation; These are the minimum and maximum values ​​of the mutation probability (0.01 and 0.15). The concentration-induced intensity coefficient, with a value ranging from 2 to 5; The maximum similarity between the current best individual and the historical elites of the memory cell bank; Using a similarity index (>0.8), forced mutation is triggered; this is the first time that the concentration regulation of the immune system and the memory cell mechanism have been introduced into an evolutionary algorithm. High concentrations lead to a saturated increase in mutation probability to introduce diversity; excessively high memory similarity forces an increase in mutation to escape the gravitational trap of historical optimal solutions. This dual nonlinear regulation outperforms the linear adaptive strategy of the native NSGA-III algorithm.

[0104] S4.4 When a new task or environmental change is detected, a local path replanning mechanism is triggered, which optimizes and adjusts only the affected route segments, and finally generates an inspection path and task scheduling scheme that satisfies multiple objectives.

[0105] In this embodiment, a local path replanning mechanism is triggered when the system detects a new task or environmental change. Local path replanning refers to optimizing only the affected route segments instead of recalculating the entire path. For example, when a new defect retesting task is added, the system only adjusts the route segments related to the task location, while keeping other segments unchanged; when increased electromagnetic interference is detected in a certain area, the system only replans the route passing through that area. This local adjustment method avoids significant changes to the overall plan, reduces computational overhead, and provides a fast response time. Finally, the algorithm outputs a set of inspection paths and task scheduling schemes that satisfy multiple objective trade-offs, allowing decision-makers to select the most suitable scheme to execute based on their actual preferences.

[0106] The algorithm uses a dual-criteria for convergence, terminating iteration when either criterion is met: first, the rate of change of the Pareto front hypervolume index of the population over 20 consecutive generations is less than 0.01%, at which point the population is considered to have converged to a stable front; second, the iteration count reaches the preset maximum of 300 generations, at which point it is forcibly terminated. The hypervolume index is calculated using the worst-case values ​​of the objective functions in the current population as a reference point.

[0107] S5. Based on the task scheduling scheme and considering the time window resource occupancy attributes, a collaborative scheduling mechanism is executed to dynamically adjust and enable multi-UAV collaborative operations while responding to task anomalies. This step is responsible for implementing the scheme generated in the previous step. Based on the task sequence and scheduled arrival time of each UAV, and considering the time window occupancy status of each node, each UAV is allocated a dedicated time slot to pass through each node to avoid multi-UAV conflicts. During execution, environmental changes and UAV status are continuously monitored. In cases of sudden wind speed changes, increased electromagnetic interference, UAV malfunctions, or new urgent tasks, the flight speed is automatically adjusted, tasks are reassigned, or flight paths are changed to ensure the safe and orderly completion of the inspection operation.

[0108] Specifically, in this embodiment, the method for implementing the cooperative scheduling mechanism includes:

[0109] S5.1 Based on the task sequence and scheduled arrival time of each UAV in the task scheduling scheme, and combined with the time window resource occupancy attributes of the upper-level nodes, allocate exclusive time segments for each UAV to pass through each node.

[0110] In this embodiment, based on the task scheduling scheme generated by the optimization algorithm, the task execution order of each drone and the estimated arrival time at each node are obtained. Combining the time window resource occupancy attributes associated with each node in the upper-level node map, the system allocates exclusive time segments for each drone to pass through each node. An exclusive time segment means that only the allocated drone can use that node during a certain period; other drones are not allowed to enter during this time. In this way, each drone knows exactly when to pass through which node before departure, laying the foundation for subsequent multi-drone collaborative operations.

[0111] S5.2 Conflicts are determined by checking whether the time windows of different drones at the same node overlap. If a conflict exists, the drones that arrive later are either waited for or their paths are replanned according to the task priority.

[0112] In this embodiment, the system determines whether a conflict has occurred by checking whether the time windows of different drones overlap at the same node. Overlapping time windows mean that the time periods occupied by two or more drones intersect, implying that they will compete for the right to use the same node at the same time. If a conflict is detected, the system processes the later-arriving drones according to task priority. Lower-priority tasks must wait for higher-priority tasks to pass first, or the drone's path is partially replanned to bypass the conflict node and find an alternative route. This conflict detection and handling mechanism ensures that multiple drones operate in an orderly manner within the same space, avoiding collisions.

[0113] S5.3 During execution, continuously monitor the flight status and environmental changes of the UAV. When it is found that the UAV deviates from the predetermined route, the node occupancy time exceeds the limit, or a new high-priority task is added, trigger dynamic adjustment and reallocate the time window resources of the affected nodes to ensure the orderly operation of multi-UAV collaborative work.

[0114] In this embodiment, during execution, the system continuously monitors the drone's flight status and environmental changes. When a drone deviates from its planned flight path due to unforeseen circumstances, stays at a node for longer than the preset time window limit, or a new high-priority task is assigned from above requiring its execution, the system automatically triggers dynamic adjustments. The specific operation of dynamic adjustments involves reallocating the time window resources of the affected nodes. This may involve adjusting the order of subsequent tasks, modifying the flight paths of some drones, or rearranging the node's occupied time period, thereby ensuring that multi-drone collaborative operations always proceed in an orderly manner.

[0115] Specifically, in this embodiment, the method for dynamically adjusting to achieve multi-UAV collaborative operation and respond to mission anomaly events includes:

[0116] S5.4 When real-time meteorological data detects a sudden increase in wind speed or a decrease in visibility, the flight speed of the affected drone is automatically adjusted and the node occupancy time window is recalculated.

[0117] In this embodiment, when a ground weather station or airborne weather sensor detects a sudden increase in wind speed or a decrease in visibility due to fog, the system automatically adjusts the flight speed of the affected drone. This speed adjustment changes the time it takes for the drone to reach each node; therefore, the system needs to synchronously recalculate the time window occupied by the drone for all nodes to ensure that the adjusted time window does not conflict with the time windows of other drones in the group. Through this adaptive speed adjustment, the drone can continue to perform its mission while ensuring flight safety, without having to completely abort or return to base.

[0118] In this embodiment, all abnormal event responses are uniformly modeled as Markov decision processes, with the goal of minimizing long-term cumulative costs to select the optimal strategy.

[0119] ,

[0120] ,

[0121] in, The optimal policy function takes the current state as input and outputs the optimal action. The system state vector includes environmental parameters, UAV battery / location, and task queue. For the joint action vector (continue / wait / return / migrate); For the expectation operator, take the expectation of all possible future state paths; This is a discount factor, with a value ranging from 0.9 to 0.99. For instantaneous cost function; For drones At any moment Cumulative delays; This is the penalty coefficient for task failure, ranging from 10 to 100. For high-priority task sets and indexes; This is the task completion indicator (1 indicates completion). Here, i is the drone index, ranging from 1 to N.

[0122] The environmental state sub-vector in the state space includes: wind speed, electromagnetic interference intensity, and visibility level. The UAV state sub-vector includes: remaining battery percentage and coordinate position. The task queue state sub-vector includes: the priority (discrete variable, with values ​​of 1, 2, and 3) and remaining completion time for each incomplete task. The action space includes six discrete actions: continue flying along the current route, wait at a node for a specified time, reduce flight speed to a safe threshold, return to base to recharge, transfer the task to a designated UAV, and call up a backup UAV. State transitions are calculated jointly using the environmental prediction model and the UAV dynamics model.

[0123] Traditional anomaly response methods employ a greedy, case-by-case strategy (e.g., immediately reducing speed upon a sudden increase in wind speed). This method, however, models all anomalies—such as sudden wind speed changes, increased electromagnetic interference, drone malfunctions, and new urgent tasks—in a unified manner, considering the cumulative costs over multiple future steps when making decisions. For example, while reducing speed ensures safety, it may delay subsequent high-priority tasks; the MDP strategy weighs these factors to determine whether a detour is worthwhile. This is the core algorithmic innovation that distinguishes this method from existing technologies.

[0124] S5.5 When the electromagnetic detector detects increased interference in a local area, the task in that area should be reassigned to a UAV with anti-interference capabilities or the flight path should be adjusted to bypass the interference area.

[0125] In this embodiment, when the electromagnetic detector detects that the electromagnetic interference intensity in a local area exceeds the standard, the system marks that area as a high-risk zone. For drones originally scheduled to perform missions through this area, the system first determines whether there are drones in the fleet equipped with anti-interference enhancement modules. These drones employ shielding designs and anti-interference communication protocols, enabling them to operate normally in strong electromagnetic environments. If such drones exist, the system reassigns the mission in that area to a drone with strong anti-interference capabilities; otherwise, it adjusts the flight path, guiding the drones to bypass the interference area and complete the mission point detection via an alternative path.

[0126] S5.6 When a drone is unable to continue its mission due to insufficient power or equipment failure, its unfinished mission is automatically migrated to other idle or low-load drones, and the time window occupancy status of the relevant nodes is updated.

[0127] In this embodiment, when a drone is unable to continue its mission due to insufficient battery power to return to base or equipment failure, the system automatically initiates a mission migration mechanism. Mission migration refers to reassigning all unfinished missions of the drone to other drones in the fleet that are idle or have low current loads. Simultaneously, the system needs to update the time window occupancy status of all relevant nodes. Time windows previously occupied by the faulty drone need to be cleared, and newly assigned drones need to re-apply for the corresponding node's time window resources and adjust their flight paths according to the new allocation scheme.

[0128] S5.7 When a new high-priority task is added, the system assesses the existing resource gap, suspends low-priority tasks if necessary to release node resources, and calls on nearby backup drones to supplement transport capacity.

[0129] In this embodiment, when a new high-priority task such as emergency repair occurs, the system first assesses whether the resources of the existing drones can meet the needs of the task, including the available flight time of each drone, the distance from the current location to the task point, and the current load status. If the existing resources are insufficient, the system will suspend some low-priority daily inspection tasks that are currently being executed, release the node time window resources occupied by these tasks, and allocate the released drones to high-priority tasks. If the needs still cannot be met, the system automatically calls on nearby standby drones in the helipad to supplement the transport capacity, ensuring that emergency tasks can be responded to as soon as possible.

[0130] S5.8 All abnormal responses aim to minimize the overall task completion time delay, ensuring the robustness of the scheduling system and the continuity of task execution.

[0131] In this embodiment, all the responses to various abnormal events described above aim to minimize the overall task completion time delay. Specifically, when faced with multiple solutions, the system estimates the impact of each solution on the overall inspection plan completion time and selects the solution with the least delay. Through this globally optimal decision-making logic, the scheduling system maintains good robustness in the face of various emergencies, that is, the system's ability to resist interference and operate stably, while ensuring that the tasks being executed will not be interrupted on a large scale due to individual anomalies, thus guaranteeing the continuity of task execution.

[0132] The calculation method for the overall task completion time delay is as follows: Let the estimated completion time of all tasks in the original scheduling plan be... The actual completion time for all tasks during execution was The overall delay = - When multiple solutions are available, the system estimates the consequences of each solution separately. ,choose The minimum acceptable flight path was implemented. The estimation methods are as follows: for speed adjustment scenarios, the time increment of the adjusted flight segment is calculated based on the wind speed-speed relationship curve; for mission migration scenarios, the additional flight time is calculated based on the migration workload and the current position of the target UAV; for path replanning scenarios, the increased flight time is calculated based on the detour distance. All estimations are based on currently known environmental data and UAV performance parameters.

[0133] Example 2

[0134] like Figure 2 As shown, this application provides an architecture diagram of an adaptive inspection path planning and task scheduling system for UAVs for new energy infrastructure, which is applied to the adaptive inspection path planning and task scheduling method for UAVs for new energy infrastructure as described in Embodiment 1, including:

[0135] The environment modeling module 210 is used to integrate multi-source sensing data to construct a dynamic environment model, acquire terrain elevation, obstacle distribution, meteorological and electromagnetic interference data, divide the flight area using the grid method, and generate a bottom-level grid map.

[0136] The map building module 220 is used to build a heterogeneous two-layer map based on the underlying raster map. The underlying layer is a raster map and the upper layer is a node map. It establishes a bidirectional mapping and consistency maintenance mechanism between the two layers and associates time window resource occupancy attributes with each upper-layer node.

[0137] The optimization evaluation module 230 is used to build a four-dimensional multi-objective optimization evaluation model based on the upper-level node map, to quantitatively evaluate the system's overall energy efficiency, task execution efficiency, cluster load balancing and task priority, and to dynamically adjust the weight coefficients of each dimension.

[0138] The path planning module 240 is used to generate the inspection path and task scheduling scheme of the UAV by taking the bottom grid map and the upper node map as inputs, the multi-objective optimization objective function and weight coefficients as criteria, and using an optimization algorithm.

[0139] The collaborative scheduling module 250 is used to execute a collaborative scheduling mechanism based on the task scheduling scheme and the resource occupancy attributes of the time window, so as to realize the collaborative operation of multiple UAVs and respond to abnormal task events through dynamic adjustment.

[0140] Figure 3 This is an electronic device provided in one embodiment of this application. For example... Figure 3 As shown, the electronic device includes at least the following components: processor 301 and memory 300, communication interface 303, and bus 302.

[0141] In this embodiment of the application, memory 300 is used to store executable instructions of processor 301, which, when configured to execute instructions, implements the method as described in the first aspect.

[0142] In embodiments of this application, a computer-readable storage medium includes instructions that instruct a device to perform the method as described in the first aspect. For example, the instructions instruct the device to perform... Figure 1 The method is shown in the process steps.

[0143] In one embodiment of this application, the program operating in the electronic device may be a program that controls a central processing unit (CPU) or similar device to achieve the functions of the above-described embodiments of the present invention (a program that enables the computer to function). Information processed by these systems is then temporarily stored in random access memory (RAM) during processing, and subsequently stored in various ROMs such as read-only memory (FlashROM) and hard disk drives (HDDs), and read, corrected, and written by the CPU as needed.

[0144] It should be noted that a portion of the electronic device described above can also be implemented using a computer. In this case, the program for implementing the control function can be recorded on a computer-readable recording medium, and the program recorded on the recording medium can be read into the computer and executed.

[0145] It should be noted that the computer mentioned here refers to a computer built into an electronic device, employing hardware including an operating system and peripheral devices. Furthermore, computer-readable recording media refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage systems such as hard drives built into the computer.

[0146] Furthermore, computer-readable recording media can include: media that dynamically stores programs for short periods of time, such as communication lines used when transmitting programs via networks like the Internet or communication lines like telephone lines; and media that store programs for fixed periods of time, such as volatile memory inside a computer that serves as a server or client in this case. In addition, the aforementioned program can be a program used to implement the above-mentioned functions, or it can be a program that can implement the above-mentioned functions by combining them with programs already recorded in the computer.

[0147] Furthermore, the electronic device in the above embodiments can also be implemented as an assembly (system group) composed of multiple systems. Each system constituting the system group can possess some or all of the functions or functional blocks of the electronic device in the above embodiments. As a system group, it is sufficient to have all the functions or functional blocks of the electronic device.

[0148] Those skilled in the art should recognize that the above embodiments are only used to illustrate this application and are not intended to limit this application. Any appropriate changes and variations made to the above embodiments within the essential spirit and scope of this application fall within the scope of protection claimed in this application.

Claims

1. A method for adaptive inspection path planning and task scheduling of unmanned aerial vehicles (UAVs) for new energy infrastructure, characterized in that, Includes the following steps: A dynamic environment model is constructed by integrating multi-source sensing data to obtain terrain elevation, obstacle distribution, meteorological and electromagnetic interference data. The flight area is divided using the grid method to generate a bottom-level grid map. A heterogeneous two-layer map is constructed based on the underlying raster map. The bottom layer is a raster map and the upper layer is a node map. A bidirectional mapping and consistency maintenance mechanism is established between the two layers, and a time window resource occupancy attribute is associated with each upper-layer node. A four-dimensional, multi-objective optimization evaluation model is constructed based on the upper-level node map to quantitatively evaluate the system's overall energy efficiency, task execution efficiency, cluster load balancing, and task priority, and dynamically adjust the weight coefficients of each dimension. Using the underlying grid map and the upper node map as input, and the multi-objective optimization objective function and weight coefficients as criteria, an optimization algorithm is used to generate the inspection path and task scheduling scheme for the UAV. Based on the task scheduling scheme and combined with the resource occupancy attributes of the time window, a collaborative scheduling mechanism is executed to achieve collaborative operation of multiple UAVs and respond to abnormal task events through dynamic adjustment.

2. The method according to claim 1, characterized in that, The construction of a dynamic environment model by fusing multi-source sensing data specifically includes: The terrain elevation and obstacle distribution are collected by airborne lidar and visual sensors, and meteorological parameters and electromagnetic interference intensity data around power transmission and transformation equipment are collected by ground meteorological stations and electromagnetic monitoring equipment. Multi-source data is spatiotemporally fused to generate a comprehensive environmental dataset that covers both static and dynamic elements; Based on the fused data, the environmental area is divided into a flightable zone, an obstacle zone, and a risk zone containing areas with strong electromagnetic interference and frequent gusts using a grid method. The division results are used as the initial attributes of the underlying grid map to provide a basis for risk avoidance in path planning.

3. The method according to claim 1, characterized in that, The method for constructing the heterogeneous two-layer map specifically includes: Based on the underlying raster map, its fine spatial representation capability is retained as the underlying layer; A node map for task topology modeling is built on top of the bottom layer, where nodes are mapped to key equipment locations and intersections in the new energy infrastructure scenario. Establish a bidirectional mapping relationship between the underlying grid cells and the upper-level nodes, so that changes in the underlying environment attributes can be transmitted to the upper-level nodes, and changes in the upper-level task allocation can also be mapped to the underlying update passage cost. At the same time, each upper-level node is associated with a time window resource occupancy attribute, recording the state of the node being occupied in different time periods, providing a basis for conflict avoidance in multi-UAV collaborative operations.

4. The method according to claim 3, characterized in that, The bidirectional mapping and consistency maintenance mechanism between the bottom-level grid cells and the upper-level nodes specifically includes: Establish a forward mapping from the bottom set of grid cells to the top set of nodes, and a reverse mapping from the top set of nodes to the bottom set of grid cells, so that changes in the environmental attributes of any grid cell can be propagated upward to its associated nodes, and changes in the task assignment of any node can be mapped downward to the grid cells it covers. An incremental update mechanism is set up so that when the changes in the attributes of the underlying raster exceed a preset threshold, the recalculation of the attributes of the upper-level nodes is automatically triggered. When the upper-level task allocation changes, it is automatically mapped to the access cost and risk level of the underlying raster to update the relevant raster, thereby maintaining data consistency between the underlying and upper layers.

5. The method according to claim 1, characterized in that, The method for constructing the four-dimensional multi-objective optimization evaluation model specifically includes: Based on the task topology represented by the upper-level node map, quantitative evaluation indicators are established from four dimensions: system comprehensive energy efficiency, task execution efficiency, cluster load balancing, and task priority. Among them, the overall system energy efficiency reflects the total energy consumption of all drones during the inspection process, the task execution efficiency reflects the longest time required for all drones to complete all tasks, the cluster load balancing reflects the degree of difference in task allocation among drones, and the task priority reflects the completion satisfaction of various inspection tasks according to their urgency. The quantitative indicators of the above four dimensions are weighted and integrated to form a multi-objective optimization evaluation model.

6. The method according to claim 5, characterized in that, The method for dynamically adjusting the weight coefficients of each dimension specifically includes: Construct a weight vector consisting of system overall energy efficiency weight, task execution efficiency weight, cluster load balancing weight, and task priority weight; During the optimization iteration process, the distribution of the weight vector is dynamically adjusted according to the improvement of the objective function of each dimension. Dimensions that improve faster are given lower weights to avoid premature convergence, while dimensions that improve slower are given higher weights to strengthen the search. At the same time, the step size of weight updates is adjusted according to the degree of environmental change, so that the weights can adapt to the search needs of different optimization stages, thereby achieving a balance between global exploration and local development.

7. The method according to claim 1, characterized in that, The method of generating UAV inspection paths and task scheduling schemes using optimization algorithms specifically includes: An improved NSGA-Ⅲ algorithm is used as the optimization engine. The fine spatial environment information provided by the bottom-level grid map and the task topology provided by the upper-level node map are used as the input of the search space. The four-dimensional multi-objective optimization objective function and dynamic weight coefficients are used as the fitness evaluation criteria. The algorithm employs a two-layer hybrid coding method, where the first layer of coding represents the allocation relationship between tasks and drones, and the second layer of coding represents the task execution order within each drone. During the evolution process, the crossover and mutation probabilities are adaptively adjusted according to the population diversity status, and the task sequence is segmented and crossover is performed according to the region or device type of the task. When a new task or environmental change is detected, a local path replanning mechanism is triggered, which optimizes and adjusts only the affected route segments, ultimately generating an inspection path and task scheduling scheme that satisfies multiple objective trade-offs.

8. The method according to claim 1, characterized in that, The specific methods for implementing the collaborative scheduling mechanism include: Based on the task sequence and scheduled arrival time of each UAV in the task scheduling scheme, and combined with the time window resource occupancy attributes of the upper-level nodes, each UAV is allocated an exclusive time segment for passing through each node. Conflicts are identified by checking whether the time windows of different drones overlap at the same node. If a conflict exists, the drones that arrive later are either waited or re-planned according to their task priorities. During execution, the flight status and environmental changes of the drones are continuously monitored. When a drone deviates from the predetermined route, a node is occupied for too long, or a new high-priority task is added, dynamic adjustments are triggered to reallocate the time window resources of the affected nodes, ensuring the orderly operation of multi-drone collaborative work.

9. The method according to claim 8, characterized in that, The specific methods for achieving multi-UAV collaborative operation and responding to abnormal mission events through dynamic adjustment include: When real-time weather data detects a sudden increase in wind speed or a decrease in visibility, the flight speed of the affected drone is automatically adjusted and the node occupancy time window is recalculated. When the electromagnetic detector detects increased interference in a local area, the task in that area is reassigned to a drone with anti-interference capabilities or the flight path is adjusted to bypass the interference area. When a drone is unable to continue its mission due to insufficient power or equipment failure, its unfinished mission is automatically migrated to other idle or low-load drones, and the time window occupancy status of the relevant nodes is updated. When a new high-priority task is added, the system assesses the existing resource gap, suspends low-priority tasks if necessary to free up node resources, and calls on nearby backup drones to supplement transport capacity. All exception responses aim to minimize overall task completion time delays, ensuring the robustness of the scheduling system and the continuity of task execution.

10. A UAV adaptive inspection path planning and task scheduling system for new energy infrastructure, applied to the method described in any one of claims 1 to 9, characterized in that, The system includes: The environmental modeling module is used to integrate multi-source sensing data to build a dynamic environment model, acquire terrain elevation, obstacle distribution, meteorological and electromagnetic interference data, divide the flight area using the grid method, and generate a bottom-level grid map. The map building module is used to build a heterogeneous two-layer map based on the underlying raster map. The underlying layer is a raster map and the upper layer is a node map. It establishes a bidirectional mapping and consistency maintenance mechanism between the two layers and associates time window resource occupancy attributes with each upper-layer node. The optimization evaluation module is used to build a four-dimensional multi-objective optimization evaluation model based on the upper-level node map, to quantitatively evaluate the system's overall energy efficiency, task execution efficiency, cluster load balancing and task priority, and to dynamically adjust the weight coefficients of each dimension. The path planning module is used to generate inspection paths and task scheduling schemes for UAVs by taking the underlying grid map and the upper node map as inputs, the multi-objective optimization objective function and weight coefficients as criteria, and using optimization algorithms. The collaborative scheduling module is used to execute a collaborative scheduling mechanism based on the task scheduling scheme and the resource occupancy attributes of the time window, so as to realize the collaborative operation of multiple UAVs and respond to abnormal task events through dynamic adjustment.

Images