Method, device and equipment for optimizing patrol path of unmanned aerial vehicle group in chemical industry park

Abstract

The invention provides a chemical industrial park unmanned aerial vehicle group patrol path optimization method, device and equipment, and relates to the technical field of unmanned aerial vehicle patrol, and the method comprises the steps: obtaining the environment data of a target chemical industrial park, and the attributes and initial patrol paths of all unmanned aerial vehicles; aiming at any unmanned aerial vehicle, acquiring each patrol position on the initial patrol path and a real-time monitoring value of the corresponding patrol position; for any patrol target in charge of the unmanned aerial vehicle, obtaining the weight when the unmanned aerial vehicle arrives at the patrol target; inputting each patrol position and the corresponding real-time monitoring value, the target position and weight of the patrol target, the environment data and the initial patrol path into a pre-constructed multi-target optimization model, and calculating a score of the initial patrol path; according to the weight, the target position, the environment data, the attribute and the initial patrol path and the position of each unmanned aerial vehicle, determining an optimal patrol path of the unmanned aerial vehicle by adopting dynamic neighborhood search; according to the method, the optimal patrol path with high efficiency, safety, high quality and high coverage rate can be generated.

Description

Technical Field

[0001] The present application relates to the field of drone inspection technology, and more specifically, to a method, device and equipment for optimizing the inspection path of a group of drones in a chemical park. Background Art

[0002] As a concentrated area of ​​hazardous chemicals, the safety management of chemical parks has long been regarded as the focus and difficulty in the field of industrial safety. The traditional manual safety inspection model is difficult to fully adapt to the actual needs of modern chemical park safety management due to its low efficiency, limited coverage and many safety hazards. In recent years, with the rapid development of drone technology, this technology has provided a new solution for safety inspections in chemical parks. With its highly flexible operating performance, wide coverage and ability to carry a variety of sensors, drones have significantly enhanced the safety of inspections while improving inspection efficiency.

[0003] However, the operating environment of chemical parks is often very complex, facing a variety of potential risk factors such as fire hazards and toxic gas leaks. It is difficult to achieve the goal of comprehensive and efficient safety monitoring by relying solely on a single drone to perform inspection tasks. Therefore, the use of drone swarms for collaborative operations has gradually become a research hotspot in academia and industry. At the same time, how to design a reasonable drone swarm inspection path to ensure inspection efficiency while taking into account the risk control level and data collection quality has become an urgent and challenging issue. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a method, device and equipment for optimizing the inspection path of a chemical park drone swarm, so as to solve the above-mentioned problems existing in the prior art and generate an optimal inspection path that takes into account efficiency, safety, high quality and high coverage, so as to control the drone swarm to perform efficient and high-quality inspections.

[0005] In a first aspect, the present invention provides a method for optimizing a patrol path of a chemical park drone group, the method comprising: Obtain environmental data of a target chemical park, as well as the properties and initial patrol paths of each drone used to patrol the target chemical park; wherein the initial patrol path corresponding to any drone includes at least one patrol target that the corresponding drone is responsible for; for any drone, control the drone to perform its respective patrol tasks according to the corresponding initial patrol path, and obtain each patrol position on the initial patrol path and the real-time monitoring value of the corresponding patrol position; for any patrol target that the drone is responsible for, obtain the weight of the drone when it reaches the patrol target; and compare each patrol position and the corresponding real-time monitoring value, the target position of the patrol target and The weights, the environmental data and the initial patrol path are input into a pre-built multi-objective optimization model to calculate the score of the initial patrol path; based on the weights, the target position, the environmental data, the attributes and the initial patrol path and position of each UAV, a dynamic neighborhood search is used to determine the optimal patrol path of the UAV; if the score of the optimal patrol path is greater than the score of the initial patrol path, the optimal patrol path is used as the new initial patrol path, and the execution step is returned: for any UAV, the UAV is controlled to patrol according to the new initial patrol path until the UAV completes the patrol task.

[0006] In an optional embodiment, the attributes include: flight speed, endurance, communication range, safe flight range, maximum weight, minimum weight, maximum neighborhood range and minimum neighborhood range; the environmental data include: all inspection targets in the chemical park and the hazardous characteristics of the corresponding inspection targets, as well as each risk area in the target chemical park and the risk level of the corresponding risk area; the initial inspection path is generated based on each inspection target and the corresponding hazardous characteristics, each risk area and the corresponding risk level, the number of drones and the flight speed, endurance, communication range and safe flight range of the corresponding drone; when the drone reaches any inspection target, the optimization number of the corresponding inspection path of the drone is increased by one.

[0007] In an optional embodiment, the method for obtaining the weight includes: determining the weight of the UAV when inspecting the target based on the total number of configured optimizations, the maximum weight, the minimum weight, and the number of optimizations of the inspection path corresponding to the UAV.

[0008] In an optional implementation, each patrol position and the corresponding real-time monitoring value, the target position and the weight of each patrol target, the environmental data, and the initial patrol path are input into a pre-built multi-objective optimization model, including: For any patrol target on the initial patrol path, the patrol position whose distance from the target position of the patrol target is less than the configured distance threshold is used as the target patrol position of the patrol target; based on the distance between the target patrol position and the corresponding patrol target, the accuracy of the real-time monitoring value corresponding to the target patrol position is determined; based on the accuracy of the real-time monitoring value corresponding to each target patrol position of the drone, the data collection quality of the drone on the initial patrol path is determined; according to the ratio of the area covered by the drone on the initial patrol path to the total area covered by the initial patrol path, the patrol coverage rate of the drone on the initial patrol path is obtained; according to the risk areas covered by the initial patrol path and the risk levels of the corresponding risk areas, the risk level of the initial patrol path is calculated; according to the flight time of the drone on the initial patrol path and the risk level of the initial patrol path, the risk degree of the drone on the initial patrol path is calculated; the data collection quality, the patrol coverage rate, the risk degree and the path length of the initial patrol path are input into a pre-constructed multi-objective optimization model to obtain the score of the initial patrol path.

[0009] In an optional embodiment, the multi-objective optimization model is constructed with minimization of path length, minimization of risk level, maximization of patrol coverage and maximization of data collection quality as objective functions, and with the flight time, communication range, safe flight range and configured patrol task allocation rules of the UAV as constraints; The inspection task allocation rule includes: each drone has a different inspection path.

[0010] In an optional implementation, according to the weight, the target position, the environmental data, the attribute, and the initial patrol path and position of each drone, a dynamic neighborhood search is used to determine the optimal patrol path of the drone, including: The neighborhood range of the UAV at the target position is determined according to the total number of configured optimizations, the maximum value of the neighborhood range, the minimum value of the neighborhood range, and the number of optimizations of the patrol path corresponding to the UAV; a neighborhood search is performed within the neighborhood range of the target position according to the environmental data to obtain multiple next candidate positions of the UAV; and the optimal patrol path is determined based on the weight, the target position, multiple next candidate positions, and the initial patrol path and position of each UAV.

[0011] In an optional embodiment, the weight includes individual weight and group weight; The method of determining the optimal patrol path based on the weight, the target position, multiple next candidate positions, and the initial patrol path and position of each drone includes: obtaining multiple alternative paths according to the multiple next candidate positions; calculating the score of each alternative path using the multi-objective optimization model; taking the alternative position corresponding to the alternative path with the highest score as the optimal alternative position; determining the first position under the influence of the individual drone according to the individual weight of the drone, the target position, and the optimal alternative position; determining the second position under the influence of the drone group according to the target position, the group weight, and the initial patrol path and position of each drone; performing weighted fusion on the first position and the second position to obtain the optimal position of the drone; and determining the optimal patrol path based on the target position and the optimal position.

[0012] In a second aspect, the present invention provides a device for optimizing the patrol path of a chemical park drone group, the device comprising: An acquisition unit, used to acquire environmental data of a target chemical park, and attributes and initial inspection paths of each drone used to inspect the target chemical park; wherein the initial inspection path corresponding to any drone includes at least one inspection target that the corresponding drone is responsible for; A control unit, for controlling any unmanned aerial vehicle to perform respective inspection tasks according to the corresponding initial inspection path, and obtaining real-time monitoring values ​​of each inspection position and the corresponding inspection position on the initial inspection path; A calculation unit is used to obtain the weight of the drone when it reaches any patrol target that the drone is responsible for; input each patrol position and the corresponding real-time monitoring value, the target position and the weight of each patrol target, the environmental data and the initial patrol path into a pre-built multi-objective optimization model to calculate the score of the initial patrol path; A determination unit, configured to determine an optimal patrol path of the UAV by using a dynamic neighborhood search according to the weight, the target location, the environmental data, the attributes, and the initial patrol path and location of each UAV; The optimization unit is used to use the optimal patrol path as a new initial patrol path if the score of the optimal patrol path is greater than the score of the initial patrol path, and return to the execution step: for any UAV, control the UAV to patrol according to the new initial patrol path until the UAV completes the patrol task.

[0013] In a third aspect, the present invention provides an electronic device, the electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; The processor is used to implement any of the methods described in the above-mentioned implementation modes when executing the program stored in the memory.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of the aforementioned embodiments is implemented.

[0015] This application effectively solves the problem of coordinated inspection of drone groups in complex environments by constructing a multi-objective optimization model that includes path length, risk level, inspection coverage, and data collection quality, and designs an improved multi-objective jellyfish search algorithm. Experimental results show that the method proposed in this application is superior to traditional algorithms in terms of inspection efficiency, risk control, and data collection quality, providing an innovative intelligent solution for safety inspections in chemical parks.

[0016] The multi-objective optimization model of this application includes four objectives. The first objective is to minimize the total path length to reduce patrol time and energy consumption; the second objective is to minimize the exposure time of high-risk areas to reduce the risks of drones and operators; the third objective is to maximize patrol coverage to ensure that key areas are fully monitored; the fourth objective is to maximize data collection quality, including flame recognition accuracy and gas detection accuracy; at the same time, the multi-objective optimization model also includes a variety of constraints, such as drone endurance, communication range, obstacle avoidance requirements, etc.; by introducing a penalty function mechanism, the degree of constraint violation is incorporated into the objective function, and the collaborative constraints between drones, such as collision avoidance and task allocation, are integrated, so that the multi-objective optimization model can fully reflect the needs of actual patrol tasks, and comprehensively and comprehensively assist drone patrol path optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a method for optimizing a chemical park drone group inspection path provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a device for optimizing the inspection path of a chemical park drone group provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] The optimization method of the patrol path of the chemical park drone group provided in the embodiment of the present application can be applied to the optimization system of the patrol path of the chemical park drone group, which includes: multiple drones equipped with flame recognition equipment and dangerous gas detection equipment, ground control station and communication equipment; the flame recognition equipment can detect and locate flame hazards in the park in real time, with high accuracy and rapid response characteristics; the dangerous gas detection equipment can detect a variety of toxic and harmful gases, such as hydrogen sulfide, ammonia, etc., and transmit concentration data in real time. The drone group maintains communication with the ground control station through a wireless network to achieve real-time data transmission and command reception.

[0021] In the embodiment of the present application, the drone can be a DJI Matrice 300 / custom quadcopter (equipped with a Pixhawk flight control) or other drones that support SDK (such as DJI MSDK / PX4 MAVSDK) and sensor expansion interface; the flame recognition device includes a camera component; the camera component can optionally use Sony IMX477 (supports 4K); the hazardous gas detection equipment uses FigaroTGS2600 (VOC detection) or other multi-sensors with unified power supply and I2C / SPI interface; the drone is equipped with NVIDIA Jetson Orin Nano (embedded GPU) for performing pre-processing such as image compression and data fusion to reduce transmission bandwidth requirements.

[0022] In the embodiment of the present application, data transmission can use Holybro SiK Radio (915MHz, 10km) or cellular network: Quectel 5G module; dual-link redundant design (data transmission + 4G) is adopted, and automatic switching is adopted; communication protocols include: MAVLink / ROS; path planning uses PX4 Avoidance / Open Motion Planning; MAVLink defines ODOMETRY and POSITION_TARGET messages, and ROS2 is used for task scheduling.

[0023] In the embodiment of the present application, the ground control station can be a cloud server (AWS EC2 / GCP instance, equipped with NVIDIA T4 GPU); running ROS / GAZEBO simulation environment, processing multi-machine data streams in real time; The ground control station can be a server or a terminal with strong computing power. The server can be a physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), as well as basic cloud computing services such as big data and artificial intelligence platforms. The terminal can be a user equipment (UE) such as a mobile phone, smart phone, laptop computer, digital broadcast receiver, personal digital assistant (PDA), tablet computer (PAD), handheld device, vehicle-mounted device, wearable device, computing device or other processing device connected to a wireless modem, mobile station (MS), mobile terminal, etc. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, which is not limited in this application.

[0024] The preferred embodiments of the present application are described below in conjunction with the drawings in the specification. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application. In addition, the embodiments and features in the embodiments of the present application may be combined with each other if there is no conflict.

[0025] Figure 1 A schematic diagram of a process flow of a method for optimizing a chemical park drone group inspection path provided in an embodiment of the present application. Figure 1 As shown, the method may include: Step S110, obtain the environmental data of the target chemical park, as well as the attributes and initial patrol paths of each UAV used to patrol the target chemical park; for any UAV, control the UAV to perform its respective patrol tasks according to the corresponding initial patrol path, and obtain the real-time monitoring values ​​of each patrol position and the corresponding patrol position on the initial patrol path.

[0026] In an embodiment of the present application, the environmental data of the target chemical park includes: all inspection targets in the chemical park and the hazardous characteristics of the corresponding inspection targets, as well as each risk area in the target chemical park and the risk level of the corresponding risk area.

[0027] In the embodiment of the present application, the initial inspection path of each drone is different, so as to complete the inspection task of the entire park by division of labor.

[0028] In the embodiment of the present application, each risk area of ​​the target chemical park and the risk level of the corresponding risk area are obtained by the following method: First, obtain the basic information of each building in the target chemical park, the basic information of the pipelines and tank areas set up in the target chemical park, the hazardous chemical storage production logistics map and other data; among them, the basic information of each building includes: location, purpose and structure, etc.; the basic information of the pipelines and tank areas includes: direction, location or area and types of hazardous chemicals stored, etc.; the hazardous chemical storage production logistics map includes: storage, transportation routes and production areas of hazardous chemicals; other data include: park population distribution, emergency facilities and surrounding environment, etc. Secondly, based on the above data obtained, the target chemical park is identified for risk, including: hazardous chemical identification, risk source identification and potential accident type identification; hazardous chemical identification includes: identifying all hazardous chemicals in the target chemical park and the hazardous characteristics of the corresponding hazardous chemicals (such as toxicity, flammability, etc.); risk source identification includes: identifying possible risk sources, such as storage tanks, pipelines, production equipment, etc.; potential accident type identification includes: identifying possible accident types, such as leakage, fire, explosion, etc.; thirdly, based on the risk identification results, accident consequence analysis, risk level classification and vulnerability analysis are carried out to obtain risk assessment results; among them, accident consequence analysis: assessing different The scope and severity of the accident; Risk level classification: Risk levels (such as high, medium, and low) are classified according to the possibility and consequences of the accident; Vulnerability analysis: Assess the vulnerability of personnel, equipment, and environment in the park to accidents; Finally, generate a risk distribution map that includes the risk areas of the target chemical park and the risk levels of the corresponding risk areas; The risk distribution map uses the architectural drawings or pipeline tank area layout drawings of the target chemical park as the base map; Mark the risk sources on the base map and use different colors or symbols to indicate the risk level; Mark the possible impact range of the accident, such as the leakage diffusion range or the explosion shock wave range; Mark emergency facilities such as fire stations and emergency material storage points.

[0029] In the embodiment of the present application, the risk distribution map has a legend, which is used to characterize the meaning of colors and symbols in the risk distribution map; the risk level and response measures of each risk source are detailed in the risk distribution map or in the appendix; the risk distribution map is updated regularly to reflect changes in the park; specifically, the risk distribution map can be obtained through the following tools or software: GIS software: such as ArcGIS, QGIS, for spatial data management and risk map drawing; CAD software: such as AutoCAD, for processing architectural drawings and pipeline layout drawings; risk assessment software: such as PHAST, ALOHA, for accident consequence simulation.

[0030] In the embodiment of the present application, the attributes of the drone include: flight speed, endurance time, communication range, safe flight range, maximum weight, minimum weight, maximum neighborhood range and minimum neighborhood range.

[0031] In an embodiment of the present application, the inspection targets include: industrial pipelines, factory buildings, tanks, warehouses and / or other facilities for storing or transporting hazardous chemicals in the target chemical park; the inspection targets may also be other buildings such as computer rooms that are prone to safety accidents; a risk level is set for each inspection target in advance based on the risk identification results and risk assessment results; the number of drones used to inspect the target chemical park is greater than or equal to 2; any drone is responsible for at least one inspection target; the inspection targets responsible for different drones may be the same or different.

[0032] In an embodiment of the present application, the initial inspection path corresponding to any UAV includes at least one inspection target that the corresponding UAV is responsible for; the inspection mission of the UAV consists of inspection speed, inspection target, inspection path and total inspection time; before all UAVs perform their respective inspection tasks, inspection tasks are generated for all UAVs in advance based on the environmental data of the target chemical park and the attributes of each UAV, and initial inspection paths are assigned; the initial inspection path corresponding to any UAV is generated based on each inspection target and the corresponding hazardous characteristics, each risk area and the corresponding risk level, the number of UAVs and the flight speed, endurance time, communication range and safe flight range of the corresponding UAV.

[0033] For example, before the inspection of the target chemical park begins, multiple drones will form a "swarm". At the same time, the initial parameters of this inspection mission are determined, such as setting the maximum flight speed of the drone to 15m / s, the maximum endurance time to 30 minutes, and the communication range to 500 meters. In addition, according to the actual situation of the park, different risk areas are divided, the locations of dangerous facilities such as toxic gas pipelines and storage tanks are marked, and a three-dimensional model of the park is constructed and imported into the simulation platform to prepare for subsequent path planning.

[0034] In the embodiment of the present application, the inspection path of the drone is composed of multiple inspection positions, and the inspection path of the drone covers at least one inspection target that the drone is responsible for; the target position of the inspection target is also a patrol position of the drone. For example, there are 2 inspection targets and 3 drones in the chemical park, and the area of ​​the chemical park is 150 acres; drone A and drone B are responsible for inspection target p1, and drone C is responsible for inspection target p2; the inspection paths of the three drones are different, and each inspection path has an inspection position every 10 meters. There are 50 inspection positions on a patrol path, and p1 overlaps with inspection position d5; drone A and drone B pass through inspection target p1 at the same time, but the other 49 inspection positions are different.

[0035] In an embodiment of the present application, each time the drone passes through an inspection location or each location on the inspection path, it is necessary to use the monitoring equipment it carries to perform real-time monitoring; specifically, the monitoring equipment includes a flame recognition device and a hazardous gas detection device; the flame recognition device is composed of an image acquisition component and an image analysis component, which is used to collect images at the inspection location and output flame recognition results through image analysis; the hazardous gas detection equipment includes: a hazardous gas acquisition component and a hazardous gas analysis component, which are used to collect air at the inspection location, and output the types of hazardous gases (such as hydrogen sulfide, ammonia, etc.) and the concentrations of the corresponding hazardous gases contained in the air through gas analysis, and compare them with the configured concentration thresholds of each hazardous gas. When the configured concentration threshold is exceeded, it indicates that the hazardous gas at the inspection location exceeds the standard.

[0036] In an embodiment of the present application, real-time monitoring values ​​include: flame recognition results and hazardous gas recognition results; the image analysis component is equipped with a YOLO flame recognition model, which can quickly identify whether the image collected at the inspection location contains a flame, and if so, outputs the flame detection frame and the confidence of the corresponding detection frame.

[0037] Step S120: For any patrol target that the drone is responsible for, obtain the weight of the drone when it reaches the patrol target; input each patrol position and the corresponding real-time monitoring value, the target position and weight of the patrol target, environmental data, and the initial patrol path into a pre-built multi-objective optimization model to calculate the score of the initial patrol path.

[0038] In the embodiment of the present application, when any UAV reaches a patrol target, the path is automatically optimized, and the number of optimization times of the patrol path corresponding to the corresponding UAV is increased by one.

[0039] In another embodiment of the present application, it can also be arranged that when any UAV reaches a patrol position, a path optimization is automatically performed, and the number of optimization times of the patrol path corresponding to the corresponding UAV is increased by one.

[0040] In the embodiment of the present application, the weight includes individual weight and group weight; specifically, the individual weight is obtained based on the total number of optimizations configured, the maximum weight, the minimum weight, and the number of optimizations of the patrol path corresponding to the drone; the specific formula is as follows: ;in, is the individual weight of the drone when it reaches any patrol target, and is the minimum and maximum weights of the configuration, obtained from the drone attributes; preferably, =0.1, =0.9; t is the current optimization times, and T is the total optimization times.

[0041] In the embodiment of the present application, group weight=1-individual weight.

[0042] In the embodiment of the present application, the weight of each drone is not fixed, but is carried out with the "iteration" (which can be understood as an optimization of the patrol path). The individual weight (flight path adjustment of each drone) and the group weight (cooperative flight strategy between multiple drones) are dynamically adjusted according to the current number of iterations (i.e., number of optimizations) and the total number of iterations (i.e., total number of optimizations).

[0043] In an embodiment of the present application, the weight of the drone is adaptively adjusted according to the number of optimizations, so that when the patrol path optimization is just beginning, the drone can be given a larger autonomous exploration space, allowing the drone to find possible patrol paths within a larger range; as the optimization nears the end, the weight of group collaboration is gradually increased, so that the drones can better coordinate with each other and jointly optimize the patrol path.

[0044] In the embodiment of the present application, each patrol position and the corresponding real-time monitoring value, the target position and weight of each patrol target, environmental data, and the initial patrol path are input into a pre-built multi-objective optimization model, including: For any patrol target on the initial patrol path, the patrol position whose distance from the target position of the patrol target is less than the configured distance threshold is used as the target patrol position of the patrol target; based on the distance between the target patrol position and the corresponding patrol target, determine the accuracy of the real-time monitoring value corresponding to the target patrol position; based on the accuracy of the real-time monitoring value corresponding to each target patrol position of the drone, determine the data collection quality of the drone on the initial patrol path; according to the ratio of the area covered by the drone on the initial patrol path to the total area covered by the initial patrol path, obtain the patrol coverage rate of the drone on the initial patrol path; according to the risk area covered by the initial patrol path and the risk level of the corresponding risk area, calculate the risk level of the drone on the initial patrol path according to the flight time of the drone on the initial patrol path and the risk level of the initial patrol path; according to the path length of the initial patrol path and the configured maximum path length, calculate the path length score; according to the risk level of the initial patrol path and the configured maximum risk level, calculate the risk level score; the data collection quality, patrol coverage, risk level and path length score are weighted and summed to obtain the score of the initial patrol path; The score of the initial patrol path is the initial fitness value of the UAV, which takes into account multiple objective functions, such as path length, exposure time in high-risk areas, patrol coverage, and data collection quality. For example, if the initial path planned by a UAV is shorter, the time spent passing through high-risk areas is less, more key areas can be covered, and the data collection quality is high, then its initial fitness value is relatively high. Specifically, the maximum risk level can be obtained by the risk levels of all risk areas in the chemical park and the longest stay time of the configured UAVs in each risk area.

[0045] In an embodiment of the present application, according to the distance between the target inspection position and the corresponding inspection target, the target accuracy of the corresponding distance is matched from the accuracy comparison table of different configured distances and different real-time monitoring values; it can also be determined according to the accuracy of the gas detection equipment or the flame identification equipment; for example, the accuracy can be set to 70% when the distance between the target inspection position and the corresponding inspection target is less than 5 meters, and to 90% when it is less than 1 meter.

[0046] In another embodiment of the present application, the attributes of the drone also include: the posture of the drone, and determining the data collection quality of the drone at the target inspection position corresponding to the inspection target based on the posture of the drone and the danger level and danger characteristics corresponding to the inspection target.

[0047] In an embodiment of the present application, the accuracy of the real-time monitoring value at the corresponding patrol position can be determined based on the distance between the patrol position and the corresponding patrol target, and the accuracy of the real-time monitoring value at each patrol position can be estimated using a pre-trained monitoring accuracy estimation model.

[0048] In another embodiment of the present application, when flames and / or hazardous gases are detected at any inspection location, the corresponding inspection location and detection results (including detection boxes, confidence levels, concentration values, etc.) can be input into a data quality assessment model that has been pre-trained using historical data to obtain the data collection quality at the inspection location, that is, the data collection quality of the real-time monitoring value.

[0049] In the embodiment of the present application, the data quality assessment model is obtained by training with historical data, and the method for collecting historical data includes: Obtain the historical monitoring values ​​collected by the drone at the corresponding historical inspection locations during the historical time period of the target chemical park, the accuracy of the corresponding historical flame recognition results, and the accuracy of the historical gas detection results; specifically, the accuracy of the historical flame recognition results is calculated through the detection results of the YOLO flame recognition model. The specific method is as follows: Flame recognition accuracy Calculated by the following formula: ; Among them: TP (True Positive) represents the number of correctly identified flames; FP (False Positive) represents the number of false positive flames; FN (False Negative) represents the number of missed flames; the above data is obtained by manually annotating and counting the images on the drone’s patrol path after the drone completes a patrol mission; for example, if the YOLO flame recognition model detects 10 flames, 8 of which are correct (TP=8), 2 are false positives (FP=2), and 1 flame is not detected (FN=1), then: ; Methods for determining the accuracy of historical gas detection results include: Gas detection accuracy Calculated by the following formula: ; in: Indicates the detected gas concentration value; Indicates the standard gas concentration value; det indicates the number of detections; for example, if 5 groups of gas concentration values ​​are detected, and the deviations from the standard values ​​are: 0.1, 0.2, 0.05, 0.15, 0.1, and the total standard value is 10, then After calculation, it is equal to 0.94; The corresponding historical data collection quality is obtained by weighted summing the accuracy of historical flame recognition results and the precision of historical gas detection results. ;in: The weight representing the flame recognition accuracy (usually set to 0.5); Indicates the weight of gas detection accuracy (usually set to 0.5); for example, if the flame recognition accuracy is 72.7% and the gas detection accuracy is 94%, then Q=0.8335; The historical monitoring value and the corresponding historical data collection quality as well as the target position or historical inspection position of the historical inspection target for collecting the historical monitoring value are taken as a historical data of the UAV, and the data quality assessment model is trained using multiple historical data of the UAV to obtain a trained data quality assessment model.

[0050] In the embodiment of the present application, the data collection quality is used as the objective function, which can guide the algorithm to optimize in the direction of obtaining high-quality data during the path planning stage. If this goal is not taken into consideration, and the path is planned only from aspects such as path length, risk level, and patrol coverage, it may cause the drone to fly frequently in some areas with poor data collection conditions, and the reliability of the acquired data is low, which cannot effectively meet the needs of safety inspections. By setting the data collection quality as the objective function, the algorithm will comprehensively weigh various factors, and while ensuring coverage of key areas and reducing risks, try to choose a path that is conducive to improving data collection quality, so that the final planned patrol path is more scientific and practical.

[0051] In practical applications, after a risk distribution assessment is made for the target chemical park, the characteristics, potential risks, and environmental conditions of different areas are known or predictable. For example, areas close to production equipment, storage tank areas, and other areas are more likely to have dangerous gas leaks and fires. Accordingly, when drones collect data in these areas, the difficulty and importance of flame identification and gas detection are also different. Based on past chemical park accident statistics, facility layout characteristics, and other information, an estimation model can be established to theoretically estimate the quality of collected data in different areas. For example, in high-risk areas, due to the large number of potential interference factors, the estimated collection is difficult and the data quality may be low; while in relatively open areas with less interference, the data quality may be higher. This estimation method can provide basic data support for incorporating data collection quality into the objective function.

[0052] At the same time, drone real-time monitoring equipment has specific performance parameters, such as the accuracy of flame recognition, the precision range of gas detection, and the detection limit. Based on these known sensor performance indicators, the quality of data collection under different conditions can be inferred. Combined with risk distribution assessment and environmental assessment, in an environment with low signal interference and obvious characteristics of the detection object, the sensor is more likely to obtain high-quality data; conversely, in cases where the signal is complex and the concentration of the detection object is close to the detection limit, the data quality may be affected. The data collection quality inference based on sensor performance does not rely on the actual collected data, but starts from the capabilities of the sensor itself, and evaluates the data quality that may be obtained by different patrol paths in advance, providing a reference for optimizing patrol paths.

[0053] In an embodiment of the present application, the risk level of the initial patrol path is calculated based on the risk areas covered by the initial patrol path and the risk levels of the corresponding risk areas, including: determining the risk value based on the proportion of the risk area to the area covered by the initial patrol path; determining the risk weight based on the risk level corresponding to the risk area; and determining the risk level of the initial patrol path based on the risk value and the risk weight.

[0054] In the embodiment of the present application, the multi-objective optimization model is constructed with minimization of path length, minimization of risk level, maximization of patrol coverage and maximization of data collection quality as objective functions, and with the drone's endurance time, communication range, safe flight range and configured patrol task allocation rules as constraints; the patrol task allocation rules include: the patrol path corresponding to each drone is different.

[0055] Specifically, the objective functions of the multi-objective optimization model formula include: ; ; ;

[0056] in, represents the distance that drone i moves from one point to the next point on path j (the line between these two points corresponds to a path segment), is a binary variable, indicating whether drone i passes through path j; represents the risk level of path j, is the dwell time of UAV i on path j; represents the area covered by UAV i on path j, is the total coverage area of ​​path j; It represents the quality of data collected by UAV i on path j.

[0057] The constraints of the multi-objective optimization model formula include: 1. Battery life constraints: ;in, It is the maximum flight time of the drone i.

[0058] 2. Communication range constraints: ;in, Indicates drone i and drone λ The distance between them, R is the communication range; 3. Obstacle avoidance constraints: ;in, represents the minimum distance between drone i and obstacles, and S is the safe distance; 4. Task allocation constraints: ; The task allocation constraint indicates that each path j can only be taken by one UAV i.

[0059] In the embodiment of the present application, the multi-objective optimization model further includes: a scoring calculation formula for the patrol path; the scoring calculation formula for the patrol path is as follows: ; Where, L represents the length of the inspection path; L max It represents the maximum length of all possible inspection paths (i.e., each path in the path set) (or the preset theoretical maximum value of the path length); Ris represents the risk level of the inspection path, which can be obtained by the weighted sum of the exposure time or risk level in the high-risk area; Ris max represents the maximum risk value among all possible inspection paths (or the preset theoretical maximum value of the path risk level); Cov represents the inspection coverage rate of the inspection path, which is the ratio of the coverage area to the total area; Cov max represents the theoretical maximum coverage rate (usually 1, i.e., full coverage); Q represents the data collection quality of the inspection path, which can be obtained by weighted summation of flame recognition accuracy and gas detection accuracy; Q max Indicates the theoretical maximum data collection quality (usually 1, i.e. 100% accurate); , , and represents the weights of path length, risk level, inspection coverage and data collection quality, ; , , and The value can be adjusted dynamically according to actual needs; if safety is the priority, increase (risk level) and (data collection quality); if efficiency is a priority, increase (path length) and (Patrol coverage) value; if there is no special requirement, the default weight is used: =0.25, =0.3, = 0.2, = 0.25; This default weight is based on the comprehensive results of experimental data, multi-objective optimization theory and the safety priority characteristics of chemical parks, ensuring that the algorithm achieves the optimal balance between path length, risk, coverage and data quality. Further calibration can be performed through simulation in actual applications.

[0060] In another embodiment of the present application, each patrol position and the corresponding real-time monitoring value, the target position and weight of each patrol target, environmental data, and the initial patrol path are input into a pre-built multi-objective optimization model, including: Determine the path set of the UAV based on the target position and weight of each patrol target and environmental data; for any patrol target on the initial patrol path, use the patrol position whose distance from the target position of the patrol target is less than the configured distance threshold as the target patrol position of the patrol target; determine the accuracy of the real-time monitoring value corresponding to the target patrol position based on the distance between the target patrol position and the corresponding patrol target; determine the data collection quality of the UAV on the initial patrol path based on the accuracy of the real-time monitoring value corresponding to each target patrol position of the UAV; obtain the patrol coverage rate of the UAV on the initial patrol path based on the ratio of the area covered by the UAV on the initial patrol path to the total area covered by the initial patrol path; and obtain the patrol coverage rate of the UAV on the initial patrol path based on the ratio of the area covered by the UAV on the initial patrol path to the total area covered by the initial patrol path. The risk level of the initial inspection path is calculated by checking the risk areas covered by the path and the risk levels of the corresponding risk areas; the path length score is calculated according to the path length of the initial inspection path and the configured maximum path length; the risk level of the UAV on the initial inspection path is calculated according to the flight time of the UAV on the initial inspection path and the risk level of the initial inspection path; the risk level of all paths in the path set is calculated; the risk level of the path with the highest risk level is taken as the maximum risk level; the ratio of the risk level of the initial inspection path to the maximum risk level is taken as the risk level score; the data collection quality, inspection coverage, risk level score and path length score are weighted to obtain the score of the initial inspection path.

[0061] In an embodiment of the present application, a path set includes multiple paths; the path set may be composed of all possible inspection paths within the chemical park, or may be composed of all possible inspection paths of the UAV. If it is composed of all possible inspection paths of the UAV, then each path in the path set has the same starting point and end point as the initial inspection path, and any path covers at least one inspection target that the UAV is responsible for.

[0062] Step S130: According to the weight, target position, environmental data, attributes, and the initial patrol path and position of each drone, a dynamic neighborhood search is used to determine the optimal patrol path of the drone.

[0063] In an embodiment of the present application, the optimal patrol path of the drone is determined by using an improved multi-objective jellyfish search algorithm based on adaptive weights and dynamic neighborhood search; in each patrol path optimization process, each drone adjusts its own patrol path according to the adaptive weights and dynamic neighborhood search strategy; at the same time, an external archiving mechanism is introduced to store the non-dominated solutions (i.e., solutions that cannot be surpassed by other solutions in all objectives, such as a path that is no worse than other paths in terms of path length, risk exposure time, patrol coverage, and data collection quality) generated in each iteration in an external archive; then, the solutions in the archive are screened and optimized through non-dominated sorting and congestion calculation to maintain the diversity of the solution set and ensure that multiple patrol path solutions with different focuses but all relatively good can be found.

[0064] Specifically, it includes: determining the neighborhood range of the UAV at the target position according to the total optimization times of the configuration, the maximum value of the neighborhood range, the minimum value of the neighborhood range and the optimization times of the patrol path corresponding to the UAV; performing neighborhood search within the neighborhood range of the target position according to environmental data to obtain multiple next candidate positions of the UAV; obtaining multiple alternative paths according to the multiple next candidate positions; calculating the score of each alternative path using a multi-objective optimization model; taking the alternative position corresponding to the alternative path with the highest score as the optimal alternative position; determining the first position under the influence of the individual UAV according to the individual weight of the UAV, the target position and the optimal alternative position; determining the second position under the influence of the UAV group according to the target position, the group weight and the initial patrol path and position of each UAV; performing weighted fusion of the first position and the second position to obtain the optimal position of the UAV; determining the optimal patrol path based on the target position and the optimal position.

[0065] In the embodiment of the present application, the neighborhood range determination formula for any drone at any target location is as follows: ; in, represents the neighborhood range of any UAV at any target location, and Indicates the maximum value and minimum value of the neighborhood range; can be set , .

[0066] In an embodiment of the present application, a dynamic neighborhood search strategy is adopted, and a larger "neighborhood range" is set for each drone in the initial stage. The neighborhood here can be understood as the range of paths of other surrounding drones that each drone refers to when looking for a better path. For example, when planning a path, a drone can refer to the path information of other drones that are farther away from it, which can expand the search range, enhance the global search capability, and find more potential high-quality patrol paths. As the iteration proceeds, the neighborhood range is gradually narrowed according to the formula of the neighborhood range, so that the drone can focus more on the local area near itself in the later stage, conduct a more detailed search near the explored better path, improve the local search accuracy, and further optimize its own patrol path.

[0067] In an embodiment of the present application, a neighborhood search may be performed within the neighborhood of the target location using methods such as random sampling, grid division, or proximity search to obtain multiple next candidate locations for the drone.

[0068] In an embodiment of the present application, after obtaining multiple alternative positions, a smooth difference method can be used to generate a smooth path between the current position and each alternative position to obtain an alternative path; or multiple alternative positions can be used to replace the next patrol position in the initial patrol path to obtain multiple alternative paths.

[0069] In the embodiment of the present application, the determination of the optimal path and the optimal alternative position is based on multiple iterations. It is not that an optimal alternative position or an optimal path is directly output, but it is necessary to determine whether it meets the termination condition. The termination condition may be that the score of the corresponding alternative path has not been significantly improved for many consecutive times. When the termination condition is met, the optimal inspection path will be output.

[0070] In the embodiment of the present application, the optimal inspection path is the path with the best comprehensive performance, which can take into account multiple objectives such as path length, risk level, inspection coverage and data collection quality, and can meet the actual needs of chemical park safety inspections to the greatest extent, and realize efficient, comprehensive and safe inspection tasks.

[0071] In the embodiment of the present application, non-dominated sorting and congestion calculation are used to maintain the diversity and convergence of the solution set: ; in, Represents an individual The congestion level, Is an individual ( represents a solution that covers the overall task allocation and path planning of the drone swarm) on the value of the kth objective function, and are the maximum and minimum values ​​of the kth objective function.

[0072] Step S140: If the score of the optimal patrol path is greater than the score of the initial patrol path, the optimal patrol path is used as the new initial patrol path, and the execution step is returned to: for any UAV, the UAV is controlled to patrol according to the new initial patrol path until the UAV completes the patrol task.

[0073] In an embodiment of the present application, after the optimal patrol path optimized for this patrol path is obtained, it is compared with the initial patrol path currently being executed by the drone. If its score is better than the initial patrol path currently being executed by the drone, it means that the optimal patrol path is better, and the optimal patrol path is used to replace the patrol path currently being executed, and the patrol task is continued according to the optimal patrol path until the next patrol target is reached, and the patrol path is optimized again; if the score of the optimal patrol path is less than the patrol path currently being executed, it means that the patrol path currently being executed is better, and the drone is controlled to continue to execute the patrol path currently being executed.

[0074] In an embodiment of the present application, after the drone completes the inspection mission, the real-time monitoring values ​​collected by the drone during the inspection process will be uniformly summarized and analyzed to facilitate the management of the chemical park.

[0075] In another embodiment of the present application, when the drone detects that the concentration of flames or dangerous gases exceeds a preset concentration threshold during the inspection process, the alarm program will be immediately activated, and the current time and location will be sent to the remote end; at the same time, the drone will first expand the detection range, collect images containing more objects and air for secondary precise detection; at the same time, the drone will descend in altitude, perform image collection and air detection again, and combine the results of the three detections to generate danger detection results for the inspection location; the danger detection results include: danger location, danger type, danger level and danger range; among them, the danger type is flame or dangerous gas; the danger level is determined according to the flame area or the degree of combustion or the type and concentration of the dangerous gas, and the danger range is determined according to the scope of the danger.

[0076] When any UAV detects danger at any patrol location, the UAV will downgrade the priority of the patrol mission it is performing; the priority of the task of detecting danger will be made the highest priority, that is, the UAV will stay at the dangerous location and will not move forward; the UAV will detect whether there is a patrol target at the dangerous location based on the image equipment, and if so, it will identify the type of patrol target and obtain the corresponding danger characteristics and danger level of the patrol target; at the same time, the UAV will conduct another patrol based on the environment around the patrol target and the location or scope of the patrol target to determine whether the danger has spread, and will promptly feedback the detection results to the remote end.

[0077] When the remote end receives the danger detection results fed back by any drone, it will call other drones that have completed the patrol mission or are in standby mode to explore the corresponding location to avoid false detection at a single location; if both drones feed back the danger detection results, a danger control plan will be generated based on the danger location, danger type, danger level and danger range, and a drone that can execute the danger control plan will be dispatched to the danger location to execute the corresponding plan; for example, when a fire is discovered, a drone that can carry fire extinguishing agents will be controlled to go to the danger location to extinguish the fire.

[0078] In an embodiment of the present application, when any drone detects the presence of danger, the weights or priorities of all patrol locations within the danger range or other patrol targets associated with the danger range (such as connected pipes), or other patrol locations within a certain range from the danger range will be increased, so that other drones that are performing patrol tasks will give priority to patrolling the above-mentioned locations to avoid the spread of danger.

[0079] In an embodiment of the present application, when individual drones hover for confirmation due to identification of data fluctuations or suspected fire leaks, or when individual drones fail to complete the original route inspection as planned due to uncontrollable reasons such as insufficient power, environmental disturbances, or failures, temporary recalls, etc., the present application can be triggered to dynamically adjust the path of the drone swarm so that the cluster can flexibly respond to emergencies and cooperate to complete the park inspection mission.

[0080] In an embodiment of the present application, when any drone detects danger at any patrol position, it will send its own patrol target, patrol path and current position to the remote end; when the remote end receives the danger detection result and the patrol target, patrol path and current position of the corresponding drone fed back by any drone, it determines whether the drone can continue to perform the patrol mission based on the danger control plan; if not, the backup drone is called to perform the patrol mission of the drone; or, the properties, current positions and patrol paths of all drones are retrieved, and the properties, current positions and patrol paths of the patrol targets that have not been patrolled by the drone and the drones that can perform the patrol mission normally are input into the multi-objective optimization model, and the drone paths are re-optimized to enable other drones that can perform the patrol mission normally to complete the patrol mission of the drone that cannot perform the patrol mission normally.

[0081] In one embodiment of the present application, if the hazard control plan cannot be executed by the drone, the drone can continue to perform the patrol mission; if the drone can execute the hazard control plan, the hazard control time required to execute the hazard control plan and the first estimated patrol completion time of the patrol mission it performs, as well as the second estimated patrol completion time of other drones that are performing patrol missions are judged; if the difference between the sum of the hazard control time and the first estimated patrol completion time and the second estimated patrol completion time is greater than the configured time threshold, the drone cannot continue to perform the patrol mission; wherein the estimated patrol completion time is obtained based on the drone's speed, patrol path, and the collection time required at each patrol location.

[0082] The technical solution of this application is further explained below in conjunction with specific experiments.

[0083] In order to verify the effectiveness of the proposed method, this application selected a typical chemical park as the experimental scene. The park contains a variety of facilities such as toxic gas pipelines, storage tanks, offices, production plants and living areas, which are highly complex and representative. This application set up drone swarms of different sizes (5, 10, and 15) for comparative experiments to compare IMOJSA with traditional NSGA-II and MOPSO algorithms.

[0084] 1.1 Experimental Setup 1.1.1 Experimental Environment The experiment selected a typical chemical park as the test scene, which contains a variety of facilities such as toxic gas pipelines, storage tanks, offices, production plants and living areas. The experimental scene was constructed using 3D modeling software and imported into the simulation platform for testing. The simulation platform is built based on Gazebo and ROS (Robot Operating System), which can simulate the flight behavior of drones and sensor data collection.

[0085] 1.1.2 Drone parameter settings Multiple quadrotor drones were used in the experiment, and their parameters were set as follows: maximum flight speed: 15 m / s; maximum endurance time: 30 minutes; communication range: 500 meters; sensors; YOLO flame recognition model, chemical gas detection equipment (capable of detecting hydrogen sulfide, ammonia, etc.).

[0086] 1.1.3 Comparison of algorithms In order to verify the performance of IMOJSA, two classic multi-objective optimization algorithms were selected for comparison: NSGA-II: non-dominated sorting genetic algorithm and MOPSO: multi-objective particle swarm optimization algorithm; 1.1.4 Evaluation indicators The experiment uses the following evaluation indicators to evaluate the performance of the algorithm: Hypervolume (HV): measures the coverage and quality of the Pareto front solution set. Inspection coverage: the proportion of key areas covered by the drone group. Risk exposure time: the total time that the drone stays in the high-risk area. Data collection quality: a comprehensive score of flame recognition accuracy and gas detection accuracy.

[0087] 1.2 Experimental Procedure 1.2.1 Algorithm Initialization First, initialize the parameters of the IMOJSA, NSGA-II, and MOPSO algorithms: population size: 50; maximum number of iterations: 100; external archive size: 100; adaptive weight range: , ; Dynamic neighborhood range: =1, =10; 1.2.2 Path Planning and Simulation Path planning: Use IMOJSA, NSGA-II, and MOPSO to plan inspection paths for the drone swarm. After each planning, record the Pareto frontier solution set. Simulation operation: Run the planned path in the Gazebo simulation platform to simulate the flight behavior of the drone and sensor data collection. Data recording: Record the inspection coverage, risk exposure time, and data collection quality of each simulation.

[0088] 1.2.3 Experimental scene setting The experiment set up three different sizes of drone swarms: 5 drones: suitable for small chemical parks; 10 drones: suitable for medium-sized chemical parks; 15 drones: suitable for large chemical parks.

[0089] 1.3 Experimental Results 1.3.1 Comparison of Hypervolume Indicators The hypervolume index (HV) is used to measure the coverage and quality of the Pareto frontier solution set. The larger the HV value, the higher the quality of the solution set. The experimental results are shown in Table 1: Table 1

[0090] As can be seen from Table 1, the HV values ​​of IMOJSA in all scenarios are higher than those of NSGA-II and MOPSO, indicating that IMOJSA can find a higher quality Pareto frontier solution set.

[0091] 1.3.2 Comparison of patrol coverage The patrol coverage rate is an important indicator to measure the proportion of key areas covered by drone groups. The results are shown in Table 2: Table 2

[0092] As can be seen from Table 2, the patrol coverage of IMOJSA in all scenarios is higher than that of NSGA-II and MOPSO, indicating that IMOJSA can cover key areas more effectively.

[0093] 1.3.3 Comparison of risk exposure time Risk exposure time is an indicator to measure the total time that the drone stays in the high-risk area. The experimental results are shown in Table 3: Table 3

[0094] As can be seen from Table 3, the risk exposure time of IMOJSA in all scenarios is lower than that of NSGA-II and MOPSO, indicating that IMOJSA can more effectively reduce the residence time of drones in high-risk areas.

[0095] 1.3.4 Comparison of data collection quality Data collection quality is a comprehensive score that measures the flame recognition accuracy and gas detection accuracy. The experimental results are shown in Table 4: Table 4

[0096] As can be seen from Table 4, the data acquisition quality of IMOJSA in all scenarios is higher than that of NSGA-II and MOPSO, indicating that IMOJSA can improve the data acquisition quality more effectively.

[0097] 1.5 Results Analysis It can be seen from the above experimental data and charts that IMOJSA shows significant advantages in multi-objective optimization performance. The specific analysis is as follows: Hypervolume index: The HV value of IMOJSA in all scenarios is higher than that of NSGA-II and MOPSO, indicating that IMOJSA can find a higher quality Pareto frontier solution set. Inspection coverage: The inspection coverage of IMOJSA in all scenarios is higher than that of NSGA-II and MOPSO, indicating that IMOJSA can cover key areas more effectively. Risk exposure time: The risk exposure time of IMOJSA in all scenarios is lower than that of NSGA-II and MOPSO, indicating that IMOJSA can more effectively reduce the residence time of drones in high-risk areas. Data collection quality: The data collection quality of IMOJSA in all scenarios is higher than that of NSGA-II and MOPSO, indicating that IMOJSA can more effectively improve the quality of data collection. In summary, IMOJSA shows significant advantages in the optimization problem of safety inspection paths of drone groups in chemical parks, and can effectively improve inspection efficiency, reduce risk exposure time and improve data collection quality.

[0098] Experimental results show that IMOJSA has significant advantages in the optimization of safety inspection paths for drone swarms in chemical parks. By introducing an adaptive weight mechanism and a dynamic neighborhood search strategy, IMOJSA can effectively balance the exploration and development capabilities of the algorithm and find a higher quality Pareto frontier solution set. In addition, IMOJSA outperforms traditional NSGA-II and MOPSO algorithms in terms of inspection coverage, risk exposure time, and data collection quality.

[0099] Experimental results show that IMOJSA has significant advantages in multi-objective optimization performance. In the scenario of 5 drones, the Pareto frontier solution set found by IMOJSA is 12% and 8% higher than that of NSGA-II and MOPSO in terms of hypervolume index, respectively. As the number of drones increases, the advantage of IMOJSA becomes more obvious. In the scenario of 15 drones, IMOJSA improves the patrol coverage index by more than 15% compared with the comparison algorithm, while maintaining a low risk exposure time and high data collection quality.

[0100] By analyzing the algorithm convergence curve and solution set distribution, this application found that IMOJSA has a faster convergence speed and better solution set diversity. This shows that IMOJSA can effectively explore the solution space and find a good balance between multiple objectives. In addition, this application also conducted actual flight tests to verify the feasibility and effectiveness of the proposed method in practical applications.

[0101] Corresponding to the above method, the embodiment of the present application also provides a device for optimizing the inspection path of a chemical park drone group, such as Figure 2 As shown, the optimization device for the chemical park drone group inspection path includes: The acquisition unit 210 is used to acquire environmental data of the target chemical park, and the attributes and initial inspection paths of each drone used to inspect the target chemical park; wherein the initial inspection path corresponding to any drone includes at least one inspection target that the corresponding drone is responsible for; The control unit 220 is used to control any UAV to perform its respective inspection tasks according to the corresponding initial inspection path, and obtain the real-time monitoring values ​​of each inspection position and the corresponding inspection position on the initial inspection path; The calculation unit 230 is used to obtain the weight of the drone when it reaches the patrol target for any patrol target that the drone is responsible for; input each patrol position and the corresponding real-time monitoring value, the target position and weight of each patrol target, environmental data, and the initial patrol path into a pre-built multi-objective optimization model to calculate the score of the initial patrol path; A determination unit 240, for determining an optimal patrol path of the UAV by using a dynamic neighborhood search based on the weight, target location, environmental data, attributes, and initial patrol paths and locations of each UAV; The optimization unit 250 is used to use the optimal patrol path as the new initial patrol path if the score of the optimal patrol path is greater than the score of the initial patrol path, and return to the execution step: for any UAV, control the UAV to patrol according to the new initial patrol path until the UAV completes the patrol task.

[0102] The functions of each functional unit of the device for optimizing the patrol path of a swarm of drones in a chemical park provided in the above-mentioned embodiment of the present application can be achieved through the above-mentioned method steps. Therefore, the specific working process and beneficial effects of each unit in the device for optimizing the patrol path of a swarm of drones in a chemical park provided in the embodiment of the present application will not be repeated here.

[0103] The present application also provides an electronic device, such as Figure 3 As shown, it includes a processor 310 , a communication interface 320 , a memory 330 and a communication bus 340 , wherein the processor 310 , the communication interface 320 , and the memory 330 communicate with each other via the communication bus 340 .

[0104] The memory 330 is used to store computer programs; the processor 310 is used to implement the following steps when executing the program stored in the memory 330: obtaining environmental data of the target chemical park, as well as the attributes and initial patrol paths of each drone used to patrol the target chemical park; wherein the initial patrol path corresponding to any drone includes at least one patrol target that the corresponding drone is responsible for; for any drone, controlling the drone to perform its respective patrol tasks according to the corresponding initial patrol path, and obtaining the real-time monitoring values ​​of each patrol position and the corresponding patrol position on the initial patrol path; for any patrol target that the drone is responsible for, obtaining the time when the drone reaches the patrol target The weight at that time; each patrol position and the corresponding real-time monitoring value, the target position and weight of the patrol target, the environmental data and the initial patrol path are input into the pre-built multi-objective optimization model to calculate the score of the initial patrol path; according to the weight, target position, environmental data, attributes and the initial patrol path and position of each UAV, the optimal patrol path of the UAV is determined by dynamic neighborhood search; if the score of the optimal patrol path is greater than the score of the initial patrol path, the optimal patrol path is used as the new initial patrol path, and the execution step is returned: for any UAV, the UAV is controlled to patrol according to the new initial patrol path until the UAV completes the patrol task.

[0105] The communication bus mentioned above can be a peripheral component interconnect standard (PCI) bus or an extended industrial standard architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used for communication between the above-mentioned electronic device and other devices. The memory may include a random access memory (RAM) and may also include a non-volatile memory (NVM), such as at least one disk storage. Optionally, the memory may also be at least one storage device located away from the aforementioned processor. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. Since the implementation methods and beneficial effects of the various devices of the electronic device in the above-mentioned embodiments to solve the problems can be referred to. Figure 1 The various steps in the illustrated embodiment are implemented, therefore, the specific working process and beneficial effects of the electronic device provided by the embodiment of the present application are not repeated here.

[0106] In another embodiment provided in the present application, a computer-readable storage medium is also provided, in which instructions are stored. When the computer-readable storage medium is executed on a computer, the computer executes the method for optimizing the inspection path of a chemical park drone group as described in any of the above embodiments.

[0107] In another embodiment provided in the present application, a computer program product comprising instructions is also provided. When the computer is run on a computer, the computer executes the method for optimizing the inspection path of a chemical park drone swarm as described in any of the above embodiments.

[0108] Those skilled in the art will appreciate that the embodiments in the present application can be provided as methods, systems, or computer program products. Therefore, the embodiments in the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments in the present application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0109] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0113] Obviously, those skilled in the art can make various changes and modifications to the embodiments in the present application without departing from the spirit and scope of the embodiments in the present application. Thus, if these modifications and variations of the embodiments in the present application fall within the scope of the claims and their equivalents in the embodiments of the present application, the embodiments of the present application are also intended to include these modifications and variations.

Claims

1. A method for optimizing the inspection path of a chemical park drone group, characterized in that: The method comprises: Acquire environmental data of a target chemical park, as well as the attributes and initial patrol paths of each drone used to patrol the target chemical park; wherein the initial patrol path corresponding to any drone includes at least one patrol target that the corresponding drone is responsible for; For any UAV, control the UAV to perform respective inspection tasks according to the corresponding initial inspection path, and obtain the real-time monitoring values ​​of each inspection position and the corresponding inspection position on the initial inspection path; For any inspection target that the drone is responsible for, obtaining the weight of the drone when it reaches the inspection target; Inputting each patrol position and the corresponding real-time monitoring value, the target position and the weight of the patrol target, the environmental data, and the initial patrol path into a pre-built multi-objective optimization model, and calculating the score of the initial patrol path; Determine the optimal patrol path of the UAV by using dynamic neighborhood search according to the weight, the target location, the environmental data, the attributes, and the initial patrol path and location of each UAV; If the score of the optimal patrol path is greater than the score of the initial patrol path, the optimal patrol path is used as the new initial patrol path, and the execution step is returned to: for any UAV, the UAV is controlled to patrol according to the new initial patrol path until the UAV completes the patrol task.

2. The method according to claim 1, characterized in that The attributes include: flight speed, endurance, communication range, safe flight range, maximum weight, minimum weight, maximum neighborhood range, and minimum neighborhood range; The environmental data include: all inspection targets of the chemical park and the hazardous characteristics of the corresponding inspection targets, and each risk area of ​​the target chemical park and the risk level of the corresponding risk area; The initial inspection path is generated based on each inspection target and the corresponding hazard characteristics, each risk area and the corresponding risk level, the number of drones and the flight speed, endurance time, communication range and safe flight range of the corresponding drones; When the UAV reaches any inspection target, the optimization times of the inspection path corresponding to the UAV is increased by one.

3. The method according to claim 2, characterized in that The method for obtaining the weight includes: The weight of the UAV when inspecting the target is determined based on the total number of configured optimizations, the maximum weight, the minimum weight, and the number of optimizations of the inspection path corresponding to the UAV.

4. The method according to claim 2, characterized in that Inputting each patrol position and the corresponding real-time monitoring value, the target position and the weight of each patrol target, the environmental data and the initial patrol path into a pre-built multi-objective optimization model, including: For any patrol target on the initial patrol path, a patrol position whose distance from the target position of the patrol target is less than a configured distance threshold is used as the target patrol position of the patrol target; Determine the accuracy of the real-time monitoring value corresponding to the target inspection position based on the distance between the target inspection position and the corresponding inspection target; Determining the data collection quality of the UAV on the initial inspection path based on the accuracy of the real-time monitoring values ​​corresponding to each target inspection position of the UAV; Obtaining a patrol coverage rate of the UAV on the initial patrol path according to a ratio of an area covered by the UAV on the initial patrol path to a total area covered by the initial patrol path; Calculating the risk level of the initial patrol path according to the risk areas covered by the initial patrol path and the risk levels of the corresponding risk areas; Calculate the risk level of the UAV on the initial patrol path according to the flight time of the UAV on the initial patrol path and the risk level of the initial patrol path; The data collection quality, the inspection coverage rate, the risk level and the path length of the initial inspection path are input into a pre-built multi-objective optimization model to obtain a score for the initial inspection path.

5. The method according to claim 4, characterized in that The multi-objective optimization model is constructed with minimization of path length, minimization of risk level, maximization of patrol coverage and maximization of data collection quality as objective functions, and with the flight time, communication range, safe flight range and configured patrol task allocation rules of the UAV as constraints; The inspection task allocation rule includes: each drone has a different inspection path.

6. The method according to claim 2, characterized in that According to the weight, the target position, the environmental data, the attribute, and the initial patrol path and position of each UAV, a dynamic neighborhood search is used to determine the optimal patrol path of the UAV, including: Determine the neighborhood range of the drone at the target location according to the total number of configured optimizations, the maximum value of the neighborhood range, the minimum value of the neighborhood range, and the number of optimizations of the patrol path corresponding to the drone; According to the environmental data, a neighborhood search is performed within the neighborhood of the target location to obtain a plurality of next candidate locations of the drone; The optimal patrol path is determined based on the weight, the target position, multiple next candidate positions, and the initial patrol path and position of each drone.

7. The method according to claim 2, characterized in that The weights include individual weights and group weights; The determining the optimal patrol path based on the weight, the target position, a plurality of next candidate positions, and the initial patrol path and position of each UAV includes: Obtaining multiple candidate paths according to the multiple next candidate positions; calculating the score of each candidate path using the multi-objective optimization model; The alternative position corresponding to the alternative path with the highest score is taken as the optimal alternative position; Determine a first position under the influence of the individual drone according to the individual weight of the drone, the target position and the optimal candidate position; Determine a second position under the influence of the drone group according to the target position, the group weight, and the initial patrol path and position of each drone; Performing weighted fusion on the first position and the second position to obtain an optimal position of the UAV; The optimal inspection path is determined based on the target position and the optimal position.

8. A device for optimizing the inspection path of a chemical park drone group, characterized in that: The device comprises: An acquisition unit, used to acquire environmental data of a target chemical park, and attributes and initial inspection paths of each drone used to inspect the target chemical park; wherein the initial inspection path corresponding to any drone includes at least one inspection target that the corresponding drone is responsible for; A control unit, for controlling any unmanned aerial vehicle to perform respective inspection tasks according to the corresponding initial inspection path, and obtaining real-time monitoring values ​​of each inspection position and the corresponding inspection position on the initial inspection path; A calculation unit is used to obtain the weight of the drone when it reaches any patrol target that the drone is responsible for; input each patrol position and the corresponding real-time monitoring value, the target position and the weight of each patrol target, the environmental data and the initial patrol path into a pre-built multi-objective optimization model to calculate the score of the initial patrol path; A determination unit, configured to determine an optimal patrol path of the UAV by using a dynamic neighborhood search according to the weight, the target location, the environmental data, the attributes, and the initial patrol path and location of each UAV; The optimization unit is used to use the optimal patrol path as a new initial patrol path if the score of the optimal patrol path is greater than the score of the initial patrol path, and return to the execution step: for any UAV, control the UAV to patrol according to the new initial patrol path until the UAV completes the patrol task.

9. An electronic device, characterized in that: The electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-7 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images