A Coalition-Based Task Distribution System and Method Based on Structural Health and Energy for Task Sharing in Autonomous Unmanned Aerial Vehicle Swarms

Abstract

This invention relates to a task distribution system (1) and method (1000) that enables swarms of numerous autonomous unmanned aerial vehicles (A) to perform task sharing in an intelligent, safe and efficient manner, based on structural health and energy-based coalition. Unlike existing structures for task sharing in autonomous unmanned aerial vehicle (A) swarms, the invention offers a holistic AI-powered decision-making mechanism that considers physical health, energy status and task priorities together. In this respect, the invention can be positioned as a highly customized, scalable, learning and safe task distribution infrastructure.

Description

1 TARIFF Task Assignment System and Method Technical Area This invention enables swarms of numerous autonomous unmanned aerial vehicles to perform tasks. sharing, in a smart, secure and 5-way manner, based on a coalition focused on structural health and energy. with a task distribution system and method that enables it to be carried out efficiently It is related. State of the Art In current applications, task sharing in autonomous drone swarms typically fixed rules, static algorithms, or simple distance and route optimization 10 This is done through decision-making mechanisms based on these principles. The most commonly used methods are: among them, greedy algorithms, predefined central control systems, task These include clustering and heuristic-based algorithms. These algorithms perform tasks. generally only the current locations of unmanned aerial vehicles, the geographic nature of the missions It assigns them based on distribution and the shortest route costs. 15 In addition, some advanced systems use multi-agent task sharing for task allocation and Auction-based models are used. In these models, each unmanned aerial vehicle, A drone submits a bid for a specific task, and the drone that submits the most suitable bid wins the mission. However, even with such systems, the status or health of the unmanned aerial vehicle's internal hardware is unknown. Profile is generally not among the evaluation criteria. 20 In addition, some reinforcement-based learning approaches have been proposed in recent years. However, most of these studies only consider environmental variables such as wind, map, and obstacles. focusing on the drone's own internal physical parameters for decision-making It does not include it in the purchasing process. The main shortcomings of these traditional methods can be summarized as follows: 25 Ignoring Hardware Health Data: Current algorithms neglect drone health data. critical health data of the vehicle such as battery level, engine temperature, load carrying limit It does not take this into account during task assignment. This situation affects those with insufficient or excessive energy. 2 assigning a mission to an overheated unmanned aerial vehicle and malfunctions during the mission This can cause it to occur. Insufficient Energy Efficiency Optimization: Focusing solely on locational or route costs. The assignment of tasks based on this leads to an uneven distribution of energy usage across the fleet. This The situation is causing some unmanned aerial vehicles to become obsolete very quickly, and some unmanned aerial vehicles 5 This causes their vehicles to become idle. Security and Scalability Issues in Centralized Structures: Centralized control systems Task sharing based on the system causes the entire system to have a single point of failure. Yes, that's right. Also, as the fleet grows, the workload on this center increases, and decision-making time lengthens. Ignoring Task Priority and Payload Type: In current techniques, tasks are often overlooked. such as priority or urgent cargo, or special transportation requirements such as sensitive cargo or high weight. The information is not taken into consideration. This leads to critical missions being undertaken with poor suitability unmanned aerial vehicles. This could lead to them being assigned to vehicles. The fundamental technical problem resulting from these shortcomings is that the tasks, unmanned aerial vehicles taking into account the current structural health, energy status and suitability of the vehicles for the type of task. 15 Increased job failure rates across the system due to appointments made without prior approval, This results in decreased energy efficiency and reduced operational safety. Furthermore, the central The lack of scalability and robustness seen in structures, coupled with the use of multiple unmanned aerial vehicles... This limits the effective use of these systems in large areas. For these reasons, the task sharing process should not only involve route and location information, but also each 20 Real-time internal health data of the unmanned aerial vehicle and multi-criteria mission analysis. It needs to be evaluated along with its features. This technical requirement must be met. The invention, developed for this purpose, eliminates the shortcomings listed above and improves the system. It significantly improves task performance, safety, and efficiency. Patent application number TR2025 / 019262, which is included in the prior art, 25 The document describes how, using a digital twin of the earthquake zone, reinforcement learning can be implemented. an autonomous drone swarm system controlled by a trained artificial intelligence model It is explained. In the system described in the relevant application document, unmanned aerial vehicles The elements and methods used in task sharing within swarms of vehicles differ. It is structured. 30 3 In conclusion, solutions that address the needs described above are relevant to the subject. Due to its shortcomings, it has become necessary to make improvements in the relevant technical field. Brief Description of the Invention The invention was created by drawing inspiration from existing situations and overcoming the aforementioned drawbacks. It aims to solve. 5 The aim of this invention is to enable swarms of numerous autonomous unmanned aerial vehicles to perform tasks. sharing, in a smart, secure and coalition-based manner, focused on structural health and energy. a task distribution system and method that enables it to be carried out efficiently It is the presentation of the matter. The system monitors the battery level and motor status of each unmanned aerial vehicle in real time. (allowing it to analyze physical health data such as temperature and load-bearing capacity) It has local assessment modules that recognize location. This allows tasks to be performed solely based on location. or not according to the route, but also according to the current operational status of the unmanned aerial vehicle. They can be assigned depending on the situation. The system developed within the scope of this invention has a decentralized decision-making structure. 15 This system allows each unmanned aerial vehicle to assess its own health status and much more. using effective artificial intelligence algorithms to take on tasks in coalitions It allows for multiple tasks, such as priority level, load type, energy consumption, and risk level. They are scored based on numerous criteria, and the most suitable coalitions are awarded the positions according to these scores. This multi-criteria evaluation of tasks, unlike classical methods, only requires 20 a more balanced and sustainable approach that does not rely on distance or route optimization It enables task sharing. The system also uses reinforcement learning-based algorithms to train unmanned aerial vehicles. This allows them to learn about past task performance and health changes. Thus, each An unmanned aerial vehicle can more accurately improve its mission suitability over time. 25 It learns to predict. In this context, especially DDQN - Double Deep Q Network and Priority. Advanced AI techniques, such as Experience Replication, can be used in the system. Additionally, the system aims to enhance the security of task assignments and health data. For this purpose, a blockchain-based verification layer can also be integrated. This structure Thanks to this, task assignments can be verified without the need for a central server, and task 30 4 This prevents manipulations. This structure enhances the cybersecurity of the system. It also enables distributed and fault-tolerant task management. The system is used in military reconnaissance operations where numerous unmanned aerial vehicles operate simultaneously. Search and rescue missions after natural disasters, agricultural observation and spraying, industrial 5 in large-scale application areas such as facility inspection and air cargo transportation It can be used, especially where operational reliability and energy efficiency are critical. In these areas, task assignment is based on the physical condition of the unmanned aerial vehicles. Doing so significantly improves mission performance and fleet lifespan. In conclusion, the invention provides insights into task sharing in autonomous unmanned aerial vehicle swarms. Unlike existing structures, physical health, energy status and task priorities are considered in 10 It offers a holistic AI-powered decision-making mechanism that considers all aspects together. In this respect, the invention is highly customized, scalable, learning, and secure. It can be positioned as a task distribution infrastructure. The structural and characteristic features and all the advantages of the invention are given in the figures below. Thanks to the detailed explanation written with references to these figures, 15 This will be understood, and therefore the evaluation should also take these forms and detailed explanations into consideration. It needs to be done by taking precautions. Figures that will help understand the invention. Figure 1 is a schematic representation of the system that is the subject of the invention. Figure 2 is the flowchart of the method described in the invention. 20 Explanation of Part References 1. System 2. Analysis module 3. Scoring unit 4. Learning unit 25 5. Distribution mechanism 6. Management layer A. Unmanned aerial vehicle Method 1000 1001. By activating the structural health analysis module, the battery of unmanned aerial vehicles, By collecting data on engine temperature, vibration, load capacity, and electrical stress, health assessments are carried out. Calculating the index and transferring it to the system as the initial state, 5 1002. Task: The multi-criteria scoring unit prioritizes tasks in the operational area. scoring and task score values ​​according to load type, risk level and energy cost. by creating and distributing it throughout the system 1003. Multi-agent reinforcement learning unit in the analysis of unmanned aerial vehicles. Using the data from the module and scoring unit, DDQL - Dual Deep Q Learning 10 The algorithm selects the most suitable mission or standby mode for each unmanned aerial vehicle. determining the action 1004. Coalition-based secure and distributed mission among unmanned aerial vehicles. The management layer communicates mission selection recommendations to the entire swarm, for each unmanned aerial vehicle. enabling the vehicle to see the health and task preferences of others, thus facilitating in-group interaction. to ensure coordination 1005. Coalition-based task allocation mechanism, energy level, load capacity and by analyzing mission requirements, the most suitable unmanned aerial vehicle for each mission. forming a coalition and deploying multiple unmanned aerial vehicles for high-risk or arduous missions mediator cooperation 20 1006. Secure and distributed task management layer for tasks selected by the Coalition. Verification and task assignment on the blockchain by smart contract safely announced to the herd 1007. The commencement of unmanned aerial vehicle missions, structural health throughout the process. The analysis module will continue to monitor the health status of the unmanned aerial vehicles, 25 In critical situations, the system removes the unmanned aerial vehicle from the coalition, and the mission is resumed. dispersal and adaptive self-updating of the herd Detailed Description of the Invention 6 In this detailed description, the system (1) and method (1000) that are the subject of the invention are preferred. Their structures are explained solely to facilitate a better understanding of the subject. This invention enables swarms of numerous autonomous unmanned aerial vehicles (A) to perform tasks. sharing, in a smart, secure and coalition-based manner, focused on structural health and energy. a task distribution system (1) and method that enables it to perform efficiently 5 It is related to (1000). The system (1) displays the battery level of each unmanned aerial vehicle (A) in real time. (allowing for the analysis of physical health data such as engine temperature and load carrying capacity) It has local assessment modules that recognize location. This allows tasks to be performed solely based on location. or according to the route, but also the current operational 10 of the unmanned aerial vehicle (A). They can be assigned depending on the situation. The system developed within the scope of the invention (1) has a decentralized decision-making structure. This structure has the health status of each unmanned aerial vehicle (A). and tasks with the help of multi-factor artificial intelligence algorithms It allows them to undertake tasks in coalitions. Tasks, priority level, payload type, energy 15 It is scored based on multiple criteria such as consumption and risk level, and the most suitable option is selected according to these scores. These tasks are undertaken by coalitions. This multi-criteria evaluation of tasks, Unlike classical approaches, it does not rely solely on distance or route optimization, It ensures a more balanced and sustainable division of labor. System (1) also enables unmanned aerial vehicles to 20 with reinforcement learning-based algorithms. (A) enables them to learn about past task performance and health changes. Thus Each unmanned aerial vehicle (A) will become more accurate in its mission suitability over time. It learns to predict in this way. In this context, especially DDQN - Double Deep Q Network and Advanced AI techniques such as Priority Experience Repetition are used in the system (1) It is available. 25 In addition, the system (1) enhances the security of task assignments and health data. For this purpose, a blockchain-based verification layer can also be integrated. This structure Thanks to this, task assignments can be verified without the need for a central server, and tasks This prevents manipulations. This structure ensures the cybersecurity of the system (1). while increasing efficiency, it also enables distributed and fault-tolerant task management. It makes it possible. 7 System (1) is a military reconnaissance system in which a number of unmanned aerial vehicles (A) operate simultaneously. operations, search and rescue missions after natural disasters, agricultural observation and spraying, large-scale applications such as industrial plant inspection and air cargo transportation. It can be used in various fields, particularly in areas such as operational reliability and energy efficiency. In these areas where efficiency is critical, the physical condition of unmanned aerial vehicles (A) is 5 Assigning tasks accordingly significantly improves task performance and fleet lifespan. It increases. In conclusion, the invention facilitates task sharing in swarms of autonomous unmanned aerial vehicles (A). Unlike existing structures in this regard, physical health, energy status and task a holistic AI-powered decision-making mechanism that considers priorities together 10 It offers a highly customized, scalable, and learning capability. and can be positioned as a secure task distribution infrastructure. The system, schematically shown in Figure 1 (1);  Battery level, motor temperature, electrical stress of unmanned aerial vehicles (A), Analyze load carrying capacity, failure history and estimated equipment lifespan data. 15 This analysis generates a health index and a job suitability score. calculating AI-based structural health analysis module (2),  Prioritizing the missions of unmanned aerial vehicles (A), payload type, energy cost and risk. Multi-criteria scoring unit that scores according to its level (3),  Decisions of unmanned aerial vehicles (A) DDQL - Dual Layer Deep Q-Learning 20 Multi-agent reinforcement learning algorithm optimized and learned through reinforcement. learning unit (4),  Unmanned aerial vehicles (A) according to energy, capacity and health index coalition-based system that divides into coalitions and assigns tasks to the most suitable groups. task distribution mechanism (5) and 25  Verifying task assignments on the blockchain, unmanned aerial vehicles (A) They communicate with each other using V2V - Vehicle-to-Vehicle communication based communication. It provides security that protects data integrity by reducing latency through edge processing. and distributed task management layer (6) It includes. 30 Analysis module (2) measures battery, motor temperature, vibration, load capacity and electrical stress. It analyzes the data, generates a health index, and calculates a task suitability score. 8 The scoring unit (3) ranks tasks according to priority, load type, energy cost and risk level. points. Formula: Task_Score = w₁·Priority + w₂·LoadMatch + w₃·EnergyCost⁻¹ + It is in the form of w₄·Risk⁻¹. Learning unit (4), decisions of unmanned aerial vehicles (A) DDQL - Double Layer Deep Optimizes with Q-Learning. Learning unit (4), 5 Learning is enabled by the formula Q(sₜ,aₜ)=Q(sₜ,aₜ)+α[rₜ+γQ(sₜ₊₁,argmaxQ)-Q(sₜ,aₜ)]. Task distribution mechanism (5), unmanned aerial vehicles (A) energy, capacity and health It divides them into coalitions according to its index and assigns tasks to the most suitable groups. The task management layer (6) verifies task assignments on the blockchain, unmanned It enables communication between aircraft (A) and reduces flight processing and delay by 10 It protects data integrity by reducing data loss. The invention addresses existing task sharing in swarms of autonomous unmanned aerial vehicles (A). Ignoring structural health data observed in buildings, energy inefficiency, the inability of centralized control structures to scale, and disregard for task prioritization Developed to eliminate technical problems such as those mentioned above, it is a multi-layered, 15 It includes a distributed and AI-based task distribution system (1). The system’s (1) basic solution approach is to determine tasks based solely on location or route cost. not according to; unmanned aerial vehicles (A) physical health status, energy level, load numerous criteria such as carrying capacity, mission priority, payload type and estimated mission duration a decentralized, coalition-based artificial intelligence that assigns tasks based on evaluation criteria. It is about presenting architecture. This solution is structured around the following fundamental elements and algorithms. The analysis module (2), which is the agent that runs on each unmanned aerial vehicle (A), unmanned aerial It analyzes the structural health data of the vehicle (A) in real time. Battery level, motor temperature, electrical stress, load carrying capacity, failure history and estimated 25 Hardware lifecycle parameters are continuously monitored. This agent converts the above data into a state vector, and this vector is then used as a task vector. It draws on a reinforcement learning model for use in suitability decisions. Example state vector: 9 s_t={battery_level,motor_temperature,load_capacity,health_index,location,task_urgency} yeti} Each unmanned aerial vehicle (A) participates in the task-sharing process as a learner. This The agents are trained by the learning unit (4) using the DDQN - Double Deep Q Network algorithm. This structure reduces the overvaluation problem experienced by classic DQN – Dual Deep Network, resulting in a more efficient system. It enables sound and accurate decisions to be made. Q-Value Update Function: 𝑄(𝑠 , 𝑎 ) = 𝑄(𝑠 , 𝑎 ) + 𝛼[𝑟 + 𝛾𝑄(𝑠 , arg max 𝑄(𝑠 , 𝑎 )) − 𝑄(𝑠 , 𝑎 )] In this learning process: s_t: Current health and environmental status of unmanned aerial vehicle (A). 10 a_t: The act of accepting or rejecting a specific task. r_t: The reward given for successful completion of the task; energy efficiency and It is weighted by task type. Additionally, by using prioritized experience repetition, more emphasis is placed on more important experiences. Learning weight is given. Each task in the system (1) is scored according to specific criteria. This 15 criteria;  priority of tasks,  load type and sensitivity,  route length and energy cost and  risk level, for example, whether it is a hazardous area or not 20 It is in this form. These criteria were determined by the task multi-criteria scoring unit (3), MCDM - Multi-Criteria Decision. It is weighted using the Giving Model. Available processes include AHP - Analytical. Hierarchy Process, TOPSIS, or Fuzzy Logic-based structures can be used. General task score formula: 25 𝑇𝑎𝑠𝑘 = 𝑤 ⋅ 𝑃𝑟𝑖𝑜𝑟𝑖𝑡𝑦 + 𝑤 ⋅ 𝐿𝑜𝑎𝑑𝑀𝑎𝑡𝑐ℎ + 𝑤 ⋅ 𝐸𝑛𝑒𝑟𝑔𝑦𝐶𝑜𝑠𝑡 + 𝑤 ⋅ 𝑅𝑖𝑠𝑘 It is in this form. Here w_1, w_2, w_3, w_4 are weighting coefficients of the system (1) manager or learning agents. It can be updated dynamically. Depending on the complexity of the tasks, multiple unmanned aerial vehicles (A) instead of a single one may be used. The unmanned aerial vehicle (A) may need to cooperate. In this case, task distribution 5 mechanism (5), coalition formation algorithm and the most suitable unmanned aerial vehicle for the mission It brings together its tools (A). Coalition formation, predefined minimum capacity It is based on meeting their needs. Every coalition;  Total battery level must be sufficient to complete the task. 10  Total load capacity must be able to support the load.  The weakest unmanned aerial vehicle (A) in the coalition, preventing completion of the mission. should not be It must meet the following conditions. After coalitions are formed, the coalition with the highest score among the tasks becomes coalition 15. It is selected by them. Task management layer (6), eliminating central control and increasing security. For this purpose, it stores task assignments on a private blockchain network. This blockchain network is unmanned. The aircraft (A) operate in a synchronized manner and the mission assignments;  prevents unauthorized modification, 20  It ensures traceability,  withdrawal of the system (1) in case of dispute makes it possible. Possible architectures;  Practical Byzantine Fault Tolerance -PBFT, 25  Directed Acyclic Graph (DAG) based structures,  Task assignment verification via smart contract It is in this form. 11 Unmanned aerial vehicles (A) are monitored as a time series of health data and in the future Their performance is predicted. For this purpose, LSTM - Long Short-Term Memory or Kalman is used. Time series modeling techniques such as filtering are used. Thus, the system (1) is used in the short term. by keeping the unmanned aerial vehicles (A) that are likely to malfunction out of the mission It increases the success rate. 5 The method presented in the flowchart in Figure 2 (1000);  By activating the structural health analysis module (2), unmanned aerial vehicles (A) battery, motor temperature, vibration, load capacity and electrical stress data by collecting the health index and the initial status of the system (1) transfer as (1001), 10  task multi-criteria scoring unit (3), tasks in the operational area Priority scoring and task score based on load type, risk level, and energy cost. By creating its values, it is published to the system in a distributed manner (1) (1002),  analysis of the multi-agent reinforcement learning unit (4), unmanned aerial vehicles (A) Using the data of module (2) and scoring unit (3) DDQL - Double Deep 15 With the Q Learning algorithm, task selection for each unmanned aerial vehicle (A) or determining the most appropriate action in the form of waiting (1003),  Coalition-based secure and distributed mission between unmanned aerial vehicles (A) The management layer (6) transmits task selection suggestions to the entire swarm, each unmanned the aircraft (A) enables others to see the health and mission preferences and 20 thus ensuring intra-herd coordination (1004),  coalition-based task distribution mechanism (5), energy level, load By analyzing the capacity and mission requirements, the most suitable unmanned aerial vehicle for each mission forming an aircraft (A) coalition and for high-risk or heavy missions Multiple unmanned aerial vehicle (A) cooperation (1005), 25  Secure and distributed task management layer for tasks selected by the coalition Verification and task assignment on the blockchain by (6) smart the safe announcement of the contract to the herd (1006) and  Commencement of mission execution by unmanned aerial vehicles (A), structural changes throughout the process health analysis module (2) health status of unmanned aerial vehicles (A) 30 Continuing to monitor the system (1) unmanned aerial vehicle (A) during critical drops removing it from the coalition, redistributing the task, and the herd adaptively self-updating (1007) 12 It includes the steps involved in the process. Method (1000) task distribution in swarms of autonomous unmanned aerial vehicles (A). to achieve this, the interactive work of five complementary fundamental elements It is based on the method (1000), the working process, firstly, the structural health analysis module. (2) each unmanned aerial vehicle's (A) battery level, motor temperature, vibration 5 Real-time analysis of physical parameters such as load capacity and electrical stress. It starts with monitoring. The analysis module (2) processes all this data and unmanned aerial It produces a health index representing the overall operational suitability of the vehicle (A) and this Value then forms the main input for subsequent task selection processes. Following this stage, the task multi-criteria scoring unit (3) in the operational area 10 It analyzes the tasks and assigns priority to each task based on factors such as load type, task duration, and regional. It subjects them to an objective scoring based on multiple criteria such as risk level and energy cost. The system (1) Task Score for this evaluation = w₁·Priority + w₂·Load Matching + Prioritize tasks using the formula w₃·EnergyCost⁻¹ + w₄·Risk⁻¹ Thus, the distribution of tasks among unmanned aerial vehicles (A) forms a numerical 15 It will be based on a solid foundation. These scores are based on the (A) self-health status of the unmanned aerial vehicles and multi-agent reinforcement. The learning unit (4) is evaluated by each unmanned aerial vehicle in the learning unit (4). Agent (A) acts as an autonomous agent and uses the DDQN algorithm to calculate Health Index+ It selects the action most appropriate to its current situation in the form of Task Points + environmental conditions. This 20 Q(sₜ,aₜ)=Q(sₜ,aₜ)+α[rₜ+γQ(sₜ₊₁,argmaxQ)-Q(sₜ,aₜ)] update equation in the selection process The system (1) is used to take on tasks, pass on tasks or join a coalition in this way. It dynamically optimizes decisions such as these. Then, depending on the nature of the tasks, coalition-based task allocation will be implemented. The distribution mechanism (5) is activated. The distribution mechanism (5) is activated by a single unmanned aerial vehicle (A) 25 too complex, risky, or high load capacity to be handled by Coalitions form groups (A) of suitable unmanned aerial vehicles for tasks requiring; Based on criteria such as Energy Level, Health Index and Load Carrying Capacity The system (1) selects the most suitable unmanned aerial vehicle (A) community for each task. It selects automatically. Thus, mission success, unmanned aerial vehicle (A) health and energy 30 The balance is maintained at the maximum level. 13 The final stage involves task assignment and coordination between unmanned aerial vehicles (A). This is performed by the secure and distributed task management layer (6). Management layer (6) eliminates the risk of changing task assignments with blockchain-based verification. removes. The management layer (6) provides communication between V2V unmanned aerial vehicles (A) and all It synchronizes the status information of coalition members and, thanks to its edge processing capabilities, 5 It enables calculations to be performed locally without delay. This way, both... Operational safety as well as communication speed and integrity are guaranteed. As a result of these five elements working together, the method (1000) provides real-time health analysis, multi-criteria evaluation of tasks, AI-based decision-making, dynamic coalition building and secure distributed mission validation processes in an integrated structure 10 It performs within it. Thus, swarms of unmanned aerial vehicles (A) are autonomous, reliable and Energy-optimized task allocation is ensured. The system used in the invention (1) is the task of swarms of autonomous unmanned aerial vehicles (A). It has a multi-layered architecture that optimizes sharing. This architecture ensures that each unmanned surface is optimized. Health analysis assessing the physical condition of the aircraft (A), mission specifications 15 A scoring structure that models numerically, based on multi-factor AI-driven decision-making. the mechanism, coalition-based task allocation and secure execution of tasks It relies on the integrated operation of distributed structures that enable verification. The process, initially... It is initiated by the analysis module (2). In the analysis module (2) unmanned aerial Battery levels, motor temperatures, vibration data, ESC telemetry and 20 of vehicle (A) All raw data from the load capacity sensors is received in timestamped form. The data is not processed directly; to reduce noise and the actual data of the unmanned aerial vehicle (A) An extended Kalman filter is applied to accurately model the physical state. This The basic state prediction model used during the process is as follows: State Prediction Equation: 25 𝑥 = 𝑓(𝑥 , 𝑢 ) + 𝑤 Measurement Equation: 𝑧 = ℎ(𝑥 ) + 𝑣 This filtering process increases the consistency of the data and the instantaneous 30 of the unmanned aerial vehicle (A). This allows for a more accurate calculation of the situation. Then, this data... 14 Operational suitability of the unmanned aerial vehicle (A) called the "Health Index" is determined using The value is calculated. The formula used for the health index is as follows: Health Index Calculation: 𝐻𝑒𝑎𝑙𝑡ℎ𝐼𝑛𝑑𝑒𝑥 = 𝛼 𝐸 + 𝛼 1 − T T + 𝛼 1 − 𝑉 + 𝛼 𝐶 This index measures the energy level, engine temperature safety, and vibration of the unmanned aerial vehicle (A). a weighted combination of fundamental parameters such as structural strength and load capacity It includes a time series-based LSTM model, energy consumption, and engine wear. By analyzing trends, it calculates the probability of a possible failure in the near future. This prediction... If the detected fault value is above a certain threshold value, the unmanned aerial vehicle (A) is safe. It is moved to the non-existent category and removed from the task pool. 10 After this physical analysis process is completed, the information is scored by the scoring unit (3). Tasks are processed. Priority level, payload type and unmanned aerial vehicle (A) compatibility, energy based on criteria such as consumption forecast, operational risk, weather effects, and mission duration. Tasks are evaluated. The final score for the tasks is calculated using the following multi-criteria formula: Task Scoring Formula: 15 𝑇𝑎𝑠𝑘_𝑆𝑐𝑜𝑟𝑒 = 𝑤 ⋅ 𝑃𝑟𝑖𝑜𝑟𝑖𝑡𝑦 + 𝑤 ⋅ 𝐿𝑜𝑎𝑑𝑀𝑎𝑡𝑐ℎ + 𝑤 ⋅ 𝐸𝑛𝑒𝑟𝑔𝑦𝐶𝑜𝑠𝑡 + 𝑤 ⋅ 𝑅𝑖𝑠𝑘 This calculation allows various tasks to be compared with each other. Furthermore... To measure the similarity between tasks, a task similarity matrix is ​​created: Task Similarity: 𝑆𝑖𝑚(𝑖, 𝑗) = 𝑒 ( , ) 20 This matrix allows similar tasks to be assigned to the same coalition, thus facilitating the operation. Its efficiency increases. Unmanned aerial vehicle (A) health indices and mission scores learning simultaneously It is used by unit (4). In this system (1) each unmanned aerial vehicle (A) is a learner It acts like an agent. The agent's Health Index + Mission Score + environmental parameters + neighboring 25 By processing the state vector it receives in the form of unmanned aerial vehicle (A) information, it optimizes the situation. It selects the task or action. The decision-making process is optimized with the DDQN approach. System (1) applies the following basic learning rule: DDQN Update Equation: 𝑄(𝑠 , 𝑎 ) = 𝑄(𝑠 , 𝑎 ) + 𝛼[𝑟 + 𝛾𝑄(𝑠 , arg max 𝑄(𝑠 , 𝑎 )) − 𝑄(𝑠 , 𝑎 )] Unmanned aerial vehicles (A) share their policies partially with their neighbors, forming the swarm of 5 This allows for more harmonious operation. This synchronization is achieved using the following equation: Policy Synchronization: 𝑄 = (1 − 𝜇)𝑄 + 𝜇 ⋅ 𝑚𝑒𝑎𝑛(𝑄 ) The reward function used in the system (1) promotes individual success, cooperative harmony and security. They evaluate flight behavior together: 10 Cooperative Award Structure: 𝑅 = 𝑅 + 𝜂𝑅 + 𝜉𝑅 Grouping process in accordance with the decisions made by unmanned aerial vehicles (A) Coalition This is carried out by the task distribution mechanism (5) based on the task distribution mechanism. At this stage, the tasks Current health and energy status of unmanned aerial vehicles (A) with the required capacity 15 They are matched. The suitability of each coalition is evaluated using the following cost function. Coalition Cost Function: 𝐶𝑜𝑠𝑡(𝐶𝑜𝑎𝑙𝑖𝑡𝑖𝑜𝑛 ) = 𝛿 ∑𝐸𝑛𝑒𝑟𝑔𝑦 + 𝛿 ∑(1 − 𝐻𝑒𝑎𝑙𝑡ℎ ) + 𝛿 ∣ 𝑀𝑖𝑠𝑚𝑎𝑡𝑐ℎ(𝑝𝑎𝑦𝑙𝑜𝑎𝑑, 𝑡𝑎𝑠𝑘) ∣ The goal is to choose the coalition that minimizes this cost. Large-scale measures are necessary when needed. Coalitions are divided into subgroups, reducing communication load and synchronization delay. is reduced. At the end of these processes, the tasks are verified and safely distributed to the entire swarm. The announcement is made by the secure and distributed task management layer (6). This The layer uses a blockchain-based “Task Commitment Protocol”. The protocol uses “proposals”, It consists of "signing" and "approval" steps. Task assignments PBFT consensus 25 It is added to the blockchain via the mechanism. Communication between V2V unmanned aerial vehicles (A) 16 thanks to the mission status of unmanned aerial vehicles (A) in real time It is updated. Thanks to its edge processing capabilities, collision avoidance models and energy predictions are possible. Algorithms and MARL - Multi-Agent Reinforcement Learning inferences are run locally and This reduces communication delay. When all these steps are combined; health analysis, task evaluation, autonomous decision 5 The layers of granting, coalition building, and secure mission verification work together as a whole. by working, unmanned aerial vehicle (A) swarms are continuously learning, adaptable and reliable This enables it to have a task distribution system (1). The system (1) has its own in each cycle. It evaluates its performance, updates its parameters based on the data it obtains, and assigns tasks. It creates a stable autonomous swarm architecture by optimizing its distribution. 10

Claims

17 REQUESTS 1. Task sharing of swarms consisting of numerous autonomous unmanned aerial vehicles (A), smart, safe and efficient, based on a coalition focused on structural health and energy. It is a task distribution system (1) which enables it to perform in this way, and its feature is;  Battery level, motor temperature, electrical stress, of unmanned aerial vehicles (A), 5 Analyze data on load carrying capacity, failure history, and estimated equipment lifespan. This analysis generates a health index and a job suitability score. calculating AI-based structural health analysis module (2),  Prioritizing the missions of unmanned aerial vehicles (A), payload type, energy cost and risk. Multi-criteria scoring unit that scores according to its level (3), 10  Decisions of unmanned aerial vehicles (A) are made using DDQL - Dual Layer Deep Q-Learning Multi-agent reinforcement learning algorithm optimized and learned through reinforcement. learning unit (4),  Unmanned aerial vehicles (A) according to energy, capacity and health index 15 coalition-based systems that divide into coalitions and assign tasks to the most suitable groups. task distribution mechanism (5) and  Verifying task assignments on the blockchain, unmanned aerial vehicles (A) They communicate with each other using V2V - Vehicle-to-Vehicle communication based communication. It provides security that protects data integrity by reducing latency through edge processing. and distributed task management layer (6) 20 It includes.

2. Task sharing of swarms consisting of numerous autonomous unmanned aerial vehicles (A), smart, safe and efficient, based on a coalition focused on structural health and energy. It is a task distribution method (1000) that enables it to perform in this way, and its feature is;  By activating the structural health analysis module (2), the unmanned aerial vehicles (A) 25 battery, motor temperature, vibration, load capacity and electrical stress data by collecting the health index and the initial status of the system (1) as (1001),  task multi-criteria scoring unit (3), tasks in the operational area Priority is scored based on load type, risk level, and energy cost, and the task score is 30. By creating its values, it is published to the system in a distributed manner (1) (1002), 18  analysis of the multi-agent reinforcement learning unit (4), unmanned aerial vehicles (A) Using the data of module (2) and scoring unit (3) DDQL - Double Deep With the Q Learning algorithm, task selection for each unmanned aerial vehicle (A) or determining the most appropriate action in the form of waiting (1003),  Coalition-based secure and distributed mission between unmanned aerial vehicles (A) 5 The management layer (6) transmits task selection suggestions to the entire swarm, each unmanned the aircraft (A) enables others to see the health and mission preferences and thus ensuring intra-herd coordination (1004),  coalition-based task distribution mechanism (5), energy level, load By analyzing the capacity and mission requirements, we select the most suitable unmanned 10 for each mission. forming an aircraft (A) coalition and for high-risk or heavy missions (1005) to realize multiple unmanned aerial vehicle (A) cooperation,  Secure and distributed task management layer for tasks selected by the coalition Verification and task assignment on the blockchain by (6) smart the safe announcement to the herd by contract (1006) and 15  Commencement of mission execution by unmanned aerial vehicles (A), structural changes throughout the process health status of (A) unmanned aerial vehicles (2) health analysis module Continuing to monitor the system (1) unmanned aerial vehicle (A) during critical drops removing it from the coalition, redistributing the task, and the herd adaptively self-updating (1007) 20 It includes the steps of the process.

Images