Coordinating multiple drones in formation requires addressing distributed decision-making challenges across varying spatial scales. Field tests demonstrate that communication latency between drones can exceed 120ms in complex environments, while positioning errors accumulate at rates of 2-5cm per minute in GPS-denied scenarios. These technical limitations constrain the scalability of swarm operations, where coordination complexity increases non-linearly with each additional unit.

The fundamental challenge lies in balancing centralized control structures against distributed autonomy while maintaining both system resilience and operational efficiency.

This page brings together solutions from recent research—including hierarchical clustering with master-slave configurations, scene-centric neural networks for trajectory prediction, environment-driven coordination systems, and multi-level path planning architectures with both swarm and individual conflict avoidance. These and other approaches provide practical frameworks for implementing drone swarms that maintain formation integrity even when facing communication degradation or environmental uncertainties.

1. Distributed Pursuit–Evasion Game Decision-Making Based on Multi-Agent Deep Reinforcement Learning

yi lin, han gao, yuanqing xia - Multidisciplinary Digital Publishing Institute, 2025

Pursuitevasion games are a fundamental framework for advancing autonomous decision-making and cooperative control in multi-UAV systems. However, the application of reinforcement learning to pursuitevasion involving fixed-wing UAVs remains challenging due constraints, such as minimum velocity, limited turning radius, high-dimensional continuous action spaces. To address these issues, this paper proposes method that integrates automatic curriculum with multi-agent proximal policy optimization. A self-play mechanism is introduced simultaneously train both pursuers evaders, enabling dynamic adaptive encirclement strategies. In addition, reward structure specifically tailored task was designed guide gradually achieving evader while ensuring their own safety. further improve training efficiency convergence, develops subgame progressively exposes agents increasingly complex scenarios, facilitating experience accumulation skill transfer. The simulation results demonstrate proposed approach improves pursuit performance under realistic UAV dynamics. This work provides practical scalable ... Read More

2. Multi-Stage Collision Avoidance System for Drones with Real-Time Imaging and Path Adjustment

SONY GROUP CORP, 2025

Dynamic collision avoidance for multiple drones operating in a shared airspace. The system employs a multi-stage approach to prevent collisions between drones, where each drone continuously monitors its surroundings and adjusts its behavior based on the presence of other drones. The system uses a combination of real-time imaging, collision detection, and path management to dynamically reorient and change drone paths to avoid collisions. This approach enables efficient collision avoidance while maintaining operational flexibility and adaptability among multiple drones.

3. AI-Based System for Predictive Management and Autonomous Response in UAV Swarm Operations

SKYWAY TECHNOLOGIES CORP, 2025

Managing a group of UAVs (drones) using AI to predict and respond to unexpected events during flight. The system uses AI to analyze UAV flight paths and predict potential issues. If an issue is identified, it takes action to mitigate it. For example, if a UAV is predicted to collide with another, it could divert that UAV's path. If a UAV loses communication, it could transfer control to a nearby UAV. The AI also helps allocate UAVs to handle unexpected events based on real-time parameters. This allows managing multiple UAVs in complex scenarios with reduced risk.

4. Distributed Model Predictive Formation Control for UAVs and Cooperative Capability Evaluation of Swarm

ming yang, xiaoyi guan, mingming shi - Multidisciplinary Digital Publishing Institute, 2025

This paper utilizes the distributed model predictive control (DMPC) method to investigate formation problem of unmanned aerial vehicles (UAVs) in obstacle environment and establishes cooperative capability evaluation metrics swarm. Based on DMPC approach, cost function is constructed adjust relative positions velocities UAVs, ensuring desired formation. Additionally, address avoidance formation, designed provide safe environment. To evaluate we design from multiple dimensions reflect swarms capability. Finally, simulation results show effectiveness with applicability metrics.

5. Reinforcement Learning-Based Optimization of Quality-of-Service and Path Planning for Multi-UAV Mobile Edge Computing

xiang li, qi liu, zhuocheng yang - Darcy & Roy Press Co. Ltd., 2025

Unmanned Aerial Vehicles (UAVs) have traditionally served as network processors in mobile networks, and more recently, servers Mobile Edge Computing (MEC). However, deploying UAVs dynamic obstacle-rich environments while ensuring efficient coordination among multiple presents significant challenges. To tackle these challenges, we propose a unified reinforcement learning framework for multi-UAV MEC platform that enhances Quality-of-Service (QoS) optimizes path planning. Specifically, our contributions include: (1) Integrating QoS optimization UAV planning into framework; (2) Modeling terminal users' demands with sigmoid-like function to enhance service quality; (3) Incorporating demand, risk factors, geometric distance the reward matrix balancing quality, mitigation, cost efficiency. Experimental results demonstrate frameworks effectiveness: simulations 3D show 2035% reduction collision incidents compared greedy algorithms 1525% improvement demand fulfillment rates under user distributions. Further analysis highlights impact of sigmoid parameter tuning on risk-QoS trade-offs,... Read More

6. Towards Optimal Guidance of Autonomous Swarm Drones in Dynamic Constrained Environments

yunes alqudsi, murat makaraci - Wiley, 2025

ABSTRACT As autonomous drone swarms become increasingly important for complex missions, there remains a critical need integrated approaches that can simultaneously handle task allocation and safe navigation in dynamic environments. This paper addresses the challenge of optimally allocating tasks generating collisionfree trajectories operating obstaclerich settings. Our proposed Swarm Allocation Route Generation (SARG) framework integrates optimal assignment with dynamically feasible trajectory planning, enabling efficient mission completion while ensuring through 3D workspaces. Using quadrotors as our experimental platform, incorporates both DronetoObstacle DronetoDrone collision avoidance algorithms, alongside modified path planning algorithm enhances simultaneous graph search efficiency. extensive experiments demonstrate SARG significantly improves performance over existing approaches. The framework, maintaining 100% rate dense environments, achieves 21.6% reduction computation time searching phase compared to conventional methods, contributing overall system Thes... Read More

7. Semi-Autonomous Drone Swarm with Dynamic Capability-Based Team Formation and Task Execution

ANDURIL INDUSTRIES INC, 2025

Dynamic grouping of semi-autonomous drones to efficiently perform tasks like surveillance, delivery, mapping, etc. The drones can self-organize into teams based on capability matching to execute complex operations. A server determines tasks and assigns drones based on capabilities. Lead drones decompose tasks and create plans. Followers execute tasks. If a leader fails, another drone takes over. This allows flexible and scalable drone swarms without centralized control.

US12293670B2-patent-drawing

8. A Hybrid Optimization Framework for Dynamic Drone Networks: Integrating Genetic Algorithms with Reinforcement Learning

mustafa ulas, anil sezgin, aytug boyaci - Multidisciplinary Digital Publishing Institute, 2025

The growing use of unmanned aerial vehicles (UAVs) in diverse fields such as disaster recovery, rural regions, and smart cities necessitates effective dynamic drone network establishment techniques. Conventional optimization techniques like genetic algorithms (GAs) particle swarm (PSO) are weak when it comes to real-time adjustment the environment multi-objective constraints. This paper proposes a hybrid framework combining reinforcement learning (RL) improve deployment networks. We integrate Q-learning into GA mutation process allow drones adaptively adjust locations real time under coverage, connectivity, energy In scenario large-scale simulations for wildfire tracking, response, urban monitoring tasks, approach performs better than PSO. greatest enhancements 6.7% greater 7.5% less average link distance, faster convergence optimal deployment. proposed allows establish strong stable networks that nature adapt mission demands with efficient coordination. research has important applications autonomous UAV systems mission-critical where adaptability robustness essential.

9. UAV Collision Avoidance via Intersection Time Calculation and Adaptive Path Adjustment

ZIPLINE INTERNATIONAL INC, 2025

UAV collision avoidance method that allows multiple UAVs to fly safely in proximity without constant communication. UAVs exchange flight information with nearby UAVs. They calculate intersection times based on the received flight info. If an intersection is imminent, they adjust their flight paths to avoid collision. This allows UAVs to dynamically adapt their paths to prevent mid-air collisions without constant communication.

10. On the Characteristics of Next Generation for Redundant Clustered Reliable Data Transmission Scheme in Critical IoT Infrastructures

grace khayat, constandinos x mavromoustakis, george mastorakis - Research Square, 2025

<title>Abstract</title> A wireless network composed of UAVs collaborating to upload data the inter-net is known as a swarm unmanned aerial vehicles (S-UAVs). are utilized extensively convey important during emergency situations like earthquakes or floods. Clustering, which divides into clusters with cluster head (CH) and members (CM), one most dependable routing strategies in S-UAVs. disconnected loss result damaged non-functional CH, topic that rarely explored. This work proposes an improved multiple redundant strategy based on three weighted parameters: energy, rewarding, propagation delay for reliable transmission through ensuring functional CH. paper exclusively uses coefficient variation set weights used calculation generates new parameter denoted index. Based index, nodes selected CH CMs, thus forming clusters. To decrease probability we propose cooperation different types 1 agents, our case. Having agents will increase systems reliability, risk non-functionality crisis scenario depends highly agents type. ensure take over duties connection stability. Every node has ensur... Read More

11. Scene-Centric Neural Network for Linear-Scale Trajectory Prediction of Multiple Agents

WAYMO LLC, 2025

Efficiently predicting the future trajectory of an agent in an environment using a scene-centric neural network instead of separate agent-centric networks for each agent. The scene-centric network encodes the environment and all agents using a single input centered on a fixed point. This reduces computational cost compared to separate agent-centric inputs for each agent, as the scene encoding scales linearly with the number of agents instead of quadratically.

US2025121857A1-patent-drawing

12. Drone Swarm Communication System with Hierarchical Clustering and Master-Slave Configuration

ICTK CO LTD, 2025

Optimizing communication in swarms of drones to enable efficient and reliable control of large numbers of drones. The optimization involves clustering the drones into groups with a master drone that communicates with a central server, and slave drones that relay messages from the master. This reduces the number of required communication channels compared to each drone directly connecting. Clustering also allows faster area coverage, obstacle avoidance, and resource sharing. If a master fails, another slave can be promoted. This enables robust swarm operation by minimizing communication breakdowns.

13. Underwater Robotic Fish Swarm with Acoustic and Optical Communication for Data Collection and Transmission

KHALIFA UNIVERSITY, TECHNOLOGY INNOVATION INSTITUTE - SOLE PROPRIETORSHIP LLC, 2025

Underwater robotic fish swarm for remote monitoring and virtual reality exploration of underwater environments. A floating platform communicates instructions from an operator to a submersible sinker and swarm of underwater drones to navigate and collect data in an underwater area. The collected data is transmitted back to the platform for display on a virtual reality headset worn by the operator. The swarm uses acoustic and optical communication, localization algorithms, and image compression to enable efficient underwater exploration and data transfer.

14. Fixed-Wing Unmanned Aerial System Navigation Using Relative Motion Estimation and Cooperative Constraints

BRIGHAM YOUNG UNIVERSITY, 2025

A system and method for navigating fixed-wing unmanned aerial systems (UAS) in environments without or with degraded global positioning System (GPS) signals. The method uses relative motion estimation and optimization to improve local navigation and leverage occasional GPS measurements and cooperative constraints from other UAS for global positioning. The UAS estimates its motion relative to the environment using an onboard sensor fusion algorithm like an extended Kalman filter. It then optimizes a back-end pose graph representing global position by incorporating local motion estimates and occasional GPS measurements as constraints. Sharing range measurements and resetting simultaneously between UAS allows leveraging cooperative constraints. This improves accuracy compared to relying solely on local sensing in GPS-denied environments.

US12277716B2-patent-drawing

15. Distributed Device Coordination System with Environment-Driven Operation Management

KABUSHIKI KAISHA YASKAWA DENKI, 2025

Coordinated control of multiple distributed devices like robots using an environment manager that monitors device operations and updates environment information. The devices monitor the environment info and if it meets a condition, they execute a specific operation. This allows devices to coordinate actions based on shared environment data without central control. The environment manager updates the environment info as devices operate, allowing coordination across devices.

US12275155B2-patent-drawing

16. Distributed Formation Planning for Unmanned Aerial Vehicles

zeming zhao, xiaozhen zhang, hao fang - Multidisciplinary Digital Publishing Institute, 2025

Formation flying of multiple unmanned aerial vehicles (UAVs) has attracted much attention for its versatility in cooperative tasks. In this paper, a distributed formation planning method is proposed UAVs. First, we design path searching algorithm, swarm-A*, which can enhance the cohesion swarm, i.e., preventing disintegration swarm when it encounters an obstacle. Then, after waypoint reallocation, trajectory optimization framework formulated. Smooth trajectories UAVs to travel safely obstacle-laden environments be obtained by solving problem. Next, tracking controller based on sliding mode control designed, ensuring that follow planned under dynamic constraints. Finally, numerical simulations and experiments are conducted validate effectiveness method.

17. Learning Verified Safe Neural Network Controllers for Multi-Agent Path Finding

mingyue zhang, nianyu li, yi chen - Association for the Advancement of Artificial Intelligence, 2025

Multi-agent path finding (MAPF) is a safety-critical scenario where the goal to secure collision-free trajectories from initial desired locations. However, due system complexity and uncertainty, integrating learning-based controllers with MAPF challenging cannot theoretically guarantee safety of learned controllers. In response, our study proposes verified safe multi-agent neural control (VSMANC) approach for MAPF, focusing on unified training Decentralized Control Barrier Functions (DCBF) enhence safety. VSMANC enables all agents concurrently learn DCBFs using loss function designed maximize safety, adhere standard policies, incorporate path-finding-related heuristics. We also propose formal verification-guided retraining process both verify properties generate counterexamples retraining, thereby providing guarantee. validate through shape formation experiments UAV simulations, demonstrating significant improvements in effectiveness complex environments.

18. Unmanned Vehicle System with Autonomous Payload Launch and Recovery in Marine and Submarine Environments

MARITIME TACTICAL SYSTEMS INC DBA MARTAC, 2025

An unmanned vehicle system that enables autonomous launch and recovery of payloads like vessels, equipment, and people in marine and submarine environments. The system uses a host vehicle like a submersible drone to carry and deploy guest vehicles like mini-subs, surface drones, and payload carriers. The host and guest vehicles can operate independently in different modes like air, surface, and sub-surface. This allows covert missions involving multiple environments and tasks. The host vehicle provides mobility, maneuverability, and launch capabilities while the guest vehicles perform specialized missions. The system enables multi-environment unmanned vehicle swarms that can be coordinated in time and space for complex missions.

19. Autonomous Drone System with Coordinated Path Navigation for Gas Turbine Engine Inspection

RTX CORP, 2025

Using a team of drones equipped with inspection sensors to simultaneously and autonomously inspect a gas turbine engine. The drones are choreographed to move along specific paths to cover the entire engine area. If an abnormality is detected, a drone can retrace the path to confirm. If a drone takes too long, another drone is summoned to finish. This allows thorough engine inspection without engine removal or downtime.

US2025117012A1-patent-drawing

20. Centralized Flight Zone Allocation System for Mobile Terminals Utilizing Multi-Network Cellular Coordination

NEC CORP, 2025

Managing flights of mobile terminals like drones that use multiple cellular networks. A central management apparatus allocates flight zones across cellular networks based on shared 3D space. This involves dividing airspace into zones identified by latitude, longitude, and altitude. The management apparatus distributes this zone allocation data to the cellular networks. Each network then uses it to coordinate flights of devices connected to their network. This ensures consistent flight management across networks when multiple devices fly in proximity.

US2025118214A1-patent-drawing

21. Collaborative Control Method for Unmanned Cluster Systems with Adaptive Robust Controllers and Dynamic Model Incorporating Parameter Uncertainty and Network Attack Impacts

22. Hierarchical Path Planning System for UAV Swarms with Swarm-Level and Individual-Level Conflict Avoidance

23. Multi-Agent Learning System with Population-Based Curriculum, Hierarchical Temporal, and Behavior Adaptation Techniques

24. Unicast Air-to-Everything Communication System for Drones with Direct Communication Request and Link Establishment Protocol

25. Centralized Trajectory Error Detection and Correction System for Multi-Component Devices with Look-Ahead Trajectory Provision

When it comes to drone swarm coordination techniques, researchers are making great progress. Drone swarms will soon be able to realize their full potential thanks to innovations like autonomous collision-avoidance controllers, centralized fleet management systems, and coordinated UAVs that monitor the surroundings in real time.

Get Full Report

Access our comprehensive collection of 118 documents related to this technology