Trend AnalysisEngineering

Drone Swarm Coordination: From Centralised Control to Decentralised Intelligence

A single drone can inspect a building. A swarm of drones can map a disaster zone, monitor a wildfire perimeter, or coordinate a search-and-rescue operation โ€” but only if they can plan paths, allocate ...

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

The Question

A single drone can inspect a building. A swarm of drones can map a disaster zone, monitor a wildfire perimeter, or coordinate a search-and-rescue operation โ€” but only if they can plan paths, allocate tasks, and avoid collisions without a centralised controller that becomes a single point of failure. The fundamental challenge is coordination under communication constraints: each drone has limited sensing range, intermittent communication with neighbours, and must make decisions in real-time. Can decentralised algorithms achieve the efficiency of centralised planning while maintaining the robustness of autonomous operation?

Landscape

Qiu et al. (2025) proposed a hybrid game theory and particle swarm optimisation (PSO) approach for multi-robot 3D path planning. They framed robot coordination as a non-cooperative game where each robot optimises its own trajectory while accounting for others' presence as constraints. The PSO component handles the continuous trajectory optimisation, while the game-theoretic framework ensures no robot has an incentive to deviate โ€” a Nash equilibrium condition that guarantees stable, collision-free paths even without explicit inter-robot communication.

Alqudsi (2024) tackled the joint problem of task allocation and trajectory optimisation for autonomous drone swarms. In practical scenarios (e.g., agricultural monitoring, infrastructure inspection), drones must simultaneously decide which tasks to perform and how to reach them. Solving these sequentially โ€” first allocate tasks, then plan paths โ€” leads to suboptimal solutions because path costs depend on allocation and vice versa. Alqudsi's synchronous approach improved total mission completion time by optimising both decisions jointly.

On the hardware side, Ganduri et al. (2025) built Woxbots, a low-cost swarm robotics platform that bridges the gap between simulation and real-world deployment. Most published swarm algorithms are validated only in simulation; Woxbots demonstrated real-time pattern formation and path planning with physical robots, revealing challenges (sensor noise, communication latency, actuator imprecision) invisible in simulation.

Methods in Action

  • Decentralised model predictive control (MPC): Each drone solves a local optimisation problem over a short prediction horizon, incorporating predicted trajectories of nearby agents. Sarathi & Kumar (2025) compared multiple guidance algorithms (PID, MPC, DRL, and their combinations with RRT*) for decentralised drone swarms, finding that integrating MPC for path planning with DRL for decision-making provided effective coordination, including under wind disturbance conditions.
  • Leader-follower architectures: Lu et al. (2025) used a leader-follower approach where designated leader drones plan primary paths while followers maintain formation and perform cooperative obstacle avoidance. This reduces computational load on follower drones but introduces vulnerability to leader failure.
  • Swarm intelligence: Bio-inspired algorithms (ant colony optimisation, PSO, artificial bee colony) encode coordination rules as simple agent-level behaviours that produce emergent group-level intelligence. The appeal is scalability โ€” adding more drones does not increase per-drone computational load.
  • Reinforcement learning: DRL agents trained in simulation can learn collision-avoidance and task allocation policies that transfer to real drones, though the sim-to-real gap remains significant.

Key Claims & Evidence

<
ClaimEvidenceVerdict
Game-theoretic coordination achieves Nash equilibrium pathsHybrid game theory + PSO produces stable, collision-free trajectories (Qiu et al. 2025)Supported in simulation; scalability to large swarms (>100 drones) untested
Joint task trajectory optimisation outperforms sequentialSynchronous approach reduces mission time vs. sequential allocation-then-planning (Alqudsi 2024)Supported; computational cost is higher
Simulation results transfer to real-world swarm behaviourWoxbots platform reveals sim-to-real gap in sensor noise and latency (Ganduri et al. 2025)Partially; gap is significant and requires explicit domain adaptation
Leader-follower reduces per-drone computationFollowers execute simpler algorithms than leaders (Lu et al. 2025)Confirmed; trade-off is robustness to leader failure

Open Questions

  • Communication failure: How should swarm algorithms degrade gracefully when inter-drone communication is jammed, spoofed, or simply out of range? Most algorithms assume reliable (if limited) communication.
  • Heterogeneous swarms: Real missions may combine drones with different capabilities (high-altitude surveillance, low-altitude manipulation, ground vehicles). How should task allocation account for heterogeneity?
  • Adversarial environments: In defence applications, swarms must coordinate while being actively targeted. Can swarm algorithms maintain mission effectiveness under attrition?
  • Regulatory integration: As swarm sizes scale from 10 to 1,000+ drones, how will air traffic management systems accommodate them? Current drone regulations are designed for individual operators, not autonomous swarms.
  • What This Means for Your Research

    For robotics engineers, the papers reviewed here highlight that the gap between simulation and physical deployment remains the field's biggest challenge. Woxbots' full-stack approach โ€” proving algorithms on real hardware โ€” should become the standard for credible swarm robotics publications. For algorithm designers, the hybrid approaches (game theory + PSO, MPC + DRL) that combine complementary strengths are outperforming single-paradigm solutions. For application developers, the joint task trajectory optimisation of Alqudsi demonstrates that practical swarm missions require co-design of planning and allocation, not sequential optimisation.

    Referenced Papers

    • [1] Qiu, H. et al. (2025). Multi-robot Collaborative 3D Path Planning Based On Game Theory and PSO Hybrid Method. J. Supercomputing. DOI: 10.1007/s11227-025-06960-1
    • [2] Alqudsi, Y. (2024). Synchronous Task Allocation and Trajectory Optimization for Autonomous Drone Swarm. IEEE ICETI. DOI: 10.1109/ICETI63946.2024.10777195
    • [3] Ganduri, K.V. et al. (2025). Woxbots: A Low-Cost Swarm Robotics Platform for Real-Time Pattern Formation and Path Planning. J. Field Robotics. DOI: 10.1002/rob.70078
    • [4] Lu, S. et al. (2025). Path Planning and Cooperative Obstacle Avoidance for Swarm Robots Based on Leader-Follower Approach. IEEE METMS. DOI: 10.1109/METMS65303.2025.11047623
    • [5] Sarathi, A.P. & Kumar, G.K.L. (2025). Multi-Vehicle Guidance for Decentralised Swarm Using Adaptive MPC, DRL, and RRT. IEEE ETAAV*. DOI: 10.1109/ETAAV66793.2025.11213270

    References (5)

    Qiu, H., Yu, W., Zhang, G., Xia, X., & Yao, K. (2025). Multi-robot Collaborative 3D Path Planning Based On Game Theory and Particle Swarm Optimization Hybrid Method. The Journal of Supercomputing, 81(3).
    Alqudsi, Y. (2024). Synchronous Task Allocation and Trajectory Optimization for Autonomous Drone Swarm. 2024 1st International Conference on Emerging Technologies for Dependable Internet of Things (ICETI), 1-8.
    Ganduri, K. V., Pathri, B. P., Pammi, S. V. N., & Swami Sairam, P. (2026). Woxbots: A Lowโ€Cost Swarm Robotics Platform for Realโ€Time Pattern Formation and Path Planning. Journal of Field Robotics, 43(2), 949-968.
    Sheng Lu, Shuai Wang, Shuai Huang et al.. Path Planning and Cooperative Obstacle Avoidance Method for Swarm Robots Based on the Leader-Follower Approach.
    AP, S., & L, G. K. K. (2025). Multi-Vehicle Guidance for Decentralised Swarm of Drone Using Adaptive MPC, DRL, and RRT for Optimal Path Planning and Obstacle Avoidance. 2025 International Conference on Emerging Technology in Autonomous Aerial Vehicles (ETAAV)*, 1-6.

    Explore this topic deeper

    Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

    Click to remove unwanted keywords

    Search 8 keywords โ†’