Multi-Robot Motion Planning and Collision Avoidance

Multi-robot motion planning and collision avoidance address the challenge of guiding teams of autonomous agents through shared environments without interference or collision. The field spans centralised frameworks, which optimise trajectories jointly but struggle with computational scalability, and decentralised schemes, which distribute computation across individual units at the cost of global optimality guarantees. Key hurdles include nonconvex collision constraints, real-time computational demands, communication limitations and dynamic feasibility under disturbance. Recent work has combined model predictive control, convex optimisation and learning-based strategies to balance safety, efficiency and scalability. Techniques such as receding-horizon trajectory generation, kinematic separation, control barrier functions and event-triggered communication have emerged to ensure robust, on-the-fly coordination. Advances in parallel computing, notably GPU-accelerated solvers and decentralised consensus algorithms, enable fast replanning for tens or even hundreds of robots. Applications range from aerial drone swarms navigating urban canyons to autonomous warehouse vehicles operating in cluttered aisles. Interdisciplinary research now integrates geometric partitioning, priority-based corridors and adaptive communication protocols, driving the field towards deployment in complex, unstructured settings.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Real-time, decentralized trajectory planning using linear spatial separations (RLSS) has demonstrated collision avoidance in static, maze-like environments without any inter-robot communication. By formulating each robot’s instantaneous trajectory as a convex quadratic optimisation problem and enforcing linear spatial separators, the method guarantees kinematic feasibility and prevents both collisions and deadlocks in simulation and physical trials. A distributed model predictive control algorithm based on the alternating direction method of multipliers has been developed for quadrotor swarms. It leverages the differential flatness of rotary-wing dynamics and a relaxed discrete-time control barrier function to ensure safety, while event-triggered peer-to-peer communication minimises overhead. Comparative studies show improved real-time performance and smooth, collision-free paths even in large swarms. Another approach employs Voronoi partitioning to derive local collision constraints for receding-horizon multi-drone replanning. By embedding orientation-aware safety corridors expressed as Bézier curves, the method ensures continuous collision avoidance without discretisation. Experiments with up to one hundred drones highlight a higher success rate compared to classical Voronoi-based planners and demonstrate scalability in dense workspaces.

Multi-Robot Motion Planning and Collision Avoidance publication trend

The graph below shows the total number of articles in multi-robot motion planning and collision avoidance across all publications each year (not limited to Nature Index journals).

Technical terms

Trajectory planning: Computation of time-parameterised paths that satisfy a robot’s kinematic and dynamic constraints while avoiding obstacles.

Decentralised algorithm: A planning scheme in which each robot computes its own trajectory using local information rather than relying on a central controller.

Model predictive control (MPC): A receding-horizon approach that solves an optimisation problem at each time step to determine control inputs over a finite future horizon.

Convex quadratic optimisation: A class of mathematical programmes in which the objective function is quadratic and all constraints are convex, allowing for efficient, reliable solutions.

Control barrier function (CBF): A safety constraint imposed on control inputs to maintain system states within a designated safe set.

Voronoi partitioning: Division of space into regions based on proximity to a set of points, used to derive local collision-avoidance constraints.

References

1. RLSS: real-time, decentralized, cooperative, networkless multi-robot trajectory planning using linear spatial separations. Autonomous Robots (2023).
2. Cooperative Safe Trajectory Planning for Quadrotor Swarms. Sensors (2024).
3. GPU Accelerated Convex Approximations for Fast Multi-Agent Trajectory Optimization. IEEE Robotics and Automation Letters (2021).
4. A Distributed Algorithm for Real-Time Multi-Drone Collision-Free Trajectory Replanning. Sensors (2022).
5. Fast Joint Multi-Robot Trajectory Optimization by GPU Accelerated Batch Solution of Distributed Sub-Problems. Frontiers in Robotics and AI (2022).
6. Multi-Robot Robust Motion Planning based on Model Predictive Priority Contouring Control with Double-Layer Corridors. Applied Sciences (2022).
7. Learning scalable and efficient communication policies for multi-robot collision avoidance. Autonomous Robots (2023).