Advances in Multi-Agent Coordination and Control

The field of multi-agent systems is rapidly advancing, with a focus on developing innovative methods for coordination and control. Recent research has explored the integration of reinforcement learning, model predictive control, and distributed optimization to improve the safety, efficiency, and robustness of multi-agent systems. Notably, the use of dynamic constraints, homotopy-aware planning, and probabilistic safety constraints has shown promise in addressing complex challenges such as collision avoidance, deadlock prevention, and uncertain agent behavior. Furthermore, the development of novel frameworks and algorithms, such as decentralized uncertainty-aware collision avoidance and multi-shot ASP-based methods, has enabled more effective and adaptive control of multi-agent systems. Some noteworthy papers in this area include: ReCoDe, which introduces a reinforcement learning-based framework for dynamic constraint design, and Homotopy-aware Multi-agent Navigation, which proposes a novel distributed trajectory planning framework. Additionally, papers such as Silent Self-Stabilising Leader Election and A Truthful Mechanism Design have made significant contributions to the development of self-stabilizing algorithms and truthful mechanisms for distributed optimization. Overall, these advances have the potential to significantly impact the development of more efficient, safe, and robust multi-agent systems in a variety of applications.

Sources

ReCoDe: Reinforcement Learning-based Dynamic Constraint Design for Multi-Agent Coordination

Reinforcement learning in pursuit-evasion differential game: safety, stability and robustness

Homotopy-aware Multi-agent Navigation via Distributed Model Predictive Control

Silent Self-Stabilising Leader Election in Programmable Matter Systems with Holes

A Truthful Mechanism Design for Distributed Optimisation Algorithms in Networks with Self-interested Agents

Decentralized Uncertainty-Aware Multi-Agent Collision Avoidance With Model Predictive Path Integral

A General Framework for Dynamic MAPF using Multi-Shot ASP and Tunnels

Deep Neuro-Adaptive Sliding Mode Controller for Higher-Order Heterogeneous Nonlinear Multi-Agent Teams with Leader

Multi-UAV Deployment in Obstacle-Cluttered Environments with LOS Connectivity

Multi-Agent Path Finding Among Dynamic Uncontrollable Agents with Statistical Safety Guarantees

Built with on top of