Decentralized shielding improves safety of multi robot systems
Decentralized Safe Multi-Agent Reinforcement Learning via Predictive Shielding
Artificial IntelligenceMultiagent SystemsRobotics
Summary
Many robots work together in shared spaces but often don’t know much about each other beforehand, which makes it hard for them to avoid accidents especially if the environment changes. The authors propose a method that helps each robot predict and avoid future dangers on its own, rather than reacting only when problems arise or relying on a central controller. This approach also includes a way for robots to resolve conflicts without needing to communicate, which is useful when they behave similarly and might get stuck. Overall, this method helps robots safely adapt to new situations while working alongside others.
What this means in practice
- •For robotics engineers: Build multi-robot systems that adapt safely to unexpected obstacles without central control or communication.
- •For automation system designers: Design decentralized control protocols that prevent deadlock situations among symmetric autonomous agents in dynamic environments.
Authors
Yacine El Yamani, Hanna Krasowski, Elena Vanneaux
Abstract
Environments are increasingly populated by multiple robots performing independent tasks with limited prior knowledge of each other. Deploying such multi-agent systems presents significant challenges. Specifically, shifts in deployment states compared to training data can lead to poor policy performance and compromised safety. While safety shields exist to mitigate these risks, they are typically reactive, which degrades performance near unseen obstacles,and centralized, limiting their scalability. To address this, we propose a decentralized framework that integrates predictive shielding with model-based finite horizon Q-learning. This approach allows agents to safely adapt their pre-trained policies during deployment. Furthermore, to mitigate livelocks in symmetric scenarios, we introduce a communication- free protocol for conflict resolution