Online Change-point Detection for Cooperative Multi-Agent Reinforcement Learning
Researchers propose Patterns of Past Rewards (PPR), a lightweight reward-based detector that flags environment shifts in cooperative multi-agent reinforcement learning training.
The paper introduces Patterns of Past Rewards (PPR), an algorithm-agnostic detector that smooths cooperative agents' return streams and applies statistical drift testing to flag environment or task changes. Evaluation in a custom Speaker-Listener environment built on the Multi-Agent Particle Environment under two non-stationarity scenarios shows PPR balances detection speed against alarm stability. It avoids the repeated alarms of a smoothed-return baseline and the missed shifts of raw-return detection, enabling MARL systems to reliably identify major changes during training.
- PPR is algorithm-agnostic and uses only reward-derived signals to detect non-stationarity.
- Evaluated in a Multi-Agent Particle Environment Speaker-Listener setup under two controlled non-stationarity scenarios.
- PPR limits redundant alarms while still detecting controlled shifts, balancing speed and stability.
- Products
- Multi-Agent Particle Environment
From the original article
1 min read at arxiv.orgCooperative multi-agent reinforcement learning (MARL) systems rely on past experience for learning coordinated behaviour, but this experience may become unreliable if the environment or task objective changes during training. In such cases, agents first need a way to recognize that the situation has changed before deciding how to adapt.
By Fatemeh Saberi Khomami · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.