Decentralized partially observable Markov decision processes (Dec-POMDPs) offer a powerful modeling technique for realistic multi-agent coordination problems under uncertainty. Prevalent solution techniques are centralized and assume prior knowledge of the model. We propose distributed reinforcement learning approach, where agents take turns to learn best responses each other’s policies. This p...

This paper reports on experiments involving simulated robot-like agents. Motivated by social insects, we investigated two approaches to learning distributed controllers for an object-clustering task: genetic algorithms and reinforcement learning. In the case of reinforcement learning, a new learning algorithm was developed in an attempt to remove the need for a global fitness observer. Results ...

Recent advances in sensing, actuation, and computation have opened the door to multi-robot systems consisting of hundreds/thousands robots, with promising applications automated manufacturing, disaster relief, harvesting, last-mile delivery, port/airport operations, or search rescue. The community has leveraged model-free multi-agent reinforcement learning (MARL) devise efficient, scalable cont...

With connections to bounded rational game theory, information theory and statistical mechanics, Product Distribution (PD) theory provides a new framework for performing distributed optimization. Furthermore, PD theory extends and formalizes Collective Intelligence, thus connectingt distributed optimization to distributed Reinforcement Learning (RL). This paper provides an overview of PD theory ...

Although behaviour-based robotics has been successfully used to develop autonomous mobile robots up to a certain point, further progress may require the integration of a learning model into the behaviour-based framework. Reinforcement learning is a natural candidate for this because it seems well suited to the problems faced by autonomous agents. However, previous attempts to use reinforcement ...