Adaptive Agents and Multi-Agents Systems 

About: Adaptive Agents and Multi-Agents Systems is an academic conference. The conference publishes majorly in the area(s): Multi-agent system & Reinforcement learning. Over the lifetime, 7058 publications have been published by the conference receiving 166678 citations.

Bayesian Reinforcement Learning in Factored POMDPs

Abstract: Model-based Bayesian Reinforcement Learning (BRL) provides a principled solution to dealing with the exploration-exploitation trade-off, but such methods typically assume a fully observable environments. The few Bayesian RL methods that are applicable in partially observable domains, such as the Bayes-Adaptive POMDP (BA-POMDP), scale poorly. To address this issue, we introduce the Factored BA-POMDP model (FBA-POMDP), a framework that is able to learn a compact model of the dynamics by exploiting the underlying structure of a POMDP. The FBA-POMDP framework casts the problem as a planning task, for which we adapt the Monte-Carlo Tree Search planning algorithm and develop a belief tracking method to approximate the joint posterior over the state and model variables. Our empirical results show that this method outperforms a number of BRL baselines and is able to learn efficiently when the factorization is known, as well as learn both the factorization and the model parameters simultaneously.

RoboCup: The Robot World Cup Initiative

Abstract: The Robot World Cup Initiative (R, oboCup) is attempt to foster AI and intelligent rohoties research by providing a standard problem where wide range of technologies especially concerning multi-agent research (:an be integrated and examined. The first RoboCup competition is to be, heht at. IJCAI-97, Nagoya. In order for a robot team to actually perform a soccer game. various technologies must I)e incorl)orated including: design principles of autononmus agents, multi-agent collaboration, strategy acquisition, real-time rea.~oning, robotics, and sensor-fllsion. Unlike AAAI robot competition, which is tuned for a single heavy-duty slow-moving robot. RoboCup is a task for a team of multiple f‘ast-moving robots under a dynamic environmen(. Although RoboCnp’s final target is a worhl cup with real robots, RoboCup offers a soft.ware platform for reseaxch on the software aspects of RoboCup. This paper describes teclini(’M challenges involw~d in RoboCup, rules, and simulation environment.

Cooperative Multi-agent Control Using Deep Reinforcement Learning

Abstract: This work considers the problem of learning cooperative policies in complex, partially observable domains without explicit communication. We extend three classes of single-agent deep reinforcement learning algorithms based on policy gradient, temporal-difference error, and actor-critic methods to cooperative multi-agent systems. To effectively scale these algorithms beyond a trivial number of agents, we combine them with a multi-agent variant of curriculum learning. The algorithms are benchmarked on a suite of cooperative control tasks, including tasks with discrete and continuous actions, as well as tasks with dozens of cooperating agents. We report the performance of the algorithms using different neural architectures, training procedures, and reward structures. We show that policy gradient methods tend to outperform both temporal-difference and actor-critic methods and that curriculum learning is vital to scaling reinforcement learning algorithms in complex multi-agent domains.

Frontier-based exploration using multiple robots

Abstract: 1 ABSTRACT Frontier-based exploration directs mobile robots to regions on the boundary between unexplored space and space that is known to be open Previously, we have demonstrated that frontier-based exploration can be used to map indoor environments where walls and obstacles may be in arbitrary orientations In this paper, we show how frontier-based exploration can be extended to multiple robots In our approach, robots share perceptual information, but maintain separate global maps, and make independent decisions about where to explore This approach enables robots to make use of information from other robots to explore more effectively, but it also allows the team to be robust to the loss of individual robots We have implemented our multirobot exploration system on real robots, and we demonstrate that they can explore and map office environments as a team

Reputation and social network analysis in multi-agent systems

Abstract: The use of previous direct interactions is probably the best way to calculate a reputation but, unfortunately this information is not always available. This is especially true in large multi-agent systems where interaction is scarce. In this paper we present a reputation system that takes advantage, among other things, of social relations between agents to overcome this problem.