Effective Methods for Reinforcement Learning in Large Multi-Agent Domains

نویسندگان

  • Martin A. Riedmiller
  • Daniel Withopf
چکیده

Robotic soccer requires the ability of individually acting agents to cooperate. The simulation league of RoboCup therefore offers an ideal testbed for evaluating multiagent methods. In this paper we discuss how Reinforcement Learning (RL) methods can be succesfully applied within the scenario of learning to cooperatively score a goal. Due to the complexity of the task, enhanced methods of learning have to be applied. We discuss several approaches from literature and also present an own approach. All approaches are evaluated on a discretized version of robotic soccer, which we call gridworld soccer. Zusammenfassung Eine wichtige Voraussetzung für erfolgreiches (Roboter-)Fußballspiel ist die Fähigkeit der einzelnen Agenten zur Kooperation. Damit stellt die RoboCup-Simulationsliga ein ideales Testszenario zur Evaluation von Multi-Agenten Lernverfahren dar. Dieser Artikel zeigt Möglichkeiten auf, wie Reinforcement Lernen erfolgreich eingesetzt werden kann, um den Agenten das gemeinsame Tore schießen beizubringen. Die hohe Komplexität dieser Aufgabe erfordert den Einsatz leistungsfähiger Lernmethoden. Im Artikel werden verschiedene aus der Literatur bekannte Ansätze diskutiert und ein neuer Algorithmus vorgestellt. Alle diese Verfahren werden in einer vereinfachten, diskretisierten Version des Roboterfußballs evaluiert.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Transfer Learning in Multi-Agent Reinforcement Learning Domains

Transfer learning refers to the process of reusing knowledge from past tasks in order to speed up the learning procedure in new tasks. In reinforcement learning, where agents often require a considerable amount of training, transfer learning comprises a suitable solution for speeding up learning. Transfer learning methods have primarily been applied in single-agent reinforcement learning algori...

متن کامل

Parallel Fault Tolerant Multi-Agent Reinforcement Learning

Reinforcement learning is a powerful tool for training an agent in a sequential decision based environment and has been successful in many simulated [6] as well as practical [5] domains. In this paper we investigate methods of strengthening the rate of convergence of a single agent RL learner by sharing observations with other independent agents. In contrast to multi-agent reinforcement methods...

متن کامل

A Modular On-line Profit Sharing Approach in Multiagent Domains

How to coordinate the behaviors of the agents through learning is a challenging problem within multi-agent domains. Because of its complexity, recent work has focused on how coordinated strategies can be learned. Here we are interested in using reinforcement learning techniques to learn the coordinated actions of a group of agents, without requiring explicit communication among them. However, t...

متن کامل

Utilizing Generalized Learning Automata for Finding Optimal Policies in MMDPs

Multi agent Markov decision processes (MMDPs), as the generalization of Markov decision processes to the multi agent case, have long been used for modeling multi agent system and are used as a suitable framework for Multi agent Reinforcement Learning. In this paper, a generalized learning automata based algorithm for finding optimal policies in MMDP is proposed. In the proposed algorithm, MMDP ...

متن کامل

Locking in Returns: Speeding Up Q-Learning by Scaling

One problem common to many reinforcement learning algorithms is their need for large amounts of training, resulting in a variety of methods for speeding up these algorithms. We propose a novel method that is remarkable both for its simplicity and its utility in speeding up Q-learning. It operates by scaling the values in the Q-table after limited, typically small, amounts of learning. Empirical...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • it - Information Technology

دوره 47  شماره 

صفحات  -

تاریخ انتشار 2005