Agent Strategy Generation by Rule Induction

This paper presents a study on a rule induction application for generating an agent strategy. It is a new approach in multi-agent systems, where reinforcement learning and evolutionary computation is broadly used for this purpose. Experimental results show that rule induction improves agent performance very quickly. What is more, rule-based knowledge representation has many advantages. It is comprehensive and clear. It allows for the examination of the learned knowledge by humans. Because of modularity of the knowledge, it also allows for the implementation of the knowledge exchange in a natural way -- only necessary set of rules can be sent. Rule induction is tested in two domains: Fish Banks game, in which agents run fishing companies and learn how to allocate ships, and Predator-Prey domain, in which predator agents learn how to capture preys. The proposed learning mechanism should be beneficial in all domains, in which agents can determine the results of their actions.

[1]  Jan D. Gehrke,et al.  Combining Rule Induction and Reinforcement Learning: An Agent-based Vehicle Routing , 2010, 2010 Ninth International Conference on Machine Learning and Applications.

[2]  Aiko M. Hormann,et al.  Programs for Machine Learning. Part I , 1962, Inf. Control..

[3]  C. Lee Giles,et al.  Learning Communication for Multi-agent Systems , 2002, WRAC.

[4]  Marek Kisiel-Dorohinicki,et al.  A Crisis Management Approach to Mission Survivability in Computational Multi-Agent Systems , 2010, Comput. Sci..

[5]  Bartlomiej Sniezynski,et al.  Resource Management in a Multi-agent System by Means of Reinforcement Learning and Supervised Rule Learning , 2007, International Conference on Computational Science.

[6]  Chris Watkins,et al.  Learning from delayed rewards , 1989 .

[7]  Jan D. Gehrke,et al.  Traffic Prediction for Agent Route Planning , 2008, ICCS.

[8]  Sean Luke,et al.  Cooperative Multi-Agent Learning: The State of the Art , 2005, Autonomous Agents and Multi-Agent Systems.

[9]  Sandip Sen,et al.  Evolving Beharioral Strategies in Predators and Prey , 1995, Adaption and Learning in Multi-Agent Systems.

[10]  Bartlomiej Sniezynski,et al.  Comparison of strategy learning methods in Farmer-Pest problem for various complexity environments without delays , 2013, J. Comput. Sci..

[11]  Bartosz Balis,et al.  Real-time Grid monitoring based on complex event processing , 2011, Future Gener. Comput. Syst..

[12]  J. Ross Quinlan,et al.  C4.5: Programs for Machine Learning , 1992 .

[13]  Sandip Sen,et al.  Learning in multiagent systems , 1999 .

[14]  Daniel Kudenko,et al.  Machine Learning and Inductive Logic Programming for Multi-agent Systems , 2001, EASSS.

[15]  Jacek Kitowski,et al.  Grid organizational memory - provision of a high-level Grid abstraction layer supported by ontology alignment , 2007, Future Gener. Comput. Syst..

[16]  Stéphane Airiau,et al.  Incorporating Learning in BDI Agents , 2008 .

[17]  Lin Padgham,et al.  Learning context conditions for BDI plan selection , 2010, AAMAS.

[18]  C Feliziani,et al.  The tragedy of commons , 2011 .

[19]  Krzysztof Cetnarowicz,et al.  Maintaining Functional Integrity in Multi-Agent Systems for Resource Allocation , 2010, Comput. Informatics.

[20]  Bartlomiej Sniezynski,et al.  Learning in a Multi-agent Approach to a Fish Bank Game , 2005, CEEMAS.

[21]  Ryszard S. Michalski,et al.  AQVAL/1 (AQ7) User's Guide and Program Description , 1975 .

[22]  R. Michalski.  Attributional Calculus: A Logic and Representation Language for Natural Induction , 2004 .

[23]  Bartlomiej Sniezynski,et al.  Agent Strategy Generation by Rule Induction in Predator-Prey Problem , 2009, ICCS.

[24]  G. Hardin,et al.  The Tragedy of the Commons , 1968, Green Planet Blues.

[25]  Allen Newell,et al.  Human Problem Solving. , 1973 .