Ipseity: An Open-Source Platform for Synthesizing and Validating Artificial Cognitive Systems in MAS
Affiliation auteurs | !!!! Error affiliation !!!! |
Titre | Ipseity: An Open-Source Platform for Synthesizing and Validating Artificial Cognitive Systems in MAS |
Type de publication | Conference Paper |
Year of Publication | 2014 |
Auteurs | Lauri F, Koukam A |
Conference Name | AAMAS'14: PROCEEDINGS OF THE 2014 INTERNATIONAL CONFERENCE ON AUTONOMOUS AGENTS & MULTIAGENT SYSTEMS |
Publisher | Assoc Comp Machinery; Int Fdn Autonomous Agents & Multi Agent Syst; Assoc Comp Machinery Special Interest Grp Artificial Intelligence; Air Force Res Lab; Microsoft Res; THALES; CEA Tech; Lab Informatique Paris 6; Univ S Carolina; Univ Fed Rio Grande Sul; |
Conference Location | 1515 BROADWAY, NEW YORK, NY 10036-9998 USA |
ISBN Number | 978-1-4503-2738-1 |
Mots-clés | artificial intelligence, cognitive systems, multi-agent system, open-source, Reinforcement Learning, software platform |
Résumé | This article presents an overview of Ipseity, an open-source platform developed in C++ with the Qt framework. The current version of the platform includes a set of plugins implementing single-agent and multi-agent environments, hard-coded controllers based on Artificial Intelligence (AI) techniques, classical Reinforcement Learning (RL) techniques like Q-Learning, Sarsa, Epsilon-Greedy combined with some linear function approximators, as well as a Machine Learning (ML) technique for Apprenticeship Learning (AL). Its architecture allows users to execute standard AI approaches as well as model-based, model-free, offline, online, standard and approximated RL algorithms. Ipseity is targeted at a broad range of users interested in AI in general, including industrial practitioners, as well as ML researchers, students and teachers. It is regularly used as a course support in Artificial Intelligence and it has been used successfully to manage power flows in simulated microgrids using multi-agent reinforcement learning. |