Learning by Demonstration of Coordinated Plans in Multiagent Systems

Abstract

One of the significant challenges in Multiagent Systems (MAS) is the creation of cooperative plans to deal with the different scenarios that present themselves in a dynamic, real-time environment composed of teams of mobile robots. This work involves capturing human knowledge to demonstrate how robot teams can better cooperate in solving the problem they must solve. The research used the environment RoboCup 3D Soccer Simulation (3DSSIM) and the collection of human demonstrations were carried out through a set of tools developed from adapting existing solutions in the RoboCup community using a crowdsourcing strategy. In addition, fuzzy clustering was used to gather human demonstrations (setplays) with the same semantic meaning, even with minor differences. With the data organized, this work used a reinforcement learning mechanism to learn a classification policy that allows agents to decide which group of setplays is best suited to each situation that presents itself in the environment. The results show the ability of the robot team to evolve, from learning the suggested setplays and their use in an appropriate way to the skills of each robot.