IIIT Hyderabad Publications
Wheeled Robots playing Catch-Catch and Chain Catch Games
Author: Garima Agrawal
Report no: IIIT/TH/2016/10
Robots playing games human are adept in playing is a challenge. It is an area where the researcher has to deal with real world properties . Catch-Catch and Chain Catch are common playground “Tag” games where a single player as a Catcher chases other players to catch them while others try to escape. To extend the game further in form of Chain Catch game, Catcher and the just caught players form a chain and chase other players in that formation. We implement the idea of robotic agents playing Catch-Catch and Chain Catch as a Multi-Agent System (MAS), where each agent is an autonomous entity with capability to communicate with other agents and to use different strategies based on their roles (Catcher, Escapee, Chain). Our solution is handled both in physical (with actual robots) and simulated environment (simulating software agents). These are fast games that require team of robotic agents to cooperate among themselves to compete with other group of agents through quick decision making. Developing such robotic agents and designing their strategies is a challenging research problem. Idea of “Catch” in our game can be seen as a variant of pursuit evasion problem, but game rules and constraints on field space make our problem more complex and hence solutions to pursuit domain are not completely suitable for Catch-Catch. Also, the extension of game into Chain Catch makes it a combination of two challenges- pursuit domain and robotic chain formation. Thus our Chain Catch game is a novel and challenging problem. In this thesis, we develop a front end Catch-Catch and Chain Catch simulator that allows us to incor- porate game rules, design strategies and simulate the game play. We use the concept of cost heuristics to design strategies for each of Escapee, Catcher and Chain. Formation plays a key role among Es-capees to evade a Catch. We propose Maximize distance, K circle, K circle with rotation and Sliding slope strategies for the Escapees. These strategies are implemented, tested and evaluated against pursuit domain strategy and random movement. Two techniques- Tagging method and Variance heuristics are developed for chain formation and its movement in Chain Catch game scenario. We also built and used our own production quality robots to implement the game play in physical environment and study the behavior of our game strategies in real robot scenario. A number of empirical experiments are per-formed to evaluate, compare and form the best strategy for each of chain, Escapee and Catcher. Our results show that Sliding slope strategy is the best strategy for Escapees whereas Tagging method is the best method for chain’s movement in Chain Catch. Our robot implementation in different scenarios shows that game strategies work as expected and a complete chain formation takes place successfully in each game.
Full thesis: pdf
Centre for Data Engineering
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved.