WMSCI 2008: 12TH WORLD MULTI-CONFERENCE ON SYSTEMICS, CYBERNETICS AND INFORMATICS, VOL I, PROCEEDINGS, pp.109-113
World Multi-Conference on Systemics, Cybernetics and Informatics (Orlando, FL, 06/29/2008–07/02/2008)
01/01/2008
Web of Science ID: WOS:000263667100020
Metrics
38 Record Views
Abstract
The G-Bots project focuses oil the development of an intelligent agent-based architecture for a real-time simulated environment. The agent architecture interacts with a Newtonian physics-based environment, given a set of sensors and actuators. The environment consists of small circular planets and other objects in two-dimensional space. In each game episode, each intelligent agent must try to eliminate the others by causing objects, such as rocks, to hit them, while minimizing hits to itself. After receiving a certain number of hits, an agent is eliminated from the episode. An agent can perform actions such as driving on a planet's surface, picking up an object, throwing an object, and launching itself at a given angle and velocity. The planets exert conflicting gravitational pulls that the agents must account for in their predictions of object motion. Artificial neural networks (ANNs) will be used to improve the agents' performance, including better aiming accuracy and trajectory prediction.