Reinforcement Learning: Developing a battle tank for the RoboCode battle simulator

Abstract

The battle simulator RoboCode is a competitive Java programming challenge, wherein one can program a competitive battle tank. By exploring the possibili- ties of using machine learning, especially reinforcement learning, a learning battle tank is designed. to design a learning battle tank. A short survey of reinforcement learning in general is made and an in depth analysis of the RoboCode environment is performed. Finally the design is implemented and tested.