Home

Description

Reinforcement Learning

Reinforcement Stimulus is a stimulus that strengthens or weakens the behavior that produced it. Reinforcement Learning applies these aspects to machine learning. For an intro in Reinforcement Learning (RL) read more here.

Master Thesis Project - RLTankAttack

This 2-player Python game consists of two boats trying to submerge their opponent in water.
Modes: Human-vs-Human, Human-vs-Computer, Computer-vs-Computer. The purpose of this game is to study learning mechanisms and their use in new game quality assurance.

Game Description

The game (demo overview) consists of two opponent boats on a canal system, trying to destroy their opponent in water. Each player can manipulate either his/her forward/backward movement, or fill/drain a level in a water tank [http://en.wikipedia.org/wiki/Lock_%28water_transport%29 lock]. A lock tank is a water column that can be drained or filled with water, and the whole water tank on which the two boats can float, consists of a finite number of lock tanks i.e. the player manipulates two things: motion (forward or backward) and a finite number of valves (drain or fill). The number of valves is double the number of lock tanks. The precise rules for the game will be documented is this [http://code.google.com/p/rltankattack/wiki/Gameplay Gameplay wiki page] and will be subject to change.

Tools

Python	*Scipy	Pygame	RL-Glue
The base	The Numeric/Science Library	The gaming platform	The AI glue

You can view the current abstract application stack [http://code.google.com/p/rltankattack/wiki/AbstractApplicationStack here].

Powered By