Reinforcement learning toolkit written in Java, with support for multi-agent environments.
Current features
multi-agent support standard RL algorithms: TD, Sarsa(lambda), Q(0), actor-critic(lambda). configurable grid world configurable, physically realistic continuous world (based on a modified version of the Phys2D physics engine: http://www.cokeandcode.com/phys2d/) episodic and continuing tasks plug-in custom reward functions plug-in custom statistics most aspects of the simulations can be configured through XML files (still a work in progress): learning algorithm, state composition (through sensors), lay-out of the world, obstacles, goal positions, rewards, etc. easy to use with RL-Glue (http://glue.rl-community.org/) (note: only for single-agent scenarios since RL-Glue does not support multiple agents!) ---
This project is supported by the YourKit Java profiler:
YourKit is kindly supporting open source projects with its full-featured Java Profiler. YourKit, LLC is the creator of innovative and intelligent tools for profiling Java and .NET applications. Take a look at YourKit's leading software products: YourKit Java Profiler and YourKit .NET Profiler.
