There are many reinforcement learning platforms available for experimenting with reinforcement learning problems in simulation. We can build and define the environment and the agent and emulate certain behaviors for the agent to see how the agent learns to solve the problem. Some of the available learning platforms are as follows:

• OpenAI Gym: This is one of the most commonly used toolkits for experimenting with reinforcement learning algorithms. It is a simple and easy to use tool with prebuilt environments and agents. It is open source and supports frameworks such as TensorFlow, Theano, and Keras. There is support for ROS using the OpenAI Gym package, which we will see in the upcoming chapters. More information can be found here: https://github.com/openai/gym.
• OpenAI Universe: This is an extension of OpenAI Gym. This tool provides more realistic and complex environments, like those of PC games. More information can be found here: https://github.com/openai/universe.
• DeepMind Lab: This is another cool tool that provides realistic and science fiction style environments that are customizable. More information can be found here: https://github.com/deepmind/lab.
• ViZDoom: This is a Doom—a PC game-based tool that supports multi-agent support, but limited to only one environment (the Doom game environment). More information can be found here: https://github.com/mwydmuch/ViZDoom.
• Project Malmo: This is a sophisticated experimentation platform built on top of the Minecraft game by Microsoft that supports research in reinforcement learning and AI. More information can be found here: https://github.com/Microsoft/malmo.

Now, let's look at reinforcement learning in robotics.