Habitat is a simulation platform for research in Embodied AI. The embodiment hypothesis is the idea that “intelligence emerges in the interaction of an agent with an environment and as a result of sensorimotor activity”. WordCraft -Test core capabilities Commonsense knowledge.The Unity Machine Learning Agents Toolkit (ML-Agents) – Create environments Curriculum learning Single-/Multi-agent Imitation learning.StarCraft II Learning Environment – Rich action and observation spaces Multi-agent Multi-task.Serpent.AI – Game Agent Framework – Turn ANY video game into the RL env.Screeps – Compete with others Sandbox MMO for programmers.RL Unplugged – Offline RL Imitation learning Datasets for the common benchmarks.RLCard – Classic card games Search and planning Single-/Multi-agent.Real-World Reinforcement Learning – Continuous control Physics-based simulation Adversarial examples.PyBullet Gymperium – Continuous control Physics-based simulation MuJoCo unpaid alternative.Procgen Benchmark – Evaluate generalization Procedurally-generated.OpenSpiel – Classic board games Search and planning Single-/Multi-agent.OpenAI Gym Retro – Classic video games RAM state as observations.OpenAI Gym – Continuous control Physics-based simulation Classic video games RAM state as observations. ![]() Multiagent emergence environments – Multi-agent Creating environments Emergence behavior.MineRL – Imitation learning Offline RL 3D navigation Puzzle-solving.Google Research Football – Multi-task Single-/Multi-agent Creating environments.DeepMind Psychlab – Require memory Evaluate generalization.DeepMind Memory Task Suite – Require memory Evaluate generalization.DeepMind Lab – 3D navigation Puzzle-solving. ![]() DeepMind Control Suite – Continuous control Physics-based simulation Creating environments.Behaviour Suite – Test core RL capabilities Fundamental research Evaluate generalization.AI Habitat – Virtual embodiment Photorealistic & efficient 3D simulator.Further down, I add a bit of description from each benchmark’s creator to show you what it’s for. The first part of this section is just a list, in alphabetical order, of all 23 benchmarks. If you don’t know what you’re interested in yet, then I suggest playing around with classic control environments in the OpenAI Gym, and reading SpinningUp in Deep RL.Įnough introduction, let’s check out the benchmarks! Benchmarks Harder environments include Robotics in the OpenAI Gym. PyBullet Gymperium is an unpaid alternative. On the other hand, if you’re more interested in algorithms specialized in continuous action spaces (DDPG, TD3, SAC, …), where the action input is, say, torque on the joints of a humanoid robot learning to walk, then you should look at the MuJoCo environments in the OpenAI Gym and DeepMind Control Suite. These include Pong, Breakout, Space Invaders, Seaquest, and more. If you’re interested in algorithms specialized in discrete action spaces (PPO, DQN, Rainbow, …), where the action input can be, for example, buttons on the ATARI 2600 game controller, then you should look at the Atari environments in the OpenAI Gym. I hope it will motivate you to keep doing good work, and inspire you to start your own project in something different than standard benchmarks! Rule of thumb Whatever your current level of knowledge, I recommend looking through the whole list. ![]() We’re going to explore 23 different benchmarks, so I guarantee you’ll find something interesting!īut first, we’ll do a short introduction to what you should be looking for if you’re just starting with RL. It’s where you run your algorithm to evaluate how good it is. The basis for RL research, or even playing with or learning RL, is the environment. In this post, I’ll share with you my library of environments that support training reinforcement learning (RL) agents.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |