SHARPIE website

Shared Human-AI Reinforcement Learning Platform for Interactive Experiments

Check out our paper Check out our code

Simple tag

Simple tag experiment, this requires 2 simultaneous users in the same room. The adversary user needs to chase the agent user. There is no learning involved in this experiment.

Simple spread

Simple spread experiment. Collaboration task: each agent should cover one target and do it as fast as possible. There is no learning involved in this experiment.

Mountain car

Mountain car experiment. Use the velocity of the car to get it to the flag as fast as possible. There is no learning involved in this experiment.

TAMER

In this experiment, the user provides a reward feedback to help the agent learn an optimal policy.