HippoGym

Human Input Parsing Platform for Openai Gym

Purpose:

HippoGym is a python library for simplifying human-ai interaction research over the web. The library provides a communicator that opens a websocket to pass environment information and recieve commands and actions from browser-based front-end.

The library is designed to be customizable for diverse research applications. The framework is naive in that it makes no assumptions about how human or agents input will be used, it simply provides the mechanism to pass along and record this information.

Installation

pip install hippogym

Examples

Lunar Lander with human agent

Copy and run the lunar lander example.

You should see:

Then connect a client to the websocket using the host and port, you can use the irll frontend: https://beta.irll.net/?server=ws://localhost:5000

You can now play with either the command pannel or the keyboard:

Minigrid with human agent

Copy and run the minigrid example.

You should see:

Then connect a client to the websocket using the host and port, you can use the irll frontend: https://beta.irll.net/?server=ws://localhost:5000

You can now play with either the command pannel or the keyboard: