This repository is a simple OpenAI Gym interface for Microsoft AirSim (https://github.com/microsoft/AirSim).

Install to your Python3 environment with

git clone https://github.com/TDYbrownrc/AirGym.git

cd air_gym

pip install .

This interface supports 2 drone control types: discrete positional control and continuous velocity control. These are initialization arguments passed into the OpenAI gym initialization script.

Create a gym environment like this:

import gym

import air_gym

env = gym.make('air_gym:airsim-drone-v0', ip_address = 'IP_ADDRESS_OF_AIRSIM_INSTANCE', control_type = {'discrete', 'continuous'}, step_length = DISTANCE_OF_DISCRETE_MOTION_OR_LENGTH_OF_TIME_FOR_VELOCITY_MOTION, image_shape = (w, h, 3), goal = [GOAL_POINT])