Open-AI Gym extension for robotics based on V-REP

Environments

1. Mobile robot navigation - the mobile robot contains ultrasonic sensors, IMU and two DC motors. Number of ultrasonic sensors depends on task type. The task of this agent is to navigate from position A to position B. There are implemented two variants presented below.

The environment can be customized with four parameters:

• navigation_type - Selects the type of the navigation algorithms from the supported ones (ideal, odometry, gyrodometry). Default is ideal.
• enable_vision - Whether to include visual feedback to the observation space. Default is False.
• goal_threshold - Defines the threshold value for reaching the goal. Default value is set to 0.05.
• collision_threshold - Defines the threshold distance for the proximity sensors, when the collision occurs. Default value is set to 0.05.

Action space are desired motor angular velocities in rad/s. They are limited to (0, 15.0) rad/s.

Environment observation space description:

• distances from n proximity sensors
• polar coordinates
• linear and angular velocities
• (optional) image from camera sensor

Getting started

Here are provided basic requirements for the project:

• CoppeliaSim 4.1.0+
• Python 3.6+
• Ubuntu 20.04
• OpenAI gym
• PyRep

Python

To install this package run following commands:

git clone https://github.com/Souphis/gym-vrep.git
cd gym-vrep
python3 setup.py install

Basic usage

To run the environment with default configuration just add follwing code to your project:

import gym
import gym_coppelia_sim

env = gym.make("RoomNavigation-v0") # or env = gym.make("DynamicRoomNavigation-v0")

Below is the example how to run environment in the headless mode:

import gym
import gym_coppelia_sim

env = gym.make("RoomNavigation-v0", headless_mode=True)

Changing the navigation algorithm can be done with following example:

import gym
import gym_coppelia_sim

env = gym.make("RoomNavigation-v0", navigation_algos="odometry")