erickTornero / gym-vrep

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

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

Environments

1. Mobile robot navigation - the mobile robot contains five proximity sensors, IMU and two DC motors. The task of this agent is to navigate from position A to position B. Four variants of this problem were implemented:

Environment	Description
MobileRobotIdealNavigation	Position is obtained from simulation engine. Collision is detected with proximity sensor
MobileRobotVisualIdealNavigation	Position is obtained from simulation engine. Collision is detected with camera sensor
MobileRobotOdometryNavigation	Position is obtained from encoders ticks. Collision is detected with proximity sensor
MobileRobotGyrodometryNavigation	Position is obtained from encoders ticks and gyroscope. Collision is detected with proximity sensor

Action space are desired motor angular velocities in rad/s. They are limited to (0, 10.0)[rad/s].

Environment state space description:

Classic env:	Visual env:
distances from 5 proximity sensors	distances from 5 proximity sensors
polar coordinates	polar coordinates
linear and angular velocities	image from camera sensor
	linear and angular velocities

Environment reward:

Where

- linear velocity of the mobile robot

- heading angle of the mobile robot

- vector od distances read from proximity sensors

Installation

Requirements:

Basic requirements:

• V-REP 3.5.0
• Python 3.6+
• Ubuntu 16.04 / Arch Linux
• OpenAI gym

V-REP

chmod +x ./install_vrep.sh
./install_vrep.sh

Python

git clone https://github.com/Souphis/gym-vrep.git
cd gym-vrep
pip install -e .