Overview
SLMRME is an open-source Python and ROS implementation supporting research into safe, mapless multi-robot exploration. It uses deep reinforcement learning to train TurtleBot3 Waffle Pi robots to navigate and avoid obstacles without pre-built maps, with training and evaluation conducted in Gazebo simulation using PyTorch.
SLMRME details
Safe Learning for Multi-Robot Mapless Exploration (SLMRME) is an open-source robotics research project and the supplementary implementation for the associated IEEE Transactions on Vehicular Technology paper. The project investigates how deep reinforcement learning can enable mobile robots to explore unknown environments, navigate autonomously and avoid obstacles without relying on a pre-built map.
The software is developed in Python within the Robot Operating System framework. It uses Gazebo to provide simulated environments and PyTorch to develop and train the deep reinforcement learning models. The supplied ROS package includes launch files for starting the TurtleBot3 Waffle Pi simulation and running the learning process. LiDAR observations and environmental feedback are used to train the robot, with the simulated laser scanner configured to use 24 range measurements as the learning input.
SLMRME provides a reproducible starting point for studying mapless navigation in simulation. It can support further development, comparative experiments and eventual validation on physical robotic platforms.
Who it’s for
SLMRME is intended primarily for robotics researchers, engineers and developers working on:
- autonomous navigation
- robot learning
- safe reinforcement learning
- multi-robot exploration
Licence
Open Source
Release date
12 June 2025
Programming language
Python
Dependencies
- ROS Noetic
- Python 3.8
- PyTorch 1.10.0
- NumPy 1.24.4
- TurtleBot3, TurtleBot3 Messages, TurtleBot3 Simulations, TurtleBot3 Machine Learning
- ROS Laser Filters
Compatible platforms / systems
Ubuntu 20.04 with ROS Noetic