Tech Stack
Python
Reinforcement Learning
ROS2
Gazebo
SLAM
Machine Learning
Description
Developed a comprehensive reinforcement learning environment for TurtleBot4 navigation and simultaneous localization and mapping (SLAM) tasks built on ROS 2 Jazzy and Gazebo Harmonic.
Implemented advanced obstacle tracking system with Model Predictive Path Integral (MPPI) controller and SMAC Hybrid planner for ultra-conservative navigation and robust obstacle avoidance.
Created Gymnasium-compliant API supporting multiple RL algorithms including PPO, SAC, and TD3 with dynamic goal assignment and completion logic for autonomous navigation training.
- Advanced RL environment with comprehensive TurtleBot4 navigation capabilities
- Built-in SLAM integration for autonomous mapping and localization with AMCL
- Dynamic obstacle detection and tracking system for complex scenarios
- Support for PPO, SAC, TD3 and other state-of-the-art RL algorithms
- Ultra-conservative safety margins with rigorous obstacle avoidance
- Gymnasium-compliant API for easy integration and fast prototyping
- Headless and visual simulation modes with optimized performance
Page Info
TurtleBot4-RL Navigation Demo
Live demonstration of TurtleBot4 autonomous navigation using reinforcement learning in Gazebo simulation environment.
TurtleBot4-RL Navigation Demo