Zhefan-Xu/NavRL

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1587 stars · 168 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

NavRL is a reinforcement learning framework for safe robot and UAV navigation in dynamic environments, published at IEEE RA-L 2025. It includes training in NVIDIA Isaac Sim, pretrained models, and ROS1/ROS2 deployment packages for velocity-controlled robots.

Use cases

• train a UAV navigation policy in Isaac Sim with reinforcement learning
• deploy a learned collision-avoidance controller on a ROS1 or ROS2 robot
• run quick demos of navigation among dynamic obstacles and multi-robot scenarios
• extend safe navigation to any velocity-controlled robot platform
• reproduce results from the NavRL RA-L 2025 paper

When to choose

• you need learned, reactive collision avoidance in dynamic environments
• your robot uses velocity-based control and runs ROS1 Noetic or ROS2 Humble
• you want GPU-accelerated training in NVIDIA Isaac Sim with a pretrained model to start from

When to avoid

• you need static-map global path planning rather than local reactive navigation
• your platform is not Linux/ROS-based or does not accept velocity commands
• you need a production-grade, commercially supported navigation stack

Facets

library · maturity active

reinforcement-learning simulation robotics robotics reinforcement-learning autonomous-vehicles simulation python cpp uav-navigation collision-avoidance isaac-sim ros1-noetic ros2-humble dynamic-environments velocity-control embodied-ai linux ros

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Zhefan-Xu/NavRL")