Robotics

Feel like TurtleBot4_Navigation is a "House of Cards" for my robot

After 6,226 meters and 2,137 dockings, a robot's navigation stack crumbles under pressure.

Deep Dive

A veteran robotics developer reports that their TurtleBot4 clone, TB5-WaLI, powered by a Raspberry Pi 5 with 8GB RAM, has been operational for 11,108 hours since January 9, 2025. During this period, it autonomously undocked, navigated 6,226 meters (nearly 4 miles) around a home, and redocked 2,137 times, with only 5 manual interventions after safety shutdowns. The robot built a reliable map using LIDAR and never lost localization. However, the developer achieved only 90% navigation reliability to 10 carefully chosen goals, and the system fails unpredictably in tight spaces, during resource starvation (e.g., DDS discovery queries), or when external nodes monitor navigation events.

The developer criticizes the TurtleBot4_Navigation stack, based on ROS 2 Jazzy and Nav2, for its fragility. Despite tuning numerous path planners, critics, and recovery planners, the robot often fails to recover from tight spots, with error messages like 'spin failed' appearing without any actual rotation attempt. The developer notes that the control loops run at only 10Hz on the Raspberry Pi 5, with CPU usage at 75% during navigation, and argues that the system should slow down or pause planning instead of failing. They conclude that the infinite parameterization and flexibility of Nav2 do not compensate for its lack of robustness in real-world, resource-constrained environments.

Key Points
  • TB5-WaLI ran for 11,108 hours, navigating 6,226 meters and completing 2,137 dockings with 90% reliability.
  • Navigation fails in tight spaces, ignores recovery behaviors, and crashes under DDS discovery query resource starvation.
  • Control loops run at 10Hz on Raspberry Pi 5 (75% CPU load), but the stack can't adapt by slowing down or pausing planning.

Why It Matters

Highlights real-world reliability issues in ROS 2 Nav2 for resource-constrained robots, urging better failure handling.