runtime_integrity v0.3-alpha: ROS2 observer catches wheel slip and localization jumps via diagnostics
Observer detects when your robot’s wheels slip or pose jumps—without touching control code.
Two weeks after pivoting from an invasive hard-braking middleware, developer Zc_Liu has released runtime_integrity v0.3-alpha for ROS2. The package is a pure observer that runs alongside existing autonomy stacks, checking whether the physical robot actually moves consistently with the commands sent to /cmd_vel. It outputs execution-integrity state as standard ROS diagnostics (/diagnostics) without requiring a new dashboard, custom protocol, or any control authority. Key detection modes include wheel slip (commanded motion vs. odometry progress), localization jumps (sudden pose discontinuities from EKF/SLAM), and command-stream timing disturbances.
The tool is intentionally conservative: it does not modify base drivers, controllers, or Nav2 behavior trees. Instead, it surfaces residuals like totalResidual and cmdOdomResidual under diagnostic topics such as runtime_integrity/execution_integrity. This design makes it immediately compatible with existing visualization tools (Foxglove Studio, rqt_robot_monitor) and diagnostic aggregators. The developer is seeking feedback on whether it should remain a standalone observer, become a Nav2-adjacent module, or integrate into ros2_control, and how best to expose short-lived events like localization jumps without turning the observer into a controller.
- Observes /cmd_vel and /odom over short windows to detect wheel slip, localization jumps, and timing inconsistencies.
- Outputs standard /diagnostics messages (e.g., diagnosticLevelName, dominantCause, residuals) consuming by Foxglove, rqt_robot_monitor, etc.
- Purely observe-only: no command interception, controller modification, or base driver changes required.
Why It Matters
For AMR/AGV fleets, catching physical vs. commanded motion mismatches early prevents collisions and improves safety compliance.