ros-navigation/navigation2

Bug report Required Info: When using nav2_smac_planner::SmacPlanner2D on Jazzy, planner_server logs: Inflation layer either not found or inflation is not set sufficiently for optimized non-circular collision checking capabilities. It is HIGHLY recommended to set the inflation radius to be at MINIMUM half of the robot's largest cross-section... ROS 2 distro: Jazzy Navigation2 version: 1.4.0 Planner plugin: nav2_smac_planner::SmacPlanner2D Costmap plugin: nav2_costmap_2d::InflationLayer On startup, planner_server emits the inflation-layer error. Removing SmacPlanner2D from planner_plugins makes the warning disappear. Runtime parameter inspection shows the inflation layer is present and configured with a big enough radius. Steps to reproduce issue Happens at startup Expected behavior SmacPlanner2D should not log an inflation-layer misconfiguration error when the warning condition is not relevant to its 2D radius-based collision checking mode. Actual behavior Additional information I t

It seems like a good idea that we run thread and address sanitizers in our CI to uncover issues which would otherwise be found in production deployments and/or users running these sanitizers manually during testing. We should compile all of Nav2 in our CI to use these with options to stop on failures, leaks, etc so our CI fails in cases that this occurs with the logs. Then, run Nav2's system tests to capture these metrics in particular and save the failure logs in the artifacts for later debugging. It may be nice to have the option to run the Nav2 System Tests as-is without this to check the behavior working and a separate parallel job to run with the sanitizers enabled (TBD) as they serve slightly different purposes and we may be able to build more sanitizer-specific tests over time or non-sanitizer based system testing.

Create a fence costmap layer which can be used instead of a static layer to re-set the bounds of a global costmap based on a zone of operations dynamically at run-time

As the description implies, we may be able to gain some performance improvements using OpenCV rather than manually rastering polygons on the costmap for the vector object server. If so and possible, it might be good to pull that source code here to remove the direct dependency on openCV. Good community improvement issue!

Summary The nightly CI is currently failing across all three RMW implementations: FastDDS, CycloneDDS, and Zenoh Affected RMWs rmw_fastrtps rmw_cyclonedds rmw_zenoh Additional Information Zenoh — may be related to #4948 CycloneDDS & FastDDS — likely an upstream issue, but not yet confirmed. Community help needed to dig in. CycloneDDS & Zenoh — both appear to have started failing after the last sync. Full logs can be found here: https://app.circleci.com/pipelines/github/ros-navigation/navigation2?branch=main Also, please make sure any fix is fully built and tested locally before submitting a PR. AI-generated PRs without any building or testing will not be accepted