graceful-controller

Add a controller plugin entry with plugin nav2_graceful_controller::GracefulController to the controller_server YAML, then tune min_turning_radius, v_linear_max, and slowdown_radius for your robot. Reference it in the behavior tree via the controller_id field of the FollowPath action.

List all plugin names in controller_plugins and define each with its own plugin class and parameters. The behavior tree selects which one runs per path segment using the controller_id field, and switching mid-path is seamless since the new controller receives the remaining path via setPlan.

GracefulController produces smooth curves to a single goal pose, ideal for docking. MPPI handles complex obstacle-dense environments with optimal control, while Regulated Pure Pursuit suits simple paths with low CPU cost. RotationShim can wrap any controller for initial rotation.

Avoid it in complex obstacle-dense environments because it has no costmap-aware trajectory scoring, and for long-distance navigation since it only approaches a single goal. Use MPPI or DWB for those scenarios instead.

No. Only the active controller runs computeVelocityCommands, so idle controllers consume no CPU. All controllers share the same local costmap, controller_frequency, progress_checker, and goal_checker settings.