smac-planner-hybrid-a-star

Add nav2_smac_planner::SmacPlannerHybridAstar as a planner_plugins entry in the planner_server YAML, then set minimum_turning_radius, motion_model_for_search, and angle_quantization_bins. A complete parameter block with penalties and smoother settings is provided in the Skill.

Differential drive robots can spin in place, but setting the radius to exactly zero produces degenerate motion primitives. Use a small value between 0.1 and 0.4 meters to get smooth arcs while retaining tight turning ability.

Dubins produces forward-only arcs and suits robots that cannot reverse. Reeds-Shepp allows forward and reverse arcs, making it preferable for differential drive robots and constrained maneuvers like parking or docking.

The 3D search space (x, y, heading) is inherently more expensive than 2D planning. Mitigate by downsampling the costmap, reducing angle_quantization_bins to 36, lowering max_planning_time, or shrinking lookup_table_size to match your map.

Wall-hugging happens when cost_penalty is too low, so the planner ignores costmap inflation costs. Increase cost_penalty from the default 2.0 to 3.0–5.0 to make the planner strongly avoid high-cost regions near obstacles.

Avoid it in simple open environments where NavFn plus controller smoothing suffices, when high-frequency replanning is required, or for holonomic robots that gain nothing from non-holonomic motion primitives.