Smac Planner 2d

name: smac-planner-2d description: 'Configure Nav2 SMAC 2D planner. Use when using 8-connected grid search, setting maximum_iterations, or cost penalties.'

SMAC Planner 2D

Plugin

nav2_smac_planner::SmacPlanner2D

Algorithm

A* search on an 8-connected 2D costmap grid with a Wavefront (breadth-first) heuristic. Operates directly on the occupancy grid without considering robot kinematics. Produces grid-aligned paths that follow cell centers, then applies an optional internal smoother.

The Wavefront heuristic precomputes exact grid distances from the goal, accounting for obstacles. This makes the heuristic admissible and consistent, guaranteeing optimal paths on the grid.

Key Parameters

Downsampling

Reduces the costmap resolution before planning, trading precision for speed.

Downsampling is effective on large outdoor maps where cell-level precision is unnecessary. For indoor robots with tight corridors, keep it disabled or use factor 2 at most.

Smoother Parameters

The built-in smoother runs after path planning to reduce jaggedness.

The smoother minimizes a cost function balancing path smoothness against fidelity to the original planned path. If w_smooth >> w_data, paths become very smooth but may cut corners near obstacles.

Complete YAML Configuration

planner_server:
  ros__parameters:
    expected_planner_frequency: 20.0
    planner_plugins: ["GridBased"]
    GridBased:
      plugin: "nav2_smac_planner::SmacPlanner2D"
      tolerance: 0.25
      max_iterations: 1000000
      max_on_approach_iterations: 1000
      max_planning_time: 2.0
      cost_travel_multiplier: 2.0
      allow_unknown: true
      use_final_approach_orientation: false
      downsample_costmap: false
      downsampling_factor: 1
      smoother:
        max_iterations: 1000
        w_smooth: 0.3
        w_data: 0.2
        tolerance: 1.0e-10

When To Use

Good fit:

• Simple indoor environments with wide corridors.
• Fast replanning needed (operates on 2D grid, no angle dimension).
• Robot kinematic constraints are handled by the controller (DWB/MPPI), not the planner.
• Holonomic or omni-directional robots where any-direction movement is valid.

Poor fit:

• Robots with non-holonomic constraints where kinematically feasible paths from the planner matter (use Hybrid-A* instead).
• Environments where path smoothness is critical and controller-level smoothing is insufficient.

Comparison with NavFn

Both operate on the 2D costmap grid. Key differences:

• SMAC 2D uses A* with a Wavefront heuristic; NavFn uses Dijkstra or A* with a potential field approach.
• SMAC 2D includes a built-in smoother; NavFn requires a separate path smoother plugin.
• SMAC 2D supports costmap downsampling for large maps.
• SMAC 2D is generally faster for point-to-point planning due to the focused A* search with a strong heuristic.

Troubleshooting

• Planner fails in cluttered space: Increase max_iterations and max_planning_time. Check that the costmap inflation radius isn't making passages impassable.
• Paths hug walls: Increase cost_travel_multiplier to penalize high-cost cells more heavily.
• Goal unreachable but looks reachable in RViz: The goal may fall inside an inflated obstacle cell. Increase tolerance or check costmap inflation parameters.
• Jerky paths from the controller: Enable the smoother or increase w_smooth. Consider adding the nav2_constrained_smoother or nav2_smoother_server downstream.