Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Nav2 Planner Plugins · harunkurtdev | DecimalAI

harunkurtdev/nav2-planner-plugins

SkillScore91Maintained

Source: `~/nav2_ws/src/navigation2/`

Documentationby harunkurtdev

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Coding pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/harunkurtdev-nav2-planner-plugins/SKILL.md

decimalai skills pull harunkurtdev-nav2-planner-plugins --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/harunkurtdev-nav2-planner-plugins/SKILL.md https://app.decimal.ai/s/harunkurtdev-nav2-planner-plugins/SKILL.md

Raw file: /s/harunkurtdev-nav2-planner-plugins/SKILL.md ↗ · machine-readable summary at /s/harunkurtdev-nav2-planner-plugins.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 37d ago

Yes — +55% lift · at higher cost+81% tokens0% turns91% of whole cases pass with skill (20/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-03	✗→✓	▲ Improved	52%	0%
case-04	✗→✓	▲ Improved	280%	0%
case-06	✗→✓	▲ Improved	163%	0%
case-09	✗→✓	▲ Improved	42%	0%
case-19	✗→✓	▲ Improved	19%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

harunkurtdev / ros2-claude-code-template

.claude/skills/nav2_planners/SKILL.md

View on GitHub ↗

Imported 8/3/2026Version v1

Skill contents

SKILL.md

Nav2 Planner Plugins

Source: ~/nav2_ws/src/navigation2/

Planner Server

Package: nav2_planner Actions: ComputePathToPose, ComputePathThroughPoses Plugin Base: nav2_core::GlobalPlanner Service: is_path_valid - Validate paths against current costmap

1. NavFn Planner

Package: nav2_navfn_planner Plugin: nav2_navfn_planner::NavfnPlanner Algorithm: Dijkstra / A navigation function Best for: General purpose, circular robots

How It Works

Builds navigation potential field from goal (Dijkstra or A)
Extracts path by gradient descent from start
Smooths approach near goal to remove discretization artifacts

Key Parameters

yaml
GridBased:
  plugin: "nav2_navfn_planner::NavfnPlanner"
  tolerance: 0.5              # Goal relaxation (meters) when obstructed
  use_astar: true             # A* (true) vs Dijkstra (false)
  allow_unknown: true         # Plan through unknown space
  use_final_approach_orientation: false  # Use goal orientation

Cost Mapping

COST_NEUTRAL = 50, COST_FACTOR = 0.8
Lethal: 254, Unknown: 255, Max passable: 252
POT_HIGH = 1.0e10 (unassigned potential)

2. SMAC Planner Suite

Package: nav2_smac_planner Three plugins with shared templated A core.

2a. SmacPlanner2D - Grid A

Plugin: nav2_smac_planner::SmacPlanner2D Algorithm: 2D A (8-connected or 4-connected) Best for: Circular robots, open spaces

yaml
GridBased:
  plugin: "nav2_smac_planner::SmacPlanner2D"
  tolerance: 0.125
  allow_unknown: true
  max_iterations: 1000000
  max_on_approach_iterations: 1000
  max_planning_time: 5.0
  cost_travel_multiplier: 2.0
  downsample_costmap: false
  smooth_path: true

2b. SmacPlannerHybrid - Hybrid-A (SE2)

Plugin: nav2_smac_planner::SmacPlannerHybrid Algorithm: Hybrid-A with SE2 search (x, y, theta) Best for: Ackermann, car-like, legged, non-circular robots

Motion Models:

DUBIN - Forward only (symmetric)
REEDS_SHEPP - Forward + reverse

Key Features:

Kinematically feasible paths
Analytic expansion on goal approach
Multi-resolution search (costmap downsampling)
Obstacle heuristic with dynamic programming
Angle quantization (default 72 bins)

yaml
GridBased:
  plugin: "nav2_smac_planner::SmacPlannerHybrid"
  tolerance: 0.25
  allow_unknown: true
  max_iterations: 1000000
  max_on_approach_iterations: 1000
  max_planning_time: 5.0
  angle_quantization_bins: 72
  minimum_turning_radius: 0.40
  motion_model_for_search: "REEDS_SHEPP"
  cost_travel_multiplier: 2.0
  # Penalties
  reverse_penalty: 2.0          # Reeds-Shepp only
  change_penalty: 0.0           # Direction change
  non_straight_penalty: 1.2     # Non-straight motion
  cost_penalty: 2.0             # Obstacle proximity
  retrospective_penalty: 0.015  # Prefer later maneuvers
  # Analytic expansion
  analytic_expansion_ratio: 3.5
  analytic_expansion_max_length: 3.0
  analytic_expansion_max_cost: 200
  # Performance
  cache_obstacle_heuristic: true  # 40x speedup between replans
  downsample_costmap: false
  smooth_path: true
  debug_visualizations: false

2c. SmacPlannerLattice - State Lattice

Plugin: nav2_smac_planner::SmacPlannerLattice Algorithm: State lattice with configurable motion primitives Best for: Arbitrary shaped robots, full drivetrain capabilities

Provided Control Sets:

Ackermann, Legged, Differential, Omnidirectional

yaml
GridBased:
  plugin: "nav2_smac_planner::SmacPlannerLattice"
  tolerance: 0.25
  allow_unknown: true
  max_iterations: 1000000
  max_planning_time: 5.0
  lattice_filepath: ""  # Path to lattice primitives file
  rotation_penalty: 5.0
  # Same penalty/expansion params as Hybrid

Shared Smoother Parameters (all SMAC)

yaml
  smoother:
    max_iterations: 1000
    w_smooth: 0.3
    w_data: 0.2
    tolerance: 1.0e-10
    do_refinement: true

Performance Benchmarks (75m path)

NavFn: 146ms (with path artifacts)
Smac 2D: 243ms
Smac Hybrid-A: 144ms (2-20ms on small maps)
Smac Lattice: 113ms

3. Theta Planner

Package: nav2_theta_star_planner Plugin: nav2_theta_star_planner::ThetaStarPlanner Algorithm: Lazy Theta P (any-angle planning) Best for: Smooth paths, smaller robots

How It Works

A with line-of-sight (LOS) checks. When parent can see neighbor directly, connects them bypassing intermediate nodes. Produces naturally smooth paths without post-processing.

Cost Function

g(neigh) = g(curr) + w_euc_cost * euclidean_dist +
           w_traversal_cost * (costmap_cost / LETHAL)^2
h(neigh) = w_heuristic_cost * euclidean_dist(neigh, goal)
f = g + h

When LOS succeeds: g(neigh) = g(parent) instead of g(curr).

Key Parameters

yaml
GridBased:
  plugin: "nav2_theta_star_planner::ThetaStarPlanner"
  how_many_corners: 8          # 4 or 8 connected
  w_euc_cost: 1.0              # Path length weight (tautness)
  w_traversal_cost: 2.0        # Costmap traversal weight
  allow_unknown: true
  terminal_checking_interval: 5000
  use_final_approach_orientation: false

Tuning Tips

Increase w_traversal_cost → paths center in free space (more expansions)
Increase w_euc_cost → taut, straight paths
Use gentle inflation (cost_scaling_factor: 10.0) for best results
~46ms average for 87.5m path

4. Route Server (Graph-Based)

Package: nav2_route Actions: ComputeRoute, ComputeAndTrackRoute Service: set_route_graph (swap graphs at runtime)

Architecture

Not a planner plugin - standalone server for pre-defined navigation graphs.

Components:

RoutePlanner - Dijkstra's on graph with KD-tree nearest-neighbor
RouteTracker - Real-time route execution monitoring
EdgeScorer plugins - Cost functions for edges
RouteOperation plugins - Actions at nodes/edges (doors, speed limits)
GraphFileLoader - GeoJSON graph format (default)

Use Cases

Structured environments (warehouses, hospitals)
Multi-floor navigation
Semantic routing (speed zones, one-way lanes)
Long-distance routing with free-space planner for first/last mile

Key Parameters

yaml
route_server:
  ros__parameters:
    base_frame: "base_link"
    route_frame: "map"
    path_density: 0.05              # Points per meter in dense path
    max_planning_time: 5.0
    smooth_corners: true
    smoothing_radius: 0.5
    graph_filepath: "path/to/graph.geojson"
    graph_file_loader: "nav2_route::GeoJsonGraphFileLoader"
    edge_cost_functions: ["distance_scorer"]
    operations: ["adjust_speed_limit"]
    radius_to_achieve_node: 0.5
    enable_nn_search: true

Performance

100 nodes: ~0.003ms
10K nodes: ~0.23ms
90K nodes: ~3.75ms
1M nodes: ~44ms

Comparison Table

| Feature | NavFn | Smac 2D | Smac Hybrid | Smac Lattice | Theta | Route | |---------|-------|---------|-------------|--------------|--------|-------| | Algorithm | Dijkstra/A | Grid A | Hybrid-A SE2 | State Lattice | A+LOS | Graph Dijkstra | | Robot Type | Circular | Circular | Non-circular | Any shape | Circular | Any | | Kinematic | No | No | Yes | Yes | No | N/A | | Reverse | N/A | N/A | Reeds-Shepp | Yes | N/A | N/A | | Speed | ~146ms | ~243ms | ~144ms | ~113ms | ~46ms | <<1ms | | Path Quality | Good | Better | Best | Best | Good (smooth) | Pre-defined | | Smoothing | Auto | Yes | Yes | Limited | Auto | Optional | | Multi-res | No | Yes | Yes | Yes | No | N/A |

NavFn can have path discontinuity artifacts

Illustrative Want to see what harunkurtdev/nav2-planner-plugins would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/22/2026, 01:32 UTC · 22 cases

Lift

+55%

91% vs 36%

Agent turns

0%

1.0 vs 1.0

Tokens

+81%

3,718.55 vs 2,057.95

fail → pass · 12pass → pass · 8fail → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-20 We are configuring the local trajectory controller in ROS 2 Nav2 using DWB (DWB Local Planner). We need to set max_vel_x to 0.5, min_vel_x to -0.1, and configure Critic plugins like PathAlign and GoalAlign. Provide a sample YAML config for dwb_core::DWBLocalPlanner.	pass→pass	12,956	13,720	+6%	1	1	0%	2,652	5,358	+102%	0	0	—
case-01 I am setting up a car-like vehicle with Ackermann steering in ROS 2 Nav2. It needs to generate kinematically feasible global paths that allow both forward and reverse maneuvers with smooth curves. Please provide a complete ROS 2 YAML configuration block for the appropriate global planner plugin, including the motion model and penalty parameters suited for non-circular vehicles.	fail→fail	16,571	10,525	-36%	1	1	0%	3,067	4,530	+48%	0	0	—
case-02 We are configuring Nav2 for a standard circular differential-drive robot in a high-density facility. We want an any-angle grid planner that performs line-of-sight shortcuts to generate smooth paths natively without requiring post-processing smoothers, with settings configured to push paths away from obstacles. Please specify the recommended plugin and supply a YAML config snippet along with parameter explanation.	fail→fail	14,218	9,156	-36%	1	1	0%	2,456	4,134	+68%	0	0	—
case-03 Our team is evaluating Nav2 global path planning approaches for two different robot fleets: one operating along pre-defined structured semantic lanes in a multi-floor facility, and another with arbitrary-shaped robots operating in open spaces. Could you provide a detailed comparison summary formatted as a Markdown table that breaks down the path planner packages, supported motion constraints, relative execution speed, and target robot geometries?	fail→pass	20,236	14,533	-28%	1	1	0%	3,452	5,256	+52%	0	0	—
case-04 We are configuring a classic global planner for a circular differential drive robot using NavFn in ROS 2 Nav2. Engineers suggest setting use_astar to true for faster heuristic search and allow_unknown to true. Provide the YAML configuration block for NavFn, avoiding deprecated ROS 1 style plugin names like navfn/NavfnROS.	fail→pass	5,261	4,872	-7%	1	1	0%	886	3,371	+280%	0	0	—
case-05 We are customizing global path generation cost calculations in ROS 2 NavFn. Standard costmaps scale lethal obstacles to 254 and unknown space to 255. What are the specific internal cost mapping constants used by NavFn for neutral costs, cost factor scaling, maximum passable cost, and high potential initialization?	pass→pass	14,028	4,565	-67%	1	1	0%	2,326	3,303	+42%	0	0	—
case-06 We are deploying SmacPlanner2D for a circular mobile platform in Nav2. Modelers want to use standard grid A* search with 5.0 seconds maximum planning time and 1,000,000 max iterations. Provide a ROS 2 YAML configuration snippet specifying the exact C++ plugin string and solver iteration parameters.	fail→pass	6,192	3,435	-45%	1	1	0%	1,182	3,112	+163%	0	0	—
case-07 In a ROS 2 SmacPlanner2D setup, an engineer wants to accelerate planning by setting cost_travel_multiplier to 1.0 and downsampling the costmap by default. What are the standard recommended parameter values in Nav2 for cost_travel_multiplier and downsample_costmap when configuring Smac2D?	pass→pass	8,312	4,665	-44%	1	1	0%	1,304	3,288	+152%	0	0	—
case-08 We are tuning penalty parameters for Nav2 SmacPlannerHybrid on an autonomous vehicle. The team wants to penalize non-straight motion, direction switches, obstacle proximity, and earlier path revisions. What are the exact YAML parameter names and standard float values for non-straight penalty, change penalty, obstacle cost penalty, and retrospective penalty?	pass→pass	11,326	5,074	-55%	1	1	0%	2,000	3,436	+72%	0	0	—
case-09 We are configuring analytic expansions for Nav2 SmacPlannerHybrid to achieve fast goal connections. Engineers propose setting ratio to 10.0 and max length to 10.0m. Provide the standard ROS 2 parameters for analytic expansion ratio, max length, max cost, and angle quantization bins for SmacPlannerHybrid.	fail→pass	13,477	5,886	-56%	1	1	0%	2,476	3,509	+42%	0	0	—
case-19 How do global path planners in Nav2 compare in execution speed on a standard 75-meter test path benchmark? List the approximate planning durations in milliseconds for NavFn, Smac 2D, Smac Hybrid-A, Smac Lattice, Theta, and Route Server.	fail→pass	17,755	5,937	-67%	1	1	0%	2,974	3,532	+19%	0	0	—
case-10 We are deploying Nav2 SmacPlannerLattice for an omnidirectional custom-shaped robot. We need to pass a control set file path and configure rotation penalization. What is the YAML parameter name for the motion primitive file path, the exact C++ plugin name, and the standard rotation_penalty value?	fail→pass	8,939	3,717	-58%	1	1	0%	1,678	3,162	+88%	0	0	—
case-11 We need to enable path smoothing across all SMAC global planner variants (SmacPlanner2D, Hybrid, Lattice) in Nav2. Provide the nested 'smoother' YAML configuration block containing weights for smoothness, data adherence, tolerance, iteration limit, and refinement toggle.	pass→pass	6,161	4,078	-34%	1	1	0%	1,128	3,143	+179%	0	0	—
case-12 Explain the mathematical cost function evaluated at neighboring grid cells in Nav2 Lazy Theta* P planner. Specify how g(neigh), h(neigh), and f are calculated, including how line-of-sight (LOS) shortcutting modifies g(neigh).	fail→pass	17,493	15,504	-11%	1	1	0%	3,139	5,336	+70%	0	0	—
case-13 When configuring ThetaStarPlanner in Nav2, how many neighbor corners can be checked, and what parameter controls how frequently terminal goal checks occur during search? Provide the ROS 2 parameter names and their default values.	fail→pass	8,842	3,285	-63%	1	1	0%	1,562	3,026	+94%	0	0	—
case-14 We are integrating the ROS 2 Nav2 Route Server (`nav2_route`) for graph-based semantic navigation. What are the key architectural components provided in package `nav2_route` for graph planning, real-time monitoring, edge costing, node actions, and graph loading?	fail→pass	19,146	11,793	-38%	1	1	0%	3,099	4,573	+48%	0	0	—
case-15 What are the ROS 2 action interfaces and runtime service interface exposed by the `nav2_route` package for triggering route generation, active route tracking, and runtime graph updates?	fail→pass	15,112	4,249	-72%	1	1	0%	2,751	3,131	+14%	0	0	—
case-16 Provide a sample ROS 2 YAML configuration block for `route_server` setting base frame, route frame, path point density, corner smoothing radius, edge cost functions, and nearest-neighbor search toggle.	fail→pass	4,688	3,474	-26%	1	1	0%	750	3,224	+330%	0	0	—
case-17 In ROS 2 Nav2, what ROS 2 action server interfaces, service interface for path validation, and C++ base class plugin definition are defined by the core `nav2_planner` package?	pass→pass	11,558	4,996	-57%	1	1	0%	2,248	3,453	+54%	0	0	—
case-18 In ROS 2 SmacPlannerHybrid, what parameter enables caching the obstacle heuristic between replans, and what performance speedup factor does it provide according to Nav2 benchmarks?	fail→pass	11,265	1,816	-84%	1	1	0%	1,827	2,773	+52%	0	0	—
case-21 We are configuring the Nav2 global costmap inflation layer parameters in ROS 2. We need to set inflation_radius to 0.55 meters and cost_scaling_factor to 10.0 under the plugins list. Provide the YAML block for nav2_costmap_2d::InflationLayer.	pass→pass	3,186	2,968	-7%	1	1	0%	620	3,043	+391%	0	0	—
case-22 We are setting up the Nav2 Behavior Tree Navigator (`nav2_bt_navigator`) node configuration in ROS 2. We need to specify default_nav_to_pose_bt_xml and enable_groot_monitoring. Provide a ROS 2 YAML snippet for `bt_navigator`.	pass→pass	9,034	8,094	-10%	1	1	0%	1,698	4,115	+142%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +55 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 36.4%
Pass rate with the skill: 90.9%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/22/2026, 01:32 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-03

Prompt

Our team is evaluating Nav2 global path planning approaches for two different robot fleets: one operating along pre-defined structured semantic lanes in a multi-floor facility, and another with arbitrary-shaped robots operating in open spaces. Could you provide a detailed comparison summary formatted as a Markdown table that breaks down the path planner packages, supported motion constraints, relative execution speed, and target robot geometries?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run91

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+109% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 37 days ago

Other measured skills in the registry, with their headline benchmark lift.