Skip to content

🧠 AI Auto-Tuner Guide

Manually tuning SLAM parameters is tedious. The Auto-Tuner uses Bayesian Optimization to find the perfect configuration for your robot.

🌊 How It Works

Auto Tuning Flowchart

🛠️ Optimization Workflow

1. Define Search Space

Create a file in configs/tuning/ defining the ranges to explore.

# configs/tuning/gmapping_search.yaml
algorithm: gmapping
parameters:
  linearUpdate: [0.1, 0.5]  # Range: 0.1m to 0.5m
  particles: [30, 200]      # Range: 30 to 200
  sigma: [0.05, 0.2]

2. Launch Tuner

From the GUI, go to the Auto-Tuner tab.

  1. Select Objective: Minimize ATE (Recommended) or Maximize Coverage.
  2. Budget: Set number of trials (e.g., 50).
  3. Run: Click Start Optimization.

Time Consumption

Each trial runs a full simulation. 50 trials @ 2 mins/run = 1h 40m total duration. Running in Headless Mode speeds this up significantly.

📈 Analyzing Results

The tuner generates a tuning_report.pdf showing:

  • Convergence Plot: How quickly the error dropped.
  • Parallel Coordinates: Which parameters matter most.
  • Best Config: The winning YAML snippet.

Best Found

linearUpdate: 0.32
particles: 85
sigma: 0.08
ATE reduced by 40% compared to default.