Nav2 Pose Sampler

Sample random start and goal poses from ROS occupancy grid maps for navigation experiments. No ROS installation required.

What It Does

Given a standard ROS map (YAML + PGM/PNG), this tool:

1. Loads the occupancy grid and computes a distance-from-obstacles field
2. Identifies valid free-space cells with configurable obstacle clearance
3. Samples random (start, goal) pose pairs with distance constraints
4. Saves poses to JSON for reproducible experiments
5. Generates visualizations showing the map, valid area, and sampled poses

Designed for automated navigation benchmarking and RCT (Randomized Controlled Trial) data collection with Nav2.

Installation

pip install numpy Pillow scipy PyYAML matplotlib

Then clone and install:

git clone https://github.com/kas-lab/random-pose-generator.git
pip install -e .

Quick Start

Command Line

# Visualize valid sampling area with 30 random pose pairs
python3 main.py --map /path/to/map.yaml --visualize --n-samples 30

# Generate 3000 poses and save to JSON
python3 main.py --map /path/to/map.yaml --generate 3000

# Both at once — generate + visualize
python3 main.py --map /path/to/map.yaml --generate 500 --visualize

# Custom distance constraints and obstacle clearance
python3 main.py --map /path/to/map.yaml --generate 1000 \
    --min-distance 2.0 --max-distance 8.0 --clearance 0.3

# Visualize an existing poses file
python3 main.py --map /path/to/map.yaml --visualize \
    --poses ./pose_data/poses.json

# Reproducible generation with a seed
python3 main.py --map /path/to/map.yaml --generate 500 --seed 42

Output Format

JSON (poses.json)

[
  {
    "id": 0,
    "start": {"x": 3.45, "y": -1.20, "yaw": 1.57},
    "goal": {"x": 7.10, "y": 2.85, "yaw": -0.78},
    "distance": 5.42
  },
  ...
]

Visualization (sampling_area.png)

The visualization shows three panels side by side:

Panel	Description
Raw Map	The original occupancy grid (white = free, black = occupied, gray = unknown)
Valid Sampling Area	Green cells where poses can be sampled (free space with sufficient obstacle clearance)
Sampled Poses	Map overlay with green dots (start), red dots (goal), and blue dashed lines connecting each pair

Parameters

Parameter	Default	Description
--map	(required)	Path to ROS map YAML file
--clearance	0.5	Minimum distance from obstacles (meters)
--min-distance	3.0	Minimum start-to-goal distance (meters)
--max-distance	15.0	Maximum start-to-goal distance (meters)
--seed	None	Random seed for reproducibility
--output	./pose_data	Output directory

Map Format

The tool reads standard ROS map_server format:

map.yaml:

image: map.pgm
resolution: 0.05
origin: [-14.2, -12.2, 0.0]
occupied_thresh: 0.65
free_thresh: 0.25
negate: 0

The image file (.pgm or .png) should be a grayscale occupancy grid where white (254) is free space and black (0) is occupied.