Skip to content

Latest commit

 

History

History
70 lines (65 loc) · 3.05 KB

README.md

File metadata and controls

70 lines (65 loc) · 3.05 KB
Raw

This repository contains code for simulating a sequence of local 2D point clouds that are scanned from a pose trajectory in a virtual 2D environment

Get Started

The simulation can be decomposed into three steps that generate

  • 2D environment
  • Pose trajectory
  • Local point clouds

Step 1: Creating 2D Environment

To create 2D environment, go to ./code and run

python sample_maps.py --n_env 2 --img_h 1024 --img_w 1024

This command simulates 2 different environments and saves them in ../results/ with 3 sub-folders (v0 and v2). In each sub-folder, there is a binary image environment.png that visualizes the simluated envrionment where the black pixels show the occupied locations and the white pixels are free-space.

Step 2: Sampling Pose Trajectory

The pose trajectory is sampled interactively using the matlab script script_samplePoseTrajectory.m. You need to specify the environment name that is simulated from step 1 and also provide a name for this pose trajectory (because you can sample multiple trajectories within the same environment).

%% settings
env_idx = 'v0';
pose_idx = 'pose0';

The sampled pose will be saved as ../results/v0_pose0/gt_pose.mat. When you run the script, a figure window will be created that displays the binary environment image. You need to select a polyline in the current figure using the mouse. This polyline is the base of the trajectory where the poses are sampled.

Other parameters you can adjust

n_pose_on_traj = 256;  % number of poses sampled on this trajectory
angle = 10; % max angular differences (in degree) two poses
displacement = 5; % controls the displacement of the sampled pose location from the polyline

Step3: Scanning Local Point Cloud

To simulate 2D point clouds from the sampled pose trajectory in the 2D environment, run

python scan_local_point_clouds.py --data_dir ../results/ --env v0 --pose pose0 --obs 256

A set of local point clouds will be saved as pcd files in ../results/v0/v0_pose0/. The argument --obs controls the number of scanned points in each point cloud.

File Structure

Once you finish the simulation, the results folder should look like (depending on how many environments and poses you simulated)

./results
├── v0
│   ├── environment.pkl
│   ├── environment.png
│   ├── v0_pose0
│   │   ├── 000.pcd
│   │   ├── 001.pcd
│   │   ├── ...
│   │   ├── ...
│   │   ├── gt_pose.mat
│   │   └── gt_pose.png
│   └── v0_pose1
│       ├── 000.pcd
│       ├── 001.pcd
│       ├── ...
│       ├── ...
│       ├── gt_pose.mat
│       └── gt_pose.png
├── v1
│   ├── environment.pkl
│   ├── environment.png
│   ├── v1_pose0
│   │   ├── 000.pcd
│   │   ├── 001.pcd
│   │   ├── ...
│   │   ├── ...
│   │   ├── gt_pose.mat
│   │   └── gt_pose.png