MedlarTea / s2s-teach-and-repeat

It's a scan-to-scan-based 2d LiDAR teach and repeat.

Scan-to-Scan-based Teach and Repeat

This work is not done yet.

Introduction

It's a scan-to-scan-based 2d LiDAR teach and repeat. In teaching step, it will store the "keyscan" after certain distance. In repeating step, it will repeat the teaching path by ICP (scan-to-scan) pose estimation and wheel odometry information.

In teach step:

• Input: odometry, 2d scan, and joystick
• Output: keyscans

In repeat step:

• Input: odometry, 2d scan, joystick and scans
• Output: cmd_vel

Hardware

Our sony joystick:

Software

Some cpp packages

• Eigen
• OpenCV
• PCL

Here, we use ETH pointmatcher to accomplish ICP process for its high speed.

1. Follow pointmatcher to install pointmatcher
2. Install libpointmatcher_ros to Connect ROS and PCL:

git clone https://github.com/MedlarTea/libpointmatcher
# then copy "libpointmatcher_ros" to your catkin_ws/src

Compile

Teach and Repeat

Observations

• Wheel odometry information is much precise than ICP pose estimation

Acknowledge

This project depends a lot on libpointmatcher and ethzasl_icp_mapping