Multi-Robot-Collision-Avoidance-with-VO

Abstract

This repository for multi-robot collision avoidance using the Velocity Obstacle.

Localization is performed through AMCL(Adaptive Monte Carlo localization) in a saved map made by Google Cartographer.

VO can be used to find the speed closest to the desired speed but not to collide. This is implemented using multiple python scripts and launch files.

Video(temp)

Limitation

This algorithm may require sufficient space. In the narrow space, robots can collide with the wall.
This is because the result of VO(velocity vector) may not follow Path Planning.

If dynamic object detection is also performed through LiDAR, human interaction is possible, but this is a different problem.

It may be a classical method, and in the ROS1, Local Path Planning is performed in another method.

Reference

• MengGuo : RVO for Multi-agent Systems
  : This is a simulation for the Holonomic multi-robot system. This repository is implemented with the help of this.
• Lee-JaeWon : Multi-Turtlebot3-Cartographer-And-Localization
  : Configuring namespace and TF Tree for multiple robots is difficult.
• ROBOTIS : turtlebot3
  
• Fiorini, P., & Shiller, Z. (1998). Motion Planning in Dynamic Environments Using Velocity Obstacles. The International Journal of Robotics Research, 17, 760 - 772.
  
• Claes, Daniel and Tuyls, Karl (2018) Multi robot collision avoidance in a shared workspace. AUTONOMOUS ROBOTS, 42 (8). 1749 - 1770.