Skip to content

Parallelize Conflict Based Search for faster multi-robot planning

Notifications You must be signed in to change notification settings

jsongCMU/parallel_CBS

Repository files navigation

Background

This repo is our final project for Carnegie Mellon University course 15-618: Parallel Computer Architecture and Programming, Spring 2023. Key contributors are Matthew Booker and Jaekyung Song.

Overview

We parallelized Conflict Based Search (CBS) in multiple different ways, each with its own branch. The key ones are:

  • HDA_SAS: implementation of Hash Distributed A-star (HDA) using Shared Address Space (SAS) model. Used to parallelize the low-level search of CBS
  • parallel_cbs_naive: parallelization of the high level search of CBS. Each core is given a subtree to expand and search within
  • parallel_cbs_v2.0: parallelization of high level serach of CBS. Cores will generate and share work, focusing on the best nodes in the constraints tree and balancing workload

Website

Website for this project can be found here

About

Parallelize Conflict Based Search for faster multi-robot planning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages