Multi-Robot Task Allocation: A Stochastic Queuing Approach

Advisor Information

Prithviraj Dasgupta

Location

Milo Bail Student Center Ballroom

Presentation Type

Poster

Start Date

8-3-2013 1:00 PM

End Date

8-3-2013 4:00 PM

Abstract

Multi-Robot Task Allocation (MRTA) is an important area of research in autonomous multi-robot systems. The main problem in MRTA is to match a set of robots to a set of tasks so that the tasks can be completed by the robots while optimizing a certain metric such as the time required to complete all tasks, distance traveled by the robots and energy expended by the robots. We consider a scenario where the tasks can appear dynamically and the location of tasks are not known a priori by the robots. Additionally, for a task to be completed, it needs to be performed by multiple robots. This setting is called the MR-ST-TA (multi-robot, single-task, time-extended assginment) category of MRTA and solving the MRTA problem for this category is a known NP hard problem. We address this problem by proposing a new algorithm that uses a stochastic queue-based model to allocate tasks between robots. We have implemented our algorithm on an accurately simulated model of Corobot robots within the Webots simulator for different numbers of robots and tasks. We investigate a decentralized implementation of our algorithm and compare our results to several other approaches baselined by an offline optimal schedule generated by applications of the Hungarian algorithm.

This document is currently not available here.

COinS
 
Mar 8th, 1:00 PM Mar 8th, 4:00 PM

Multi-Robot Task Allocation: A Stochastic Queuing Approach

Milo Bail Student Center Ballroom

Multi-Robot Task Allocation (MRTA) is an important area of research in autonomous multi-robot systems. The main problem in MRTA is to match a set of robots to a set of tasks so that the tasks can be completed by the robots while optimizing a certain metric such as the time required to complete all tasks, distance traveled by the robots and energy expended by the robots. We consider a scenario where the tasks can appear dynamically and the location of tasks are not known a priori by the robots. Additionally, for a task to be completed, it needs to be performed by multiple robots. This setting is called the MR-ST-TA (multi-robot, single-task, time-extended assginment) category of MRTA and solving the MRTA problem for this category is a known NP hard problem. We address this problem by proposing a new algorithm that uses a stochastic queue-based model to allocate tasks between robots. We have implemented our algorithm on an accurately simulated model of Corobot robots within the Webots simulator for different numbers of robots and tasks. We investigate a decentralized implementation of our algorithm and compare our results to several other approaches baselined by an offline optimal schedule generated by applications of the Hungarian algorithm.