Presentation Title

Task Assignment for Multi-Robot Systems of Heterogeneous Ability Using Negotiation

Start Date

November 2016

End Date

November 2016

Location

HUB 302-#20

Type of Presentation

Poster

Abstract

As robots become cheaper to manufacture in large numbers, multi-robot systems using teams of simple robots to complete tasks are becoming more appealing. Many multi-robot systems involve homogeneous or heterogeneous groups, the latter where robots differ only in their capabilities. Heterogeneous groups of robots with multiple capabilities and varying abilities to express those capabilities have been mostly overlooked. We attempt to solve the problem of how to dynamically and efficiently assign subtasks to these diverse groups of robots. For the purpose of this work, capability is the existence of the capacity of an agent to perform some action, and ability is the extent which a capability is expressed. Many algorithms for task assignment of heterogeneous robots deal with only the capability to perform some set of actions and managing the overlap of capabilities in a swarm. We look at heterogeneous ability to express capabilities and how taking this into account affects system performance. In our example application, we investigate the heterogeneous ability of robots to carry materials to a depot after collecting them from the environment, where the robots have differing ability to store materials and speed of movement. Our work is a step towards applications involving more dynamic heterogeneous multi-robot systems. Taking into account differing ability to express a capability allows for better fault tolerance and dynamic upgrades to a system. We simulate a multi-robot system of vehicles moving at differing speeds and carrying differing amounts of materials. To test performance, we implement an existing negotiation algorithm that takes into account only distances between robots, objects, and the depot as well as our modified algorithm which takes also into account abilities of individuals. We compare the completion times of each, looking at how each scales with the number of robots and degree of heterogeneity.

This document is currently not available here.

Share

COinS
 
Nov 12th, 4:00 PM Nov 12th, 5:00 PM

Task Assignment for Multi-Robot Systems of Heterogeneous Ability Using Negotiation

HUB 302-#20

As robots become cheaper to manufacture in large numbers, multi-robot systems using teams of simple robots to complete tasks are becoming more appealing. Many multi-robot systems involve homogeneous or heterogeneous groups, the latter where robots differ only in their capabilities. Heterogeneous groups of robots with multiple capabilities and varying abilities to express those capabilities have been mostly overlooked. We attempt to solve the problem of how to dynamically and efficiently assign subtasks to these diverse groups of robots. For the purpose of this work, capability is the existence of the capacity of an agent to perform some action, and ability is the extent which a capability is expressed. Many algorithms for task assignment of heterogeneous robots deal with only the capability to perform some set of actions and managing the overlap of capabilities in a swarm. We look at heterogeneous ability to express capabilities and how taking this into account affects system performance. In our example application, we investigate the heterogeneous ability of robots to carry materials to a depot after collecting them from the environment, where the robots have differing ability to store materials and speed of movement. Our work is a step towards applications involving more dynamic heterogeneous multi-robot systems. Taking into account differing ability to express a capability allows for better fault tolerance and dynamic upgrades to a system. We simulate a multi-robot system of vehicles moving at differing speeds and carrying differing amounts of materials. To test performance, we implement an existing negotiation algorithm that takes into account only distances between robots, objects, and the depot as well as our modified algorithm which takes also into account abilities of individuals. We compare the completion times of each, looking at how each scales with the number of robots and degree of heterogeneity.