Multi-Robot Informed Path Planning under Communication Constraints

Presenter Information

Brad WoosleyFollow

Advisor Information

Raj Dasgupta

Location

UNO Criss Library, Room 232

Presentation Type

Oral Presentation

Start Date

2-3-2018 10:30 AM

End Date

2-3-2018 10:45 AM

Abstract

In the unmanned exploration of extraterrestrial surfaces, or when collecting valuable information after disasters, teams of autonomous robots can be deployed to collect and communicate the information to a central server. The environment is often initially unknown and communications between robots are unreliable and intermittent. We propose a unified solution to this problem where each robot uses Gaussian processes (GPs) to model the distribution of information entropy and communication signal strength in the environment. The two GPs are combined into a single objective function representing the utility of different candidate locations to explore and solved as a constrained utility maximization problem. Robots periodically share their collected information and exploration locations with each other to avoid repeated exploration. Initial simulation experiments show that our proposed approach improves on the distance required to reach similar estimations of the phenomena of interest compared to an approach based on information entropy and distance alone.

This document is currently not available here.

COinS
 
Mar 2nd, 10:30 AM Mar 2nd, 10:45 AM

Multi-Robot Informed Path Planning under Communication Constraints

UNO Criss Library, Room 232

In the unmanned exploration of extraterrestrial surfaces, or when collecting valuable information after disasters, teams of autonomous robots can be deployed to collect and communicate the information to a central server. The environment is often initially unknown and communications between robots are unreliable and intermittent. We propose a unified solution to this problem where each robot uses Gaussian processes (GPs) to model the distribution of information entropy and communication signal strength in the environment. The two GPs are combined into a single objective function representing the utility of different candidate locations to explore and solved as a constrained utility maximization problem. Robots periodically share their collected information and exploration locations with each other to avoid repeated exploration. Initial simulation experiments show that our proposed approach improves on the distance required to reach similar estimations of the phenomena of interest compared to an approach based on information entropy and distance alone.