An Initial Evaluation of Approaches to Building Entry for Large Robot Teams

  • Authors:

  • Paul Scerri;Zheng Ma;Shih-Yi Chien;Huadong Wang;Pei-Ju Lee;Michael Lewis;Katia Sycara

  • Affiliations:

  • School of Computer Science, Carnegie Mellon University, Pittsburgh, USA 15213;School of Information Sciences, University of Pittsburgh, Pittsburgh, USA 15260;School of Information Sciences, University of Pittsburgh, Pittsburgh, USA 15260;School of Information Sciences, University of Pittsburgh, Pittsburgh, USA 15260;School of Information Sciences, University of Pittsburgh, Pittsburgh, USA 15260;School of Information Sciences, University of Pittsburgh, Pittsburgh, USA 15260;School of Computer Science, Carnegie Mellon University, Pittsburgh, USA 15213

  • Venue:
  • Journal of Intelligent and Robotic Systems
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Teams of robots offer potential reductions in risk for rescuers and improved rescue rates for search and rescue in collapsed buildings and other restricted environments. However, one aspect of deployment of such teams that is often overlooked is the initial entry of the robots into the building. If many robots enter the building from the same entrance, they are likely to interfere with each other and dramatically slow the initial phase of exploration. In this paper, we use a low fidelity simulation to evaluate several approaches to overcoming this initial congestion and identify heuristics that might allow the robots to most quickly clear the entry area and begin their actual mission. Our results show that unless the exits to the initial entry area are small, over a wide range of parameters, the most effective technique was for every robot to move in the direction of the longest open space it senses. In more cluttered environments with smaller exits, wall following, a strategy less dependent on the quality of sensing, proved more effective. Informal comparisons of these algorithms in a high fidelity simulation with limited trials and no variation in parameters suggested that simply promoting dispersion might be a more effective strategy for realistically complex dynamics.