Control and Random Searching with Multiple Robots

This paper presents the work of several ongoing studies to determine the effectiveness of multiple small robotic vehicles for performing mine field clearance, and the related problem of clearing unexploded ordnance from areas of interest. Many issues are implied in this opening sentence. Not the least of these is knowing, out the many items cluttering a battlefield, which ones need to be cleared. There is the problem of transiting through dangerous areas with the threat of detonation, the difficulties of open field navigation and rough terrain, and the dangerous task of picking up unexploded charges. Current technology employs brute force, is often overt, or the use of human hands with the potential for loss of life and / or limb. It is of interest then, to explore whether improvements in safety and performance can be made using small smart machines that have the capability of transiting an open area, obstacle avoidance, and picking up an piece of ordnance, or placing a charge that could be detonated upon command. Results given in this paper includes the performance of clearance operations with behavior based robots in random search. This is where the formations are random, allowing for low cost controllers. Included are the effects of the use of multiple vehicles, the influence of various levels of detection probability, and some estimation of the losses suffered under a given probability that detonation will occur upon ordnance recovery. INTRODUCTION The Navy's Explosive Ordnance Disposal (EOD) research and development department has recently been active in the pursuit of small robots a Basic Unexploded Ordnance Gathering System (BUGS) , as an aid to EOD technicians who are required to enter the battlefield , or test firing range, to clear improved conventional munitions(ICM) [1]. The munitions do not all detonate upon delivery leaving about five percent in a dangerous state. Two methods of clearance are to pick up the offending objects and place hem on a pile for later disposal, or to simply blow them up in place. The pick up and carry away scenario is the subject of this study. We will discuss the clearance performance of multiple robots in performing random search. Since any field of interest will also be littered with obstacles, reliable obstacle avoidance methods are essential, and target detection sensor(s) are integral to every concept. Additionally, candidate robots must have a reliable capability to pick up the selected object and return to the designated pile point. RANDOM SEARCH Given a purely random search for unknown targets within an area A, using a perfect sensor of detection radius, r, traveling at speed U, we may assume that the probability of detection is proportional to the mean target density, A / ) t ( n , times the area sweep rate [2]. With an imperfect sensor where the probability of detection, conditioned on target presence is p, we can deduce that the expected rate of target acquisition, ) t ( q& is ) A / ) t ( n ( pN ) r 2 ( U ) t ( q = & . Related to the above, n ( t ) is the average number of targets remaining at time t, so that,