Lane finding using homogeneous groups of cooperating autonomous vehicles

Abstract

This paper describes the problem of finding a clear lane through a potentially mined region. Using a group of cooperating AUVs, we propose several solutions and compare their advantages and disadvantages. The goal is to find straight paths clear of obstacles from one edge of a rectangular search area to the other. Homogeneous robots are used to find paths clear of all obstacles. The number of robots, their communications abilities and their sensory abilities are varied. The interleaved method shows the effectiveness of a team of interleaved robots working their way across the region. In this approach, the robots start in neighboring swaths and as they find blocked swaths (blocked by either a mine or-an obstacle) they break off and move to a new swath. This approach uses the least shared information and results in long search times that decrease predictably with the number of robots used and with the width of the sensor swath. The regional method takes a different approach, where each robot is assigned a single territory within the search area. There is little cooperation between robots; one robot would only change its path if information from its neighbor indicates the possible presence of a clear lane in a new area within its region. Search times using this approach tend to be long, and cooperation rarely shortens the search time. Preliminary results show that this cooperative approach can substantially increase the time necessary to survey the region if there are partial lanes at region boundaries. However, the regional simulation also shows a substantial decrease in the amount of time it takes to determine whether or not clear lanes are possible, compared to the interleaved approach. In the hybrid method the search area is divided into smaller regions, defined as a specific number of swaths. The robots are initially assigned to neighboring regions, and as a region is invalidated or completed, the affected robot would move to the next unassigned region. This approach forms a bridge, with the interleaved approach at one extreme (one swath per region) and the regional approach at the other (total number of swaths/number of robots per region). All three approaches were simulated in ALWSE MATLAB. The hybrid approach was also implemented in autonomous ground vehicles as a physical simulation before implementation in target (amphibious) hardware.