Terrain Leveling by a Swarm of Simple Agents

In this paper we present a new algorithm for leveling a discrete terrain using autonomous agents of minimal complexity. Our primary interest is in developing robust strategies for eventual application in underwater construction. Terrain leveling underwater typically involves the use of large and expensive machinery that may be cumbersome to operate. We propose that a swarm of autonomous robots could be used instead. Each agent is implemented with minimal capabilities for perception and actuation. The goal of each agent is to estimate the average height of the terrain, and move material it finds above that average to locations below the average height. We implement the algorithm in a realtime three-dimensional simulator, as well as an offline simulator for the purpose of performance analysis. From these simulations we find that the algorithm functions correctly, and we provide preliminary performance data for various parameters. We conclude that terrain leveling may be a viable application for a swarm of minimalistic robots.