Environmental Sampling with the Boustrophedon Decomposition Algorithm

The automation of data collection via mobile robots holds promise for increasing the efficacy of environmental investigations, but requires the system to autonomously determine how to sample the environment while avoiding obstacles. Existing methods such as the boustrophedon decomposition algorithm enable complete coverage of the environment to a specified resolution, yet in many cases sampling at the resolution of the distribution would yield long paths with an infeasible number of measurements. Downsampling these paths can result in feasible plans at the expense of distribution estimation accuracy. This work explores this tradeoff between distribution accuracy and path length for the boustrophedon decomposition algorithm. We quantify algorithm performance by computing metrics for accuracy and path length in a Monte-Carlo simulation across a distribution of environments. We highlight conditions where one objective should be prioritized over the other and propose a modification to the algorithm to improve its effectiveness by sampling more uniformly. These results demonstrate how intelligent deployment of the boustrophedon algorithm can effectively guide autonomous environmental sampling.