Abstract:
Manual collection of environmental data over a large area can be a time-consuming, costly, and even dangerous process, making it a perfect candidate for automation with mobile robots. Despite this clear suitability and numerous advances in robotics resulting in decreased costs, improved reliability, and increased ease of use, the problem of powering autonomous robots has proved to be an effective deterrent to their widespread use in the field. Although many survey scenarios involve domains with ample ambient energy present in the form of winds or currents that could be exploited by a robot operating given an appropriate strategy, past path planning research has neglected the study of energy-efficient methods in these domains, in lieu of continued pursuit of time- and length-optimal planning algorithms. Furthermore, much of the limited work addressing this topic relies on prior knowledge of the energy distribution within the domain, which can be particularly difficult and expensive to determine, especially when moving fluids are involved. In this thesis we address the problem of planning energy-efficient paths that exploit ambient energy in the absence of complete a priori knowledge of the domain.

Although work on energy-efficient planning continues, the methods developed consistently rely on a priori models of the vehicle or environment to achieve energy savings. This gap in research is particularly stark when energy-efficient coverage path planning is considered; a significant portion of the past work on this problem makes use of vehicle dynamics models and generally results in coverage plans that optimize the number of turns or the velocity along the path, with just a few studies considering the harvest of ambient energy during coverage execution. This thesis investigates the development of coverage planning techniques that integrate the gathering of highly practical domain knowledge with its exploitation to achieve autonomous energy-efficient information gathering. To this end we improve upon existing LSPIV current measurement methods and contribute a novel constraint-based coverage path planner, which given even a few fuzzy domain energy constraints and incomplete domain knowledge, is believed to produce energy-efficient coverage plans that will outperform plans produced by traditional methods. The addition of information gain constraints can be used to bias the vehicle towards exploration to acquire additional domain knowledge that may further improve energy-efficiency, particularly when initial domain knowledge is limited.

The particular motivating application behind this work is the dense mapping of environmental parameters in riverine environments using autonomous surface vehicles (ASVs) while exploiting evolving surface current knowledge to improve energy-efficiency throughout the process. To address this problem, we apply our coverage planner to compute complete coverage strategies around energy and information gain constraints provided by our enhanced LSPIV surface current measurement system. In order to motivate and validate this work, we describe and present results from its application to a scenario where an ASV is deployed to survey the bathymetry in a section of river using an energy-efficient coverage strategy, which is initially computed with incomplete surface current data and later improved by opportunistic deviation from the initial plan.

More Information

Thesis Committee Members:
John Dolan, Co-chair
Paul Scerri, Co-chair
George Kantor
Mel Siegel
Jordi Albó, La Salle University