Human-autonomy teaming for improved diver navigation

Submitted in partial fulfillment of the requirements for the degree of Master of Science at the Massachusetts Institute of Technology and the Woods Hole Oceanographic Institution February 2022.Diving operations are inherently complex due to navigation and communication limitations. Until recently, fixed-beacon acoustic localization techniques have served as the primary means of improving diver navigation. However, modern artificial intelligence and acoustic modem technologies have enabled accurate relative navigation methods between a diver and an autonomous vehicle. Human-robot collaboration takes advantage of each member’s strengths to create the most effective team. This concept proves especially advantageous within the ocean domain, where humans are naturally deficient navigators. Yet humans serve as the team’s creative spirit, offering the critical thinking and flexibility needed to succeed in an unpredictable and dynamic environment. Recent underwater human-robot cooperative navigation systems typically rely on autonomous surface vehicles (ASVs), specially designed underwater vehicles, or stereo cameras. This thesis proposes a diver navigation method exhibiting significantly improved accuracy over dead reckoning without relying on a surface presence, cameras, or fixed acoustic beacons. Specifically, we develop and evaluate the communication architecture and autonomous behaviors required to guide a diver to a target location using subsurface humanautonomous underwater vehicle (AUV) teaming with no requirement for ocean current data or exact diver speeds. By depending on acoustic communication and commercial AUV navigation capabilities, our method has increased accessibility, applicability, and robustness over former techniques. We utilize the Woods Hole Oceanographic Institution (WHOI) Micromodem 2’s twoway-travel-time (TWTT) capability to enable range-only single-beacon navigation between two kayaks serving as proxies for the diver and Remote Environmental Monitoring Units (REMUS) 100 AUV. During processing, a nonlinear least-squares (NLS) method, called incremental smoothing and mapping 2 (iSAM2), utilizes odometry and range measurements to provide real-time diver position estimates given unknown ocean currents. Field experiments demonstrate an average online endpoint error of 4.53 meters after transits four hundred meters long. Additionally, simulations test our method’s performance in more challenging situations than those experienced in the field. Overall, this research progresses the interoperability of divers and AUVs.The United States Navy funded my graduate education. The Office of Naval Research also partially supported this work under grant N00014-18-1-2832