Evaluation of Dead Reckoning Navigation for Underwater Drones using ROS

Abstract Hurricane modelling research requires oceanic measurements taken before and during hurricanes, and improving these models requires more data points. Increasing the number of oceanic sensors available can facilitate a greater understanding of oceanic conditions before and after hurricanes. The motivation for this project is the creation of an Autonomous Underwater Vehicle (AUV) for the National Oceanic and Atmospheric Administration's (NOAA) 2021 Ocean Observing Prize, a competition aimed at developing innovative oceanic monitoring platforms. In order to facilitate the collection of data relevant to hurricane modeling, the proposed AUV must operate in open ocean, presenting challenges to navigation. The deep ocean environment prevents the use of the ocean floor as a reference for oceanic navigation due to the relatively shallow depth at which the AUV will operate. Instead, a dead reckoning approach is used to determine heading and distance between navigation waypoints based on an inertial navigation system (INS). The INS estimates the drones position through a combination of collected magnetometer data and orientation derived from accelerometer and gyroscope data. While the INS is capable of operating under the surface of the water, the error accumulation makes it impractical for tracking position over the full mission duration. This error is accounted for with continual correction by the drone resurfacing and reading its current GPS position. Through a dead-reckoning approach the system is capable of estimating its position even while under the water's surface. The navigation system presented was developed for a Robot Operating System (ROS) control platform, enabling performance evaluation within the UUV Simulator Gazebo package. Through the simulation of sensor inputs and ocean conditions, we developed and tested the navigation algorithm prior to the assembly of a prototype AUV. The structure of ROS also enabled algorithm verification with a car-like ground vehicle without significant modifications. The final algorithm was tested on a small scale AUV utilizing off-the-shelf sensors and a Raspberry Pi 4 at its core. This paper presents the navigation system developed, the simulation setup and results from testing various dead-reckoning approaches, and the experimentation evaluation of the navigation algorithm using a small scale AUV implemented using mostly COTS components. . Experimental performance evaluation showed an acceptable match between simulation and the experiments. This project was part of the 2021 installment of NSF Research experience for Undergraduates (REU) program in the Electrical and Computer Engineering Department at Oakland University.