DOES: A Deep Learning-Based Approach to Estimate Roll and Pitch at Sea

Fabiana Di Ciaccio,Paolo Russo,Salvatore Troisi

doi:10.1109/access.2022.3158971

Abstract

The use of Attitude and Heading Reference Systems (AHRS) for orientation estimation is now common practice in a wide range of applications, e.g., robotics and human motion tracking, aerial vehicles and aerospace, gaming and virtual reality, indoor pedestrian navigation and maritime navigation. The integration of the high-rate measurements can provide very accurate estimates, but these can suffer from errors accumulation due to the sensors drift over longer time scales. To overcome this issue, inertial sensors are typically combined with additional sensors and techniques. As an example, camera-based solutions have drawn a large attention by the community, thanks to their low-costs and easy hardware setup; moreover, impressive results have been demonstrated in the context of Deep Learning. This work presents the preliminary results obtained by DOES, a supportive Deep Learning method specifically designed for maritime navigation, which aims at improving the roll and pitch estimations obtained by common AHRS. DOES recovers these estimations through the analysis of the frames acquired by a low-cost camera pointing the horizon at sea. The training has been performed on the novel ROPIS dataset, presented in the context of this work, acquired using the FrameWO application developed for the scope. Promising results encourage to test other network backbones and to further expand the dataset, improving the accuracy of the results and the range of applications of the method as a valid support to visual-based odometry techniques.

Highlights

T HE pose estimation problem consists in estimating the position and orientation of a vehicle, device, human or robot with respect to a reference frame, through the use of different kinds of internal or external sensors
The Root Mean Square Error (RMSE) is generally higher than the Mean Absolute Error (MAE), and the greater is the difference between them, the greater will be the variance in the individual errors of the samples; if the RMSE is close to the MAE, all the errors are of the same magnitude
The Standard Deviation (STD) values of the three methods show that the results obtained by DOES are significantly more clustered than the others, meaning that they are closer to the mean value and as such can be considered more reliable

Summary

Introduction

T HE pose estimation problem consists in estimating the position and orientation of a vehicle, device, human or robot with respect to a reference frame, through the use of different kinds of internal or external sensors. In the last years the use of low-cost technologies is becoming widely spread in numerous applications: this means that the accuracy of the pose obtained by these systems can be affected by even more disturbing factors than the traditional high-performing methods In these circumstances, the development of accurate and reliable orientation estimation algorithms can still be considered a very challenging task, being at the basis of the localization process and of the consequent performances of the device employed for any specific task. The same goes for Unmanned Surface Vehicles (USVs), which are mainly employed in environmental monitoring, safety or navigation support and research operations In this case, a non accurate estimation of the orientation can severely compromise the ultimate success of the mission, especially when paired to low-cost sensors and poor GNSS support

Objectives

Methods

Results

Conclusion