Abstract
This paper addresses Visual Odometry (VO) estimation in challenging underwater scenarios. Robot visual-based navigation faces several additional difficulties in the underwater context, which severely hinder both its robustness and the possibility for persistent autonomy in underwater mobile robots using visual perception capabilities. In this work, some of the most renown VO and Visual Simultaneous Localization and Mapping (v-SLAM) frameworks are tested on underwater complex environments, assessing the extent to which they are able to perform accurately and reliably on robotic operational mission scenarios. The fundamental issue of precision, reliability and robustness to multiple different operational scenarios, coupled with the rise in predominance of Deep Learning architectures in several Computer Vision application domains, has prompted a great a volume of recent research concerning Deep Learning architectures tailored for visual odometry estimation. In this work, the performance and accuracy of Deep Learning methods on the underwater context is also benchmarked and compared to classical methods. Additionally, an extension of current work is proposed, in the form of a visual-inertial sensor fusion network aimed at correcting visual odometry estimate drift. Anchored on a inertial supervision learning scheme, our network managed to improve upon trajectory estimates, producing both metrically better estimates as well as more visually consistent trajectory shape mimicking.

Abstract
The big growth of electrical demand by the countries required larger and more complex power systems, which have led to a greater need for monitoring and maintenance of these systems. To overcome this problem, UAVs equipped with appropriated sensors have emerged, allowing the reduction of the costs and risks when compared with traditional methods. The development of UAVs together with the great advance of the deep learning technologies, more precisely in the detection of objects, allowed to increase the level of automation in the process of inspection. This work presents an electrical assets monitoring system for detection of insulators and structures (poles and pylons) from images captured through a UAV. The proposed detection system is based on lightweight Convolutional Neural Networks and it is able to run on a portable device, aiming for a low cost, accurate and modular system, capable of running in real time.

Abstract
The global navigation satellite system (GNSS) constitutes an effective and affordable solution to the outdoor positioning problem. When combined with precise positioning techniques, such as the real time kinematic (RTK), centimeter-level positioning accuracy becomes a reality. Such performance is suitable for a whole new range of demanding applications, including high-accuracy field robotics operations. The RTKRCV, part of the RTKLIB package, is one of the most popular open-source solutions for real-time GNSS precise positioning. Yet the lack of integration with the robot operating system (ROS), constitutes a limitation on its adoption by the robotics community. This article addresses this limitation, reporting a new implementation which brings the RTKRCV capabilities into ROS. New features, including ROS publishing and control over a ROS service, were introduced seamlessly, to ensure full compatibility with all original options. Additionally, a new observation synchronization scheme improves solution consistency, particularly relevant for the moving-baseline positioning mode. Real application examples are presented to demonstrate the advantages of our rtkrcv_ros package. For community benefit, the software was released as an open-source package.

Abstract
Emerging opportunities in the exploration of inland water bodies, such as underwater mining of flooded open pit mines, require accurate real-time positioning of multiple underwater assets. In the mining operation scenarios, operational requirements deny the application of standard acoustic positioning techniques, posing additional challenges to the localization problem. This paper presents a novel underwater localization solution, implemented for the VAMOS! project, based on the combination of raw measurements from a short baseline (SBL) array and an inverted ultrashort baseline (iUSBL). An extended Kalman filter (EKF), fusing IMU raw measurements, pressure observations, SBL ranges, and USBL directional angles, estimates the localization of an underwater mining vehicle in 6DOF. Sensor bias and the speed of sound in the water are estimated indirectly by the filter. Moreover, in order to discard acoustic outliers, due to multipath reflections in such a confined and cluttered space, a data association layer and a dynamic SBL master selection heuristic were implemented. To demonstrate the advantage of this new technique, results obtained in the field, during the VAMOS! underwater mining field trials, are presented and discussed.

Abstract
This paper addresses Three-Dimensional (3D) reconstruction of historical sites with an Unmanned Aerial Vehicle (UAV), combining the information from a visible spectrum camera with a Light Detection and Ranging (LiDAR). The developed solution was validated in two sites located in Monastery of Tibães (Braga, NW Portugal), within the scope of MineHeritage project, which intends to reach society on the importance of raw materials through a historical approach. The outputs obtained from the datasets, resulted in a successfully 3D reconstruction of the two studied sites on the Monastery. Although the research is still ongoing on this topic, this paper is a starting point and an important contribution to this field and this type of scenarios. © CLAWAR Association Ltd.

Abstract
A novel dead reckoning algorithm conceived for localization of small inspection rail vehicles in Global Navigation Satellite System (GNSS) denied environments is presented. This work focus on simplifying the rail vehicle localization task, taking into account restrictions on movement imposed by the railroad tracks. Considering that dead reckoning techniques accumulate errors over time, leading to increasing global uncertainty, a method was designed to correct the estimates and also smooth trajectory errors backwards in time, through visualization of global landmarks. Results show the effectiveness of this approach in reducing long-term position errors. The current document reports real railroad experiments, featuring a specially designed non-motorized mobile modeling vehicle.