Abstract
The growing dependence of modern-day societies on electricity increases the importance of effective monitoring and maintenance of power lines. Endowing UAVs with the appropriate sensors for inspecting power lines, the costs and risks associated with the traditional foot patrol and helicopter-based inspections can be reduced. However, this implies the development of algorithms to make the inspection process reliable and autonomous. Visual methods are usually applied to locate the power lines and their components, but poor light conditions or a background rich in edges may compromise their results. To overcome those limitations, we propose to address the problem of power line detection and modeling based on LiDAR. A novel approach to the power line detection was developed, the PL2DM -Power Line LiDAR-based Detection and Modeling. It is based in a scan-by-scan adaptive neighbor minimalist comparison for all the points in a point cloud. The power line final model is obtained by matching and grouping several line segments, using their collinearity properties. Horizontally, the power lines are modeled as a straight line, and vertically as a catenary curve. The algorithm was validated with a real dataset, showing promising results both in terms of outputs and processing time, adding real-time object-based perception capabilities for other layers of processing.

Abstract
3D path planning with unmanned aerial vehicles in search and rescue scenarios is an important research area, due to the ability to explore damage areas that could be inaccessible for vehicles like ground robots. This paper presents two innovative real-time path planning algorithms based on PRM (Probabilistic Road Map) able to be implemented in UAV’s denoted by Grid Path Planning Roadmap Planning (GPRM) and the Particle Probabilistic Roadmap (PPRM). With the requirement of being implemented in a real search and rescue scenario like the EuRathlon competition, the GPRM method will produce a roadmap building step with obstacles inside a predefined grid while PPRM will follow a different approach by introducing an associated probability to each computed path in order to support the next sampling step path planning iteration. Both methods were evaluated and compared with the well known 3D path planning PRM in a search and rescue earthquake simulation environment developed in MORSE (Modular Open Robots Simulation Engine). © 2018 by World Scientific Publishing Co. Pte. Ltd.

Abstract
This paper address the use of deep learning approaches for visual based navigation in confined underwater environments. State-of-the-art algorithms have shown the tremendous potential deep learning architectures can have for visual navigation implementations, though they are still mostly outperformed by classical feature-based techniques. In this work, we apply current state-of-the-art deep learning methods for visual-based robot navigation to the more challenging underwater environment, providing both an underwater visual dataset acquired in real operational mission scenarios and an assessment of state-of-the-art algorithms on the underwater context. We extend current work by proposing a novel pose optimization architecture for the purpose of correcting visual odometry estimate drift using a Visual-Inertial fusion network, consisted of a neural network architecture anchored on an Inertial supervision learning scheme. Our Visual-Inertial Fusion Network was shown to improve results an average of 50% for trajectory estimates, also producing more visually consistent trajectory estimates for both our underwater application scenarios.

Abstract
Zooplankton plays a key -role on Earth's ecosystem, emerging in the oceans and rivers in great quantities and diversity, making it an important and rather common topic on scientific studies. Given the numbers of different species it is not only necessary to study their numbers but also their classification. In this paper a possible solution for the zooplankton in situ detection and classification problem in real-time is proposed using a portable deep learning approach based on Convolutional Neural Networks deployed on INESC TEC's MarinEye system. For detection a Single Shot Detection model with MobileNet was used, and ZooplanktoNet for classification. System portability is guaranteed with the use of MovidiusTMNeural Compute Stick as the deep learning motor.

Abstract
For the context of a mobile tracking system, an underwater acoustic positioning system was developed, using three hydrophones to compute the direction of an acoustic source relative to an Autonomous Surface Vehicle (ASV). The paper presents an algorithm for the Direction of Arrival (DoA) of an acoustic source, which allows to estimate its position. Preliminary results will be shown in this paper relative to the detection and identification (ID) of the acoustic sources, as well as an analysis of the proposed algorithm. The solution allows the position estimation of an acoustic source, which can be used in tracking solutions. The system can be applied in an ASV or fixed buoys, as long as the baseline's hydrophones are at equal angular distances. The main objective is to track targets with the DoA algorithm as well to estimate their position, improving what was done in [1].

Abstract
Teaching robotics based on challenge of our daily lives is always more motivating for students and teachers. Several competitions of self-driving have emerged recently, challenging students and researchers to develop solutions addressing the autonomous driving systems. The Portuguese Festival Nacional de Rob ' otica (FNR) Autonomous Driving Competition is one of those examples. Even though the competition is an exciting challenger, it requires the development of real robots, which implies several limitations that may discourage the students and compromise a fluid teaching process. The simulation can contribute to overcome this limitation and can assume an important role as a tool, providing an effortless and costless solution, allowing students and researchers to keep their focus on the main issues. This paper presents a simulation environment for FNR, providing an overall framework able to support the exploration of robotics topics like perception, navigation, data fusion and deep learning based on the autonomous driving competition.