Antoni Grau

By Edmundo Guerra, Yolanda Bolea, Rodrigo Munguia and Antoni Grau

This chapter describes an approach to improve the feature initialization process in the delayed inverse-depth feature initialization monocular Simultaneous Localisation and Mapping (SLAM), using data provided by a robot’s camera plus an additional monocular sensor deployed in the headwear of the human component in a human-robot collaborative exploratory team. The robot and the human deploy a set of sensors that once combined provides the data required to localize the secondary camera worn by the human. The approach and its implementation are described along with experimental results demonstrating its performance. A discussion on the usual sensors within the robotics field, especially in SLAM, provides background to the advantages and capabilities of the system implemented in this research.

Part of the book: Recent Advances in Robotic Systems

By Rodrigo Munguia, Edmundo Guerra, Sarquis Urzua, Yolanda Bolea and Antoni Grau

This chapter describes the development of an autonomous fluid sampling system for outdoor facilities, and the localization solution to be used. The automated sampling system will be based on collaborative robotics, with a team of a UAV and a UGV platform travelling through a plant to collect water samples. The architecture of the system is described, as well as the hardware present in the UAV and the different software frameworks used. A visual simultaneous localization and mapping (SLAM) technique is proposed to deal with the localization problem, based on authors’ previous works, including several innovations: a new method to initialize the scale using unreliable global positioning system (GPS) measurements, integration of attitude and heading reference system (AHRS) measurements into the recursive state estimation, and a new technique to track features during the delayed feature initialization process. These procedures greatly enhance the robustness and usability of the SLAM technique as they remove the requirement of assisted scale initialization, and they reduce the computational effort to initialize features. To conclude, results from experiments performed with simulated data and real data captured with a prototype UAV are presented and discussed.

Part of the book: Aerial Robots

By Edmundo Guerra, Jordi Palacin, Zhuping Wang and Antoni Grau

Robust perception is generally produced through complex multimodal perception pipelines, but these kinds of methods are unsuitable for autonomous UAV deployment, given the restriction found on the platforms. This chapter describes developments and experimental results produced to develop new deep learning (DL) solutions for industrial perception problems. An earlier solution combining camera, LiDAR, GPS, and IMU sensors to produce high rate, accurate, robust detection, and positioning of pipes in industrial environments is to be replaced by a single camera computationally lightweight convolutional neural network (CNN) perception technique. In order to develop DL solutions, large image datasets with ground truth labels are required, so the previous multimodal technique is modified to be used to capture and label datasets. The labeling method developed automatically computes the labels when possible for the images captured with the UAV platform. To validate the automated dataset generator, a dataset is produced and used to train a lightweight AlexNet-based full convolutional network (FCN). To produce a comparison point, a weakened version of the multimodal approach—without using prior data—is evaluated with the same DL-based metrics.

Part of the book: Industrial Robotics