Landing safety is a challenge heavily engaging the research community recently, due to the increasing interest in applications availed by aerial vehicles. In this paper, we propose a landing safety pipeline based on state of the art object detectors and OctoMap. First, a point cloud of surface obstacles is generated, which is then inserted in an OctoMap. The unoccupied areas are identified, thus resulting to a list of safe landing points. Due to the low inference time achieved by state of the art object detectors and the efficient point cloud manipulation using OctoMap, it is feasible for our approach to deploy on low-weight embedded systems. The proposed pipeline has been evaluated in many simulation scenarios, varying in people density, number, and movement. Simulations were executed with an Nvidia Jetson Nano in the loop to confirm the pipeline's performance and robustness in a low computing power hardware. The experiments yielded promising results with a 95% success rate.

In numerous real-world situations, acquiring extensive sets of labeled data poses a formidable challenge. This study introduces an innovative approach for enhancing weakly supervised learning (WSL) through anomaly-informed weighted training (WT). The method not only is tested in diverse benchmark datasets such as CIFAR-10 and Fashion-MNIST by simulating a binary classification problem with only prior knowledge of some samples belonging to one class but also is applied to a real-world scenario which specifically aims to detect surface defects, like pitting, on ball screw surfaces. The proposed method leverages anomaly detection techniques to refine the training processes in a WSL setting, thereby effectively addressing the challenge of limited labeled data availability. The results demonstrated significant enhancements in the classification metrics, showing the potential of this method for industrial applications.

The explosion of the digitisation of traditional industrial processes and procedures is consolidating a positive impact on modern society by offering a critical contribution to its economic development. In particular, the dairy sector consists of various processes, which are very demanding and thorough. Therefore, it is crucial to leverage modern automation tools and through-engineering solutions to increase efficiency and continuously meet challenging standards. Towards this end, an intelligent algorithm based on machine vision and artificial intelligence, which identifies dairy products within production lines, is presented in this work. Furthermore, to train and validate the model, the novel YogDATA dataset was created that includes yoghurt cups captured within a production line. Specifically, we evaluate two deep learning models (Mask R-CNN and YOLO v5.0) to recognise and detect each yoghurt cup in a production line to automate product packaging processes. According to our results, the performance precision of the two models is quite similar, fulfilling the identification requirement of Dairy 4.0 production systems.

In many real-world scenarios, obtaining large amounts of labeled data can be a daunting task. Weakly supervised learning techniques have gained significant attention in recent years as an alternative to traditional supervised learning, as they enable training models using only a limited amount of labeled data. In this paper, the performance of a weakly supervised classifier to its fully supervised counterpart is compared on the task of defect detection. Experiments are conducted on a dataset of images containing defects, and evaluate the two classifiers based on their accuracy, precision, and recall. Our results show that the weakly supervised classifier achieves comparable performance to the supervised classifier, while requiring significantly less labeled data.

As unmanned aerial vehicles (UAVs) are becoming increasingly popular, their safety features turn out to be vital, not only for protecting the aircraft itself, but also people and property on the ground. Emergency landings are fairly common fail-safe scenarios, where a UAV is compelled to land at the closest safe area. This can be achieved by detecting landing obstacles on the ground, as in our previously safety landing work. Building upon that, we propose a multi-criteria decision module to quantify the danger each obstacle poses to landing operation. We evaluated our system using Gazebo Simulator in scenarios consisting of both moving and static people. The results were promising, showing the robustness of our pipeline in different test-scenarios.

The digitalization of traditional industrial processes has profoundly influenced every step of the manufacturing value chain during the past two decades, having as its main goal to achieve zero-defected products. Moreover, since dairy production is at the heart of food industry, it is critical to leverage innovative technologies to increase their efficiency and continuously meet the demanding standards from the farm level to market and reduce the amount of waste. Towards this end, we propose a Dairy 4.0 architecture capable of utilising information to detect and prevent flaws to the final dairy products. The architecture layers are based on machine vision and the digital twins technologies, while it respects the zero defect manufacturing (ZDM) approach. The proposed frameworks is structured on a four layer architecture: (i) the physical layer, which consists of dairy farming, dairy production, and dairy storage and logistics, (ii) the acquisition layer that is responsible for collecting contextual information, (iii) the digital twin layer which uses data from the vision system and the physical system to anticipate future occurrences, and finally (iv) the ZDM layer, which functions as an orchestrator and binding agent for all the processed data.

During Space Weather events, Single Event Effects (SEEs) caused by energetic ions on spacecraft electronics can severely damage satellites in space. At geostationary orbit (GEO), Space Weather events are continuously monitored, and spacecraft launched to GEO are equipped with radiation hardened electronics that undergo rigorous radiation testing. In the present NewSpace era of Low Earth Orbit (LEO) mega-constellations, tens of thousands of planned spacecraft are shifting usage of space to LEO, however, unlike GEO, Space Weather prediction and monitoring at LEO is still inadequate. At the same time, most LEO spacecraft have minimal radiation hardness requirements so as to minimize size and cost, making them vulnerable to Space Weather effects, which are underestimated but ever-present at LEO. Furthermore, modern semiconductor-based devices keep shrinking in size, which makes them vulnerable to ionizing particles of even lower energies than in the past. As part of the ReTiMo satellite mission concept, a novel, low power, radiation-hardened SEE detection system is designed for real-time reporting of SEE rates. In this paper we will present the concept of a System on Chip (SoC) for the on-board, calibrated SEE rate detection and reporting that can be used for spacecraft housekeeping and operations. This SoC consists of a radiation-hardened memory controller and memory data evaluator and of a bank of Single Event Upset (SEU) sensitive memories with programmable Linear Energy Transfer (LET) onset. In addition, the SoC can interface to external memories and analog components in order to further measure SEUs, Single Event Transients (SETs) and Single Event Latch Ups (SELs).

Aiming to recognize familiar places through the camera measurements during a robot’s autonomous mission, visual loop-closure pipelines are developed for navigation frameworks. This is because the main objective for any simultaneous localization and mapping (SLAM) system is its consistent map generation. However, methods based on active vision tend to attract the researchers’ attention mainly due to their offered possibilities. This paper proposes a BK-tree structure for a visual loop-closure pipeline’s generated database when active vision is adopted. This way, we address the drawback of scalability in terms of timing occurring when querying the map for similar locations while high performances and the online nature of the system are maintained. The proposed method is built upon our previous work for visual place recognition, that is, the incremental bag-of-tracked-words. The proposed technique is evaluated on two publicly-available image-sequences. The one is recorded via an unmanned aerial vehicle (UAV) and selected due to its active vision characteristics, while the second is registered via a car; still, it is chosen as it is among the most extended datasets in visual loop-closure detection. Our experiments on an entry-level system show high recall scores for each evaluated environment and response time that satisfies real-time constraints.

Unmanned aerial vehicles (UAVs) are at the fore-front of this century's technological shift, becoming ubiquitous in research and market areas. Similarly, nowadays, 3D printing is a fast-emerging, widely used technology that allows individ-uals to design prototypes that fulfil their needs. This paper presents an autonomous UAV designed and implemented to be fully modular and 3D printable. Furthermore, suitable areas for landing are recognized using a lightweight deep learning architecture while a Gazebo model for simulation purposes is also given to the research community. Finally, its fly and surface recognition processes are evaluated exhaustively in real-world and simulation scenarios.

This article presents the design, manufacturing and test results of an on-chip CMOS oscillator, using a ring-oscillator, VCO based architecture. The oscillator generates a configurable square waveform clock signal to be used internally or externally to the IC that integrates it, with very low area (320 transistors, 112x148 mu m) and power overhead (975 mu W). The oscillator is integrated in a mixed signal IC which has been qualified for space applications, at a commercial 250nm process. It enables the standalone operation of the IC without external oscillator and gives the possibility to clock other components and systems. In addition, it reduces the noise interference at PCB and chip level, optimising the performance of sensitive analogue parts. It was validated by radiation tests according to ESA standards' procedures that the oscillator's functionality and characteristics do not deteriorate with TID levels up to 1Mrad. This approach can be easily adjusted to a wide range of frequencies, while significantly reducing the cost and power budget of space qualified systems with small design effort trade-off.

The CAN bus standard is widely used in the space industry to interconnect subsystems. Its main advantage is the performance in terms of reliability, due to its sophisticated error handling mechanisms and electrical noise robustness. This paper presents a novel approach to the design of a CAN interface (including the CAN Controller—data link layer—and the CANopen—application layer) optimized for radiation-hardness, power consumption and other qualities desired for space systems electronics. The CAN interface was fabricated on a commercial 0.25 μm technology, integrated in a SoC with analog and digital subsystems. Tests have shown radiation hardness above 1 Mrad and no SEEs up to 57 MeV, with 3 mW power consumption, and a −55 to +125 °C operating temperature range, features greatly optimized in comparison to current implementations in the space industry.

This paper presents an innovative two-processor computer architecture, developed for the data processing unit (DPU) of the Magnetospheric IMaging Instrument (MIMI), on-board the Cassini spacecraft mission to Saturn. The main advantages of this architecture are its high performance and reliability, and its intelligence. The high performance is justified by the following: 1) optimum combination of two powerful Harris RTX 2010 processors; 2) adoption of two independent main bus structures used for the communication of the processors with the various instrument interfaces and subsystems; 3) adoption of two additional local buses on each processor board used to speed the on-board operations of the processors; 4) high speed interprocessor communication port. The high reliability is justified by the following: 1) simplicity of hardware/software structures; 2) fault tolerance capabilities; 3) capability for on-flight hardware/software reconfiguration by ground command. Moreover, the on-board intelligence is justified by the following: 1) sophisticated fault protection, data handling, and instrument control software; 2) intelligent interfaces [implemented using held programmable gate arrays (FPGAs)]; 3) capability for autonomous on-flight hardware/software reconfiguration in case of an unrecoverable failure in one processor. The advantages of this architecture make it the best choice for the DPU of the complex, sophisticated scientific MIMI instrument, compared to the traditional master-slave (low reliability-single point failure) and common shared bus (low performance, hardware and software complexity) architectures.