Publications

A classifier that provides a technical diagnosis of complex equipment is proposed. Predicting the remaining life of the equipment is implemented to determine the time of diagnosis and inclusion of equipment in the repair program. The procedure for forming a repair program based on the forecast value of the remaining resource and financial losses associated with equipment failure is described.

Temperature monitoring during thermal therapies is used to regulate the amount of heat distributed to the cancerous tissue and therefore improve the clinical outcome of the oncological treatment. For the development of a hybrid hyperthermia system with non-invasive temperature monitoring by means of ultra-wideband (UWB) imaging, the optimal configuration of sensing antennas and heating applicators has to be investigated. In this paper we present the results of numerical simulations of several possible antenna arrangements, which are then validated by experiments with the radar system. The performance of each channel configuration was analyzed and benefits for the different clinical scenarios were specified.

In recent years, deep neural networks (DNNs) have accomplished impressive success in various applications, including autonomous driving perception tasks. However, current deep neural networks are easily deceived by adversarial attacks. This vulnerability raises significant - concerns, particularly in safety-critical applications. As a result, research into attacking and defending DNNs has gained much coverage. In this work, detailed adversarial attacks are applied on a diverse multi-task visual perception deep network across distance estimation, semantic segmentation, - motion detection, and object detection. The experiments consider both white and black box attacks for targeted and un-targeted cases, while attacking a task and inspecting the effect on all others, in addition to inspecting the effect of applying a simple defense method. We conclude this paper - by comparing and discussing the experimental results, proposing insights and future work. The visualizations of the attacks are available at https://youtu.be/6AixN90budY.

Over the last decades, automobiles have developed from predominantly mechanical machines to driving computers, consisting of a large number of sensors, actuators, and electronic control units that use various types of communication busses to form large and complex cyber-physical systems which provide a variety of comfort and safety features for the driver. Such E/E-systems (electric/electronic systems) do not only require high availability and reliability, but also overall efficient architectures and topologies. Because of the already high (and constantly increasing) level of complexity of such systems, the evaluation and optimization of their corresponding architectures has become a big challenge. The need for evaluation and optimization methods that are capable of abstracting the systems' complexity is evident for the automotive industry. In general, when it comes to designing and developing hard- and software systems, the paradigm of model-driven engineering emphasizes and supports the measures of abstraction, flexibility, and reusability to be able to grasp the complexity of modern systems. This paper presents a model-driven application for evaluating and optimizing automotive E/E-architectures as part of an integrated, model-based toolchain, developed with the Eclipse Modeling Framework.