Christoph Steup

Assessing neural networks for sensor fault detection

The idea of “smart sensing” includes a permanent monitoring and evaluation of sensor data related to possible measurement faults. This concept requires a fault detection chain covering all relevant fault types of a specific sensor. Additionally, the fault detection components have to provide a high precision in order to generate a reliable quality indicator. Due to the large spectrum of sensor faults and their specific characteristics these goals are difficult to meet and error prone. The developer manually determines the specific sensor characteristics, indicates a set of detection methods, adjusts parameters and evaluates the composition. In this paper we exploit neural-network approaches in order to provide a general solution covering typical sensor faults and to replace complex sets of individual detection methods. For this purpose, we identify an appropriate set of fault relevant features in a first step. Secondly, we determine a generic neural-network structure and learning strategy adaptable for detecting multiple fault types. Afterwards the approach is applied on a common used sensor system and evaluated with deterministic fault injections.

Bridging Physical and Digital Traffic System Simulations with the Gulliver Test-Bed

We propose a cyber-physical platform that combines road traffic simulation, network simulation, and physically simulated vehicles to facilitate extensive testing on various levels of vehicular systems. Our design integrates physical and digital vehicle simulation into a common development and testing environment. This paper describes the platform design and presents prototypical implementations that use Simulator of Urban Mobility (SUMO), TinyOS Simulator (TOSSIM), a 3D sensor simulation environment, and a test-bed of miniature vehicles called Gulliver. As a prototypical implementation, we demonstrate the development of cooperative applications, and by that we achieve: (a) a cyber-physical system that provides a common environment for physically and digitally simulated vehicles, (b) a platform to interface communication between physically and digitally simulated vehicles, and (c) the ability to tailor testing scenarios in which some system components are simulated digitally and some physically.

Online evaluation of manipulation tasks for mobile robots in Industry 4.0 scenarios

The concepts of “Industry 4.0” are founded on Cyber-Physical Systems (CPS) that interact flexibly with each other. The system is no longer an individual robot equipped with a predefined set of sensors, drives and manipulators. Furthermore, it represents a task-specific selection of all available CPS in a certain area. This adaptive composition provides flexible handling of varying environmental conditions and stabilizes the perception quality related to the current task requests. But the permanent adaptation includes a number of challenging tasks - the selection and adjustment of the involved CPS has to be done on run-time now. In this paper we propose a new approach for analyzing CPS configurations for manipulation tasks. If a mobile robot has to handle an object, the algorithm explores all available actuators and sensors. The proposed concept applies a graph-based model to define the individual failures of each component and the geometrical links in between. Based on this representation, all suitable combinations are determined, the expected failure level is calculated and compared in relation to the requested handling precision.

Technical evaluation of the Carolo-Cup 2014 - A competition for self-driving miniature cars

The Carolo-Cup competition conducted for the eighth time this year, is an international student competition focusing on autonomous driving scenarios implemented on 1:10 scale car models. Three practical sub-competitions have to be realized in this context and represent a complex, interdisciplinary challenge. Hence, students have to cope with all core topics like mechanical development, electronic design, and programming as addressed usually by robotic applications. In this paper we introduce the competition challenges in detail and evaluate the results of all 13 participating teams from the 2014 competition. For this purpose, we analyze technical as well as non-technical configurations of each student group and derive best practices, lessons learned, and criteria as a precondition for a successful participation. Due to the comprehensive orientation of the Carolo-Cup, this knowledge can be applied on comparable projects and related competitions as well.

Design and implementation of a small size robot localization system

The position of a mobile robot can be determined very precise today. A large number of high level sensor systems in combination with processing algorithms running on powerful hardware can provide (nearly) every requirement. But in case of limited financial and computational resources new approaches beside laser scanners and stereo cameras are necessary. In this paper we propose a localization system motivated by RoboCup Junior competitions but also suitable for similar applications. The paper describes the theoretical investigation of a multi sensor system, based on this the implementation with real hardware and its validation.

Energy Aware Particle Swarm Optimization as search mechanism for aerial micro-robots

This paper presents the Energy Aware PSO (EAPSO) as a search mechanism for aerial micro-robots with limited energy capacity. The proposed model is an extension of the search concept of Particle Swarm Optimization (PSO) that additionally considers the energy levels of the individuals for an efficient movement. One major contribution of this paper is that the energy efficiency results from a multi-criteria decision making process performed by the individuals. The energy consumption model in EAPSO is adapted from a real hardware scenario and has been tested on three known landscapes which are very similar to search terrains by the aerial micro-robots. The results show that EAPSO can reduce the total energy consumption of the swarm with negligible degradation of the search results.

On the security of international data exchange services for e-governance systems

The effects of globalisation and information welfare combined with the increasing mobility of individuals lead to a number of challenges to modern states. In order to guarantee a smooth, secure, uninterrupted organisational flow, governments and their subsidiaries need to cooperate and exchange data on individuals and organisations across national borders. However, insufficiently secured communication of such data imposes security threats which may endanger the individual’s privacy. Currently, several states within Europe develop and operate e-governance systems. These are primarily designed to allow the exchange of data within the institutions of one state. However, examples such as the Estonian e-governance backbone system X-Road strive towards an EU-wide expansion. Technical solutions for the transnational exchange of data between governmental institutions are an unavoidable part of the future of cyberspace. Despite the fact that EU specifications exist, the discrepancy between specification and implementation becomes immanent. This article explores some of the general aspects of the design of secure transnational data exchange frameworks. A comparative analysis of existing e-governance systems within Europe is given based on defined security aspects. It is explored how decisions made in the design may affect the security of the underlying network and its components. The challenges of transnational data exchange frameworks are discussed.

Phase optimization for control/fusion applications in dynamically composed sensor networks

The variable acquisition of distributed sensor perception promises an effective utilization of these data across different applications. Multiple observations enable an increase in the quality of the system output. However, the dynamic composition disables all off-line optimization approaches, especially for sensor-application-scheduling. This paper addresses the need for an online adjustment of periodically working sensors and fusion/control applications. Based on a number of common goals - e.g., minimization of the variance of sensor data or the age of data sets - we deduce different metrics. For one aspect, the number of input counts, we propose a mathematical description and apply related optimizations. We use an an example analysis to illustrate further research goals.

Adaptive environment perception in cyber-physical systems

The concept of an adaptive acquisition of environment data in distributed scenarios promises a number of benefits. If an application aggregates and uses all available sensing information in an intelligent environment it may provide a higher precision and an increased fault-tolerance. Unfortunately, the application developer has to cope with a number of additional challenges compared to static sensor evaluation. It is not possible to generate an optimized sensor application schedule for a dynamic system at design-time. Due to the adaptive selection process, this has to be executed at runtime. In this paper we propose a general approach for this problem based on a two-level analysis. The first level compares sensor parameter sets (periods, offsets, delays) and application requirements (number of measurements, quality) based on a worst/best case analysis. If a more precise evaluation is necessary, the second level needs to be started. This one considers additional, situation-specific properties like phase shift of sensor periods, communication delays and jitter. At the end, it provides an online optimization of common goals e.g., minimization of the age of data and a constant number of input counts.