IUI Logo

Automotive IUI

Die Arbeiten sind von neu nach alt sortiert.

Integration des Three-Vehicle Platoon Tests in OpenDS

Gleb Banas

Bachelor Thesis by Gleb Banas advised by Rafael Math and Christian Müller


A Framework for Designing, Conducting and Recording User Studies as an Extension of an Existing Driving Simulation Solution

Till Maurer

Bachelor Thesis by Till Maurer advised by Angela Mahr, Rafael Math and Christian Müller


Design and Implementation of a Graphical Driving Task Editor for the Carmina Driving Simulator


Master Thesis by Otávio HS Biasutti advised by Rafael Math and Christian Müller.

In this thesis a new conception for the definition of a driving task will be presented through studies made with the primal objective of implementing a piece of software for driving task creation. The reader might be familiarized with the existence of car driving simulators. Inside the virtual driving environment provided by such simulators the driver has assignments to complete in order to have his performance measured. Usually, there is a group of researchers behind this process who, in fact, designed the driving scenario and the tasks with the purpose of validating an initial hypothesis. The study that was made tried to analyze how such a task is designed, and how would a computer program enhance the process during which the original researcher's hypothesis develops into a proper driving task inside the car simulator.

The thesis starting point is the current car driving simulator which was developed by DFKI's Automotive Group. The goal was to provide a tool to create and edit driving task scenarios compatible with this existent simulator. The new implemented editor offers an environment that enables a researcher without any coding skills, to create a driving task entirely from scratch. In addition, to simplify the user's handling of the tool, there exists an underlying driving task representation model which is divided into three major layers. This division can improve the user‟s focus on a smaller piece of work and subsequently help in the creation of more complex driving tasks.

The Crossmodal Stroop Effect - Is the Detection of Blue Accelerated by Hearing "Blue"?


Bachelor Thesis by Tanja Schneeberger advised by Angela Mahr and Prof. Wentura

The aim of this study was to analyze the reason for the Crossmodal Semantic Priming Effect. We repeatedly found crossmodal Semantic Priming Effects for spoken words in object classification in former visual search experiments. Results denote crossmodal influences of spoken words (colors, object labels) on visual detection and classification of target objects. Results clearly show facilitated responses in congruent trials.

In incongruent trials, however, responses might be slowed compared with neutral trials. In order to find out, whether object detection in visual search is facilitated (e.g. by increased detection sensitivity), or whether response priming is the reason for the effects, we used an object absent/present-decision task and signal detection theory. Thereby, response priming cannot be responsible for congruency effects in this experiment. The results showed that congruent verbal color-cues (SOA = 100 ms) improved the detection of target colors (sensitivity parameter d’) within filler colors, in comparison with incongruent or neutral verbal cues, and also in comparison with silence. This confirms, that congruent speech primes can increase detection sensitivity and thus lead to faster responses. Transferred to the automotive context, this supports the assumption, that auditory speech hazard warnings can enhance their detection or might even lead to faster responses by drivers.

Object-Based Auditory-Visual Crossmodal Integration: Effects of Auditory Speech Primes on the Detection of Visual Icons


Bachelor Thesis by Verena Johann advised by Angela Mahr and Prof. Wentura


Development of a User-Adaptive UIML to Flash Renderer for the Automotive Domain


Master Thesis by Anand Ramkumar Shanmugham advised by Michael Feld, Angela Mahr and Christian Müller in collaboration with ZMMI, Kaiserslautern.


Physical & Spiritual Proximity: Linking Car2Car Communication with Online Social Networks


Master Thesis by Monika Mitrevska supervised by Sandro Castronovo and Christian Müller


A Driver Model for Exploring the Parameters Space of Human Subject Studies on the Example of Persuasive EcoDrive Assistants


Bachelor Thesis by Bernd Mechenbier advised by Jan Miksatko, Angela Mahr and Christian Müller

Summary: Developing interfaces to be used by humans is a time consuming and expensive task. This stems mainly from having to test in a highly tailored environment with human subjects, meaning extensive preparation in setup and design of the experiment to be done. The large space of parameters related to the experiment, concerning the functionality of the system to be evaluated as well as the evaluation of the results, intensifies this problem. Scoring systems, using positive as well as negative scores, weights for the scores and many other details can and have to be tuned and adjusted to gain insight into the recorded data and extract useful results. The higher the complexity of the system in question is, the larger the space of the parameters which can be adjusted. Floor and ceiling effects further complicate these adjustments.

To avoid such problems, a certain amount of pre-testing is performed to gain better understanding of how to set scores and weights and of how to design the whole experiment, to gain insight into the questions the experiment should deliver the answers for. Naturally this pre-testing may become complex if the tested system itself is complex and it may be necessary to have multiple iterations of pre-testing. Using human subjects in all these pre-tests is costly in terms of time, money and possibly other resources. It also depletes the pool of subjects who can be used for the final and real test where the experiment data is gained.

We think that by substituting a fitting software agent instead of the human subject it is possible to do cheaper and quicker testing. The agent can be made adaptive, to fit a variety of human subject types. Virtualizing the environment, e.g. running a simulation instead of testing with the real device, may also give the possibility to run the experiments in faster-than-real time. Thus it becomes possible to tune and adapt many parameters in a short amount of time and saving costs. In this particular case, the implemented system to be tested in the experiment is a software agent as well. Thus we can see this as a multiagent scenario. We used a scenario of a driver interacting with an Eco-Drive assistant to show the feasibility of our approach.

Incorporating Touch-Free 2D Microgestures into a Multimodal Interaction Concept for Drivers


Master Thesis by Praveen Chundi advised by Christoph Endres and Christian Müller


User-Adaptive Warning Systems: Online Assessment of Driving Performance and Driver Stress on the Example of Local Danger Alerts

Master Thesis by Donjeta Ibrahimi advised by Michael Feld and Christian Müller

With the introduction of network technologies inside the car, information on local dangers can be available to drivers and the system, long before they come into sight. In a highly cognitively demanding environment as the driving environment, the warnings issued are the only way to reach the driver and their contribution to safety is not questionable. With this valuable information at hand, providing action suggestions as ”Brake”, ”Change Lane” or ”Slow down” can help the drivers in the decision making process.

However, giving suggestions to drivers, might give them the feeling of being patronized by the system and might decrease system acceptance. The aim of this thesis is to investigate adaptive system behavior as a solution for the trade-off between safety and driver satisfaction in the case of local danger alerts. An algorithm for online measuring of drivers stress is developed based on an existing database of offline measures. Using physiological sensors and driving performance, elicited from drivers position inside the lane, action suggestions are presented only when drivers stress is estimated to be high.

Analyzing Theremin Sound for Touch-Free Gesture Recognition


Bachelor Thesis by Svilen Dimitrov advised by Christoph Endres and Christian Müller

The increasing number of car accidents caused by distraction of the drivers mostly by the increasing amount of devices in every car, leads to rising interest in innovative human computer interaction technology. The two most studied methods are gesture and speech recognition. In the gesture recognition field, numerous techniques are using different expensive devices and complex algorithms or combination of them to increase robustness. This thesis investigates a new method for two dimensional gesture recognition of index finger movement, having one dimensional input by using magnetic fields. The gestures that are about to be recognized, vary from single straight moves to more complex geometric figures. The way to do this is straightforward, flexible and low cost. The core input component studied in this work is an early electronic musical instrument named Theremin after its founder. It is placed in immediate proximity of a steering wheel providing gesture based interaction without requiring the driver to take off hands. Despite of all the benefits this method has, it is still in early development phase and can be further improved by adding more antennas and by using modern classification algorithms to obtain greater accuracy.


Net-Bike: An Approach for Incorporating Bicycles into the Vehicle2X Communication Framework Using Consumer Device


Bachelor Thesis by Nils Schnabel advised by Sandro Castronovo and Christian Müller

This bachelor thesis presents a system to encorporate bicycles into Vehicle2X (V2X)  networks. The system proposed is a composition of two components: One of them is a  device based on standard router hardware, responsible for running safety applications and providing the wireless (V2X) connectivity. The other component consists of a smartphone  which offers a user interface for the bicyclist and positioning sensor data required for V2X communication.

This combination of devices is an attempt to reuse smartphone technology many users  already have at hand, to reduce cost and installation effort. Additionally, this combination  allows to extend the bicycle side with additional hardware components, as well as standards-compliant V2X communication.


Thin-client Vehicle2X-communication compliant visibility-enhancement for bicycles


Bachelor Thesis by Tobias del Fabro advised by Sandro Castronovo and Christian Müller

 In the past few years the demand for economically friendly and non-polluting, so called "green", solutions has increased. More and more people tend to use their bicylces and "go green" instead of using the car for everyday business. Companies even encourage their employees to come to work on bike and provide a bike-friendly environment such as showers and changing rooms. Many cities try to catch up with this trend and have integrated bike-paths into the infrastructure. These allow cyclists to reach their destinations more quickly than they would do via public transport and simultaneously reduce the traffic on normal roads.

One of those cities is Münster (Germany) - a university town where according to statistics about 47% of the 280.000 inhabitants use mostly their bicycle as mean of transportation and about 77% use a bicycle daily or several times per week. But this increasing number of cyclists represents also a new amount of vulnerable road users in the already dense traffic of a city. Not everywhere bike paths and special signal lights exist and thus the well-known safety issues arise: a bicycle suddenly appears in front of a driving car and they crash because the driver hasn't seen it. Be it lapse of concentration, bad illumination or any other kind of distraction. Today for cyclists there are more possibilities to make themselves more visible: bright LEDs are used for spotlights and for clothing, bike jackets can even display changes in motion as for example breaking or indicate changes of direction.

In this thesis a novel form of visibility enhancement is developed in order to protect vulnerable road users (VRUs), here: bicyclists. The system uses standard compliant Vehicle-2-X (V2X) Communication technology to incorporate bicycles into the existing network. Current approaches turn on flashing lights and horns in order to increase the visibility of motorcycles. It can be argued that this approach has various drawbacks: First, the solution is visible and audible by other, uninvolved road users as well. This could lead to serious distraction. Second, current solutions are expensive and not tailored to the needs of bicylclists. Third, the power consumption is designed for motorcycles not for bicycles where powerful batteries are missing. In this paper we present a solution which addresses these facts: This novel approach to visibility enhancement is ``directed'' to the specific addressee and low cost since  consumer hardware is used. It is energy-efficient for the use on bicycles.


Optische Kopfhaltungserkennung im Auto als Wissensquelle für multimodale Mensch-Maschine-Schnittstellen


Master Thesis by Chakib Bensajjay advised by Michael Feld and Christian Müller

 Diese Arbeit beschäftigt sich mit kamera-basierten Methoden zur nicht-intrusiven Wissensakquisition für Benutzer im Fahrzeug. Diese finden einerseits Einsatz als Eingabemodalität in multimodalen Systemen, darüber hinaus dient das durch sie ermittelte Wissen als Basis für die Adaption bestehender Mensch-Maschine-Schnittstellen. Konkret wurde ein Bildverarbeitungs-Verfahren zur Ermittlung der Kopfhaltung des Fahrers entwickelt, implementiert und evaluiert. Außerdem wurde ein Framework erstellt, welches die Ansteuerung von mehreren im Auto angeschlossenen Kameras und die Bereitstellung von Einzelbild- und Videodaten für Anwendungen erlaubt.


Seamless Positioning of Nomadic Devices for In-Car and In-Door Applications


Master Thesis by Bahjat Saliba advised by Tim Schwartz and Christian Müller

 In this thesis an evaluation method on indoor positioning system called LORIOT is presented. This positioning system combines two technologies (RFID and IR) for positioning depending on Geo-referenced dynamic bayesian networks. LORIOT allows the users to calculate their position on their own device without sending any data to a server responsible for calculating the position. This property provides less complexity and fast calculation. This positioning method is developed by placing the tags in the environment and letting the user carry the sensors that are used to read data from these tags. The user is then able to choose either to pass the positioning data to any third party application or not. The main focus here is to check the actual accuracy and performance of indoor positioning systems using the proposed evaluation method which is tested on LORIOT. Most of the evaluation methods that have been used to test the level of accuracy of indoor positioning systems are biased and not good enough. For instance, the system is tested under optimal conditions of the environment. To achieve this goal, the evaluation method will be used to test LORIOT in a natural environment and by using data of natural traces of people walking in the environment without giving them any task to do. This type of evaluation criteria improves the results because the system would be installed in an environment which has the same properties that the environment has in this study, (where the evaluation tests are done). In addition, the system will position people while walking naturally (unlike most evaluation methods which test indoor positioning systems not while walking).


An Extendable and Standard-Compliant Driving Simulator Software for Experimental Assessment of Driver Distraction and Reaction


Master Thesis by Saied Terani advised by Angela Mahr and Christian Müller

Driving simulators are important tools for performing experiments in automotive research. But sophisticated simulators are very expensive. The available low cost simulators, however, do not always meet all the requirements needed for specific experiments. For the present master's thesis, an extendable low cost simulator called ExDS has been developed that is based on open source technology. Its suitability for research has been verified by performing an experiment based on the ISO-standardized Lane Change Test (LCT), which is an experiment-framework for the analysis of driver distraction. The results of the experiment are compared with those of another popular simulator that has been developed at the company Daimler AG. The comparison confirms ExDS's capabilities and suitability for driver distraction experiments. In contrast to the simulator developed at Daimler, ExDS can be extended with new functionalities.


A Combined Uniform and Heuristic Search Algorithm for Finding Shortest Paths in Unknown Highly Dynamic Graphs


Master Thesis by Björn Kunz advised by Sandro Castronovo and Christian Müller

 There is a significant amount of research focused on Cooperative Vehicles in projects like e.g. SimTD, PReVENT and SAFESPOT]. Cooperative Vehicles communicate via Wireless Ad Hoc Networks which are in this special case called Vehicular Ad Hoc Networks because of the high velocity with which the network participants move. Due to this high movement speed and the environment, which in an urban scenario consists of roads between buildings, the vehicles in the network frequently establish and loose connections. If a vehicle wants to, e.g. query information on the number of free spaces in a parking lot when it is not yet in direct transmission range with a road-side unit (short RSU) installed at the parking lot the packet needs to be sent via intermediate vehicles. The process of communcation via intermediate nodes, in this case vehicles, is called routing and the algorithm deciding, which of the neighbouring nodes should be forwarding the message is called a routing algorithm.

When looking at routing in Vehicular Ad Hoc Networks we can easily transfer this problem to graph theory as path computation on fully dynamic graphs. A fully dynamic graph is a special type of graph that changes over time according to an allowed set of operations, e.g. deleting or inserting edges. The network formed by the vehicles can be viewed as a graph with the vehicles constituting the vertices of the graph and edges denoting the connections  between these vehicles. By abstracting the problem to graph theory we can give a general solution that is applicable to domains other than Vehicular Ad Hoc Networks.  The problem of shortest path searches is well known on non-dynamic graphs but gets more complicated when the graph becomes dynamic. In the domain we're interested in, Cooperative Vehicles in Vehicular Ad Hoc Networks, we can't rely on the assumption that we  have complete knowledge about the topology of the graph and the operations applied to it. Therefore the problem becomes one of computing shortest paths on unknown fully dynamic graphs.

This thesis adds to the research already conducted in Computer Science and Artificial Intelligence as well as Vehicular Ad Hoc Networks by abstracting the problems faced in the domain of Vehicular Ad Hoc Networks onto the higher level of path searching on an unknown fully dynamic graph as stated before. We show that our solution is abstract enough such that an application to other domains than Vehicular Ad Hoc Networks is possible. We evaluate our algorithm specifically in the domain of Cooperative Vehicles in Vehicular Ad-hoc Networks against algorithms from the domain itself. Our algorithm outperforms both reference algorithms with respect to Path Discovery Ratio (short PDR) and outperforms one of the reference algorithm with respect to Path Discovery Time (short PDT) when a certain traffic density is reached.

Rule-Based Priority Management for Advanced Driver Assistance Systems


Master Thesis by Michael Maurer advised by Christoph Endres and Christian Müller


Physiological measurement of driver stress induced by car2X-based local danger warnings


Master Thesis by Veronika Dimitrova advised by Angela Mahr and Christian Müller

Advanced Driver Assistance Systems (ADAS) aim at providing safety and comfort to the driver and have become an integral part of our daily life. By designing such a system, a number of important aspects (such as understanding physical and cognitive demand of driving) should be considered, in order to provide the driver a reliable and trustworthy product. An important question for the development is also what kind of influence a system has on the driver, in the sense of his physiological state or, more precisely, whether differences in her/his stress level can be detected. Physiological measurements are a tool that helps answering this question.

In this thesis, offine measurements with physiological sensors were done in order to test how the stress level of the driver changes with using a special in-vehicle information system for presenting driver relevant warnings. Heart rate, respiration, skin conductance, and temperature measurements were applied to drivers in an experimental study, in which they were instructed to drive under different conditions. Furthermore, the measured sensor data was processed and analysed. The results indicate that by presenting a warning message for upcoming obstacles on the road, the stress level of the driver decreases in comparison to normal driving without warning presentation. Comparing the different modalities of information presentation, did not yield significant differences in the physiological states of the driver. However, a trend was noticeable in favor for visual presentation combined with a blinking bar in order to attract the attention and with action suggestion. Further analysis of the physiological measurements showed that after unanticipated failure of the in-vehicle information system the drivers' stress levels increased significantly, compared to the working system.

Referring to the Outside Environment from Within a Moving Vehicle: Adapting Multimodal Dialog to Vehicle Speed and Role of the User

mehdi_copy referring_to_outside_objects

Master Thesis by Mohammad Mehdi Moniri advised by Sandro Castronovo and Christian Müller

 This thesis investigates various modalities for referring to the outside environment from within a moving vehicle. A brief theoretical study is conducted to explore the different aspects of the vehicle context, regarding the role of the user and the speed of the car. Based on this study, following modalities were explored in order to create the reference set: Eye Gaze, Head Pose, Pointing Gesture, Camera View and the user's View Field. All of these modalities were implemented and tested in a moving vehicle in a real life traffic. Afterwards, using all of the named modalities, a data collection in the city Saarbrücken was conducted. During the data collection, the passenger used each modality to refer to the buildings in the city center. The driver was not involved in the data collection procedure. For analysis and visualization purposes 528 buildings of the city were modeled in 2.5D by using an airborne LIDAR scan of the city center, Google Earth, and a spatial database. Using this model it is possible to visualize the three dimensional interaction vector of the user in Google Earth.


  • What are the requirements of a multimodal object reference component (touch, pointing, eye gazes, speech) depending on
    • the speed of the vehicle ?
    • the role of the user (characteristics of the primary task) ?
  • How can a system be designed in order to meet these requirements (an how can it be evaluated) ?

In this thesis, a new algorithm for spatial reference resolution together with a scanning mechanism is developed as well. The scanning mechanism is based on a visibility analysis so that the system computes which parts of the buildings in the surrounding environment are visible to the user. The developed spatial reference resolution algorithm is used to detect the target buildings in the collected data of four modalities. This algorithm is designed in a way that considers various aspects of car context. The results of applying this algorithm with five different variable assignments to the data collected is also presented and analyzed. Finally, the potential and planned future works are described.

For further information please contact Mohammad Mehdi Moniri.

Combination of Speech and Tangible Interfaces for Automotive Dialog Systems: An Experimental Study


Master Thesis by Margarita Pentcheva, finished September 2009. Advised by Angela Mahr and Christian Müller

In recent years, the complexity of on-board and accessory devices, infotainment services, and driver assistance systems in cars has experienced an enormous increase. This development emphasizes the  need for new concepts for advanced human-machine interfaces that support the intuitive and efficient use of this large variety of devices and services. This thesis describes an experimental study of driver distraction induced by the interaction with the car's comfort functions under three conditions: manual (conventional), speech-only, and multimodal interaction. The latter is realized as a combination of speech and a tangible interface device (turn-and-push dial ).

The driver distraction is assessed via subjective rating by the subjects using a standardized questionnaire (DALI ) and objective measurement  of the driving performance (LCT ). The Lane Change Test measures distraction of a primary task while performing a secondary task. This method allows to research whether both speech-only and combined conditions outperform the manual condition in terms of distractiveness and convenience. Results
show that the driver distraction in both speech-only and multimodal conditions are significantly lower than with manual (conventional) interaction. While there is no significant difference between speech-only and multimodal in the objective task, the latter is subjectively rated as more demanding by the subjects. In terms of task completion (number of successful interactions performed), the manual method outperforms the other two.