Dietrich Manstetten

Predicting lane keeping behavior of visually distracted drivers using inverse suboptimal control

Driver distraction strongly contributes to crash-risk. Therefore, assistance systems that warn drivers if their distraction poses a hazard to road safety, promise a great safety benefit. Current approaches either seek to detect critical situations using environmental sensors or estimate a drivers attention state solely from his/her behavior. However, this neglects that driving situation, driver deficiencies and compensation strategies altogether determine the risk of an accident. This work proposes to use inverse suboptimal control to predict these aspects in visually distracted lane keeping. In contrast to other approaches, this allows a situation-dependent assessment of the risk posed by distraction. Real traffic data of seven drivers are used for evaluation of the predictive power of our approach. For comparison, a baseline was built using established behavior models. In the evaluation our method achieves a consistently lower prediction error over speed and track-topology variations. Additionally, our approach generalizes better to driving speeds unseen in training phase.

Behaviour Prediction and Intention Detection in UR:BAN VIE – Overview and Introduction

The recognition of driver intentions can enhance the driver-vehicle interaction and offer more intuitive assistance and driving support. The assistance should display comfortably timed warnings only in situations where they are really needed and has to act in accordance to the driver’s intentions. This was the framework of the UR:BAN sub-project VIE on detecting driver’s intention and predicting his behaviour. This introductory chapter starts with a description of the project’ objectives. It gives an overview on the working method during the project and summarises the results. Finally, it gives some indications on the main topics of the following chapters presenting parts of the project’s activities in detail.

Traffic Simulation for the Development of Traffic Management Systems

This paper describes the industrial application of traffic simulation at Robert Bosch GmbH. The role of traffic simulation in the design process of traffic management systems and the way of working of Bosch R&D are presented. The main features of Bosch’s simulation environment ARTIST (Advanced Research Tool for Indoor Simulation of Traffic) are described. Sample applications in the area of freeway management systems and urban traffic light control demonstrate how traffic simulation is currently used in the product development process. ARTIST’s flexibility allows the usage of domain specific traffic models that are best suited to the specific task.

Predicting Strategies of Driving in Presence of Additional Visually Demanding Tasks: Inverse Optimal Control Estimation of Steering and Glance Behaviour Models

Driver distraction strongly influences the accident risk on both motorways and urban streets. In this context, visual distraction - long glances off the road - is the most contributing factor. However, in natural driving engagement in visually distracting activities is very frequent compared to the small number of critical incidents. This indicates that drivers apply situational-adaptive gaze and driving strategies that can provide a certain amount of driving safety. Yet, most state-of-the-art mitigation systems assess driver distraction based on fixed thresholds on glance duration. This chapter presents an approach for prediction of situation specific human behaviour in distracted driving. Here, we apply a driver model based on sub-optimal control. Taking into account driver strategies and their potential insufficiencies in the current driving context, our method has the potential to greatly improve assistance systems, by reducing unneeded warnings and interventions. This holds true especially in urban scenarios that are characterized by a broad variety of driving situations.

1st Workshop on Understanding Automation: Interfaces that Facilitate User Understanding of Vehicle Automation

This workshop addresses how in-vehicle interfaces could be designed to support humans in understanding how highly automated vehicles (HAVs) operate. Current practices describe levels of automation in terms of technological limitations and expect users to accommodate. However, humans might not be able to understand the implications of technical limitations. Therefore we discussed how automation could be designed to understand the behavioral limitations and proclivities of human users. It also addresses how human-machine interfaces could provide users with an accurate mental model of automation. While transparency is often promoted as a crucial design principle for human-automation interfaces, doing so without thought can give rise to information overload. As outcomes, we identified potential misunderstandings that humans might hold of automated systems, how these misunderstandings can be resolved with novel interfaces, and what measures could be taken to develop automated systems that are easily understandable and capable of understanding their users in return.

Exact Maximum Entropy Inverse Optimal Control for modeling human attention switching and control

Maximum Causal Entropy (MCE) Inverse Optimal Control (IOC) has become an effective tool for modeling human behavior in many control tasks. Its advantage over classic techniques for estimating human policies is the transferability of the inferred objectives: Behavior can be predicted in variations of the control task by policy computation using a relaxed optimality criterion. However, exact policy inference is often computationally intractable in control problems with imperfect state observation. In this work, we present a model class that allows modeling human control of two tasks of which only one be perfectly observed at a time requiring attention switching. We show how efficient and exact objective and policy inference via MCE can be conducted for these control problems. Both MCE-IOC and Maximum Causal Likelihood (MCL)-IOC, a variant of the original MCE approach, as well as Direct Policy Estimation (DPE) are evaluated using simulated and real behavioral data. Prediction error and generalization over changes in the control process are both considered in the evaluation. The results show a clear advantage of both IOC methods over DPE, especially in the transfer over variation of the control process. MCE and MCL performed similar when training on a large set of simulated data, but differed significantly on small sets and real data.

Taking CHI for a drive: interaction in the car

With the increasing number of cars on the road, longer commutes, and the proliferation of complex information and entertainment features, there is a greater need for careful interaction design in the car. The automobile is a challenging environment for designing and deploying good user interfaces. Interaction designers must balance brand identity, safety, legislation, and manufacturability, among other issues. In this panel, practitioners and researchers from industry, industrial labs, and academia will discuss the challenges of interaction design in an automotive environment. While some members of the CHI community are active in the automotive field, the general CHI community may not be aware of this work, the open research issues, and opportunities for collaboration in this area. This panel will provide an introduction into HCI research in the automotive industry. Some successful examples of interaction design will be discussed, as well as a few not-so-successful examples. Questions and comments from the audience are welcomed.