Mark Eilers

Probabilistic and Empirical Grounded Modeling of Agents in (Partial) Cooperative Traffic Scenarios

The Human Centered Design (HCD) of Partial Autonomous Driver Assistance Systems (PADAS) requires Digital Human Models (DHMs) of human control strategies for simulations of traffic scenarios. The scenarios can be regarded as problem situations with one or more (partial) cooperative problem solvers. According to their roles models can be descriptive or normative . We present new model architectures and applications and discuss the suitability of dynamic Bayesian networks as control models of traffic agents: Bayesian Autonomous Driver (BAD) models. Descriptive BAD models can be used for simulating human agents in conventional traffic scenarios with Between-Vehicle-Cooperation (BVC) and in new scenarios with In-Vehicle-Cooperation (IVC). Normative BAD models representing error free behavior of ideal human drivers (e.g. driving instructors) may be used in these new IVC scenarios as a first Bayesian approximation or prototype of a PADAS.

Further Steps towards Driver Modeling According to the Bayesian Programming Approach

The Human Centered Design (HCD) of Partial Autonomous Driver Assistance Systems (PADAS) requires Digital Human Models (DHMs) of human control strategies for simulating traffic scenarios. We describe first results to model lateral and longitudinal control behavior of drivers with simple dynamic Bayesian sensory-motor models according to the Bayesian Programming (BP) approach: Bayesian Autonomous Driver (BAD) models. BAD models are learnt from multivariate time series of driving episodes generated by single or groups of users. The variables of the time series describe phenomena and processes of perception, cognition, and action control of drivers. BAD models reconstruct the joint probability distribution (JPD) of those variables by a composition of conditional probability distributions (CPDs). The real-time control of virtual vehicles is achieved by inferring the appropriate actions under the evidence of sensory percepts with the help of the reconstructed JPD.

Predicting the focus of attention and deficits in situation awareness with a modular hierarchical Bayesian driver model

Situation Awareness (SA) is defined as the perception of elements in the environment within a volume of time and space, the comprehension of their meaning, and the projection of their status in the near future [1]. Lacking SA or having inadequate SA has been identified as one of the primary factors in accidents attributed to human error [2]. In this paper we present a probabilistic machine-learning-based approach for the real-time prediction of the focus of attention and deficits of SA using a Bayesian driver model as a driving monitor. This Bayesian driving monitor generates expectations conditional on the actions of the driver which are treated as evidence in the Bayesian driver model.

Modelling Aspects of Longitudinal Control in an Integrated Driver Model

Simulating and predicting behaviour of human drivers with Digital Human Driver Models (DHDMs) has the potential to support designers of new (partially autonomous) driver assistance systems (PADAS) in early stages with regard to understanding how assistance systems affect human driving behaviour. This paper presents the current research on an integrated driver model under development at OFFIS within the EU project ISi-PADAS. We will briefly show how we integrate improvements into CASCaS, a cognitive architecture used as framework for the different partial models which form the integrated driver model. Current research on the driver model concentrates on two aspects of longitudinal control (behaviour a signalized intersections and allocation of visual attention during car following). Each aspect is covered by a dedicated experimental scenario. We show how experimental results guide the modelling process.

Development of a Lane Change Assistance System Adapting to Driver's Uncertainty During Decision-Making

This paper presents the development of a lane change assistance system adapting to both criticality and drivers uncertainty during decision-making in overtaking scenarios on simulated two-lane highways. Based on information about the traffic situation, the proposed system relies on a probabilistic model of drivers uncertainty to classify whether a driver is unsure or not and on a safety analysis based on drivers preferences to suggest appropriate overtaking actions, which together trigger corresponding advices on the Human Machine Interface. Different to existing lane change assistance systems using traffic light colors to encode criticality and warn accordingly, the proposed system uses colored abstract faces with emotional expressions encoding both criticality and drivers uncertainty to provide suggestions to either overtake or decelerate. The proposed system has been implemented in a driving simulator. The qualitative results of a not yet analyzed evaluation study with 20 participants show that the proposed system is more accepted and trusted than reference systems that do not consider drivers uncertainty.

Integrating Anticipatory Competence into a Bayesian Driver Model

Background We present a probabilistic model architecture combining a layered model of human driver expertise with a cognitive map and beliefs about the driver-vehicle state to describe the effect of anticipations on driver actions.

Building driver's trust in lane change assistance systems by adapting to driver's uncertainty states

Drivers uncertainty during decision-making in overtaking results in long reaction times and potentially dangerous lane change maneuvers. Current lane change assistance systems focus on safety assessments providing either too conservative or excessive warnings, which influence drivers acceptance and trust in these systems. Inspired by the emancipation theory of trust, we expect systems providing information adapted to drivers uncertainty states to simultaneously help to reduce long reaction times and build the overall trust in automation. In previous work, we presented an adaptive lane change assistance system based on this concept utilizing a probabilistic model of drivers uncertainty. In this paper, we investigate whether the proposed system is able to improve reaction times and build trust in the automation as expected. A simulator study was conducted to compare the proposed system with an unassisted baseline and three reference systems not adaptive to drivers uncertainty. The results show while all systems reduce reaction times compared to the baseline, the proposed adaptive system is the most trusted and accepted.