HyBS-HD

HyBS-HD -

Hybrid Behaviour Simulation of Human Drivers

To what extent is the formal modeling approach of hybrid automata suitable for describing psychological models? To what extent are reinforcement value-based behavioral models suitable for describing human behavior in traffic situations?

Funding

Funded by the Zukunftsfond of the TU Braunschweig

Project partner

  • Institut für Programmierung und Reaktive Systeme (Prof. Dr. U. Goltz)

Project team member (IPMB)

Dr. Anke Schwarze

Dr. Matthias Buntins

Aim

The aim of the project is to formally and quantitatively describe the behavior when driving onto a highway using hybrid automata and to empirically substantiate this description. To achieve this goal, the language of hybrid automata must be extended. So far, we have already added unknown functions and an associated amplifier value variable to be maximized. In the medium term, however, we will also integrate multi-agent approaches into the model in order to be able to describe larger traffic situations reasonably. We are also developing software tools to support model development and simulation.

Publikationen

[BSEG11a] M. Borgstede, J. Schicke, F. Eggert, and U. Goltz
Verhaltensbasierte Fahrermodellierung mittels Hybrider Automaten
TeaP 2011

Abstract:
A new modeling approach is presented, which realizes the behavior of car drivers not via cognitive decision and control processes, but as a result of adaptive learning processes that lead to the optimization of a situation-dependent reinforcement value function. Hybrid automata, which model both qualitative and quantitative state changes, serve as the formal framework for the simulation. An initial application of the approach for driving onto the highway delivers promising results: The trajectory of the oncoming vehicle, as well as the speed progression, are realistically predicted. In addition, the model predicts the initiation and execution of complex driving maneuvers such as "merging" or "overtaking", depending on the traffic situation on the freeway, without ever having specified these.

[BSEG11] M. Borgstede, J. Schicke, F. Eggert, and U. Goltz Hybrid Automata as a Modelling Approach in the Behavioural Sciences
Proceedings of the HAS Workshop, 2011

Abstract:
Hybrid modelling techniques have evolved as a tool to design, verify and guide the implementation of embedded systems. However they can ­ and we think should ­ be used to express quantitative models about hybrid systems in other domains, e.g. empirical sciences. Since the formal structure of hybrid automata corresponds well to sequentially organized behaviour chains in living organisms, we argue for an application of hybrid modelling techniques in the behavioural sciences and, especially, Psychology. We try to answer the question how human drivers move onto a freeway and at the same time use this research as our testbed for hybrid automata in the behavioural sciences.