New technologies are becoming available to reduce the frequency of crashes. They may be vehicle based or road based and will involve a variety of levels of information provision to drivers and increasing levels of control over the vehicle. Vehicle systems under development include intelligent speed control, lane keeping, adaptive cruise control, night vision, driver drowsiness detection while road based systems include information services and signalling. Systems development is made on the basis of technological factors, experimental studies and human factors approaches. While improved safety is a prime objective of a number of these systems, there are currently few methods available to systematically assess the in-depth application of the technologies in specific accident situations. The paper reports on a new methodology for vehicle and traffic simulation that reproduces the drivers' and vehicles' actions in the period leading to a crash. Autonomous driver agents are used to simulate the observations, behaviour and decision making of the driver while vehicle dynamics modules and road modules place the driver within the traffic and road context. To enable virtual drivers to emulate some of the unpredictable behaviour of their human counterparts, each driver agent has the capability to perceive their environment, make decisions based on what they 'see' and take appropriate actions. So far, several aspects of the model have been validated against experimentally derived data. The model has been used to simulate the pre-crash events leading to cases examined within the UK On-the-Spot Accident study. Case studies are presented and other applications of the simulation methodology relating to driving simulators, virtual road design and other transport applications will also be discussed. The oral presentation will include video run throughs of real-world scenarios and their simulations. For the covering abstract see ITRD E825082.
Samenvatting