Simulation of Vehicle-Pedestrian Interaction

An important consideration for the process chain is that traffic accidents are statistically rare events and that each accident is a unique event (see Chap. 1). Detailed accident data are available in various databases and are very helpful in determining the potential of safety measures. However, critical traffic situations or near-accidents are not included in any representative databases. Field Operational Tests, like eu-roFOT [22], or Naturalistic Driving Studies, like the 100-Car Naturalistic Driving Study [23], collect data of such events, but often lack representativity for the traffic system as a whole. A possible solution can be provided, for example, by methods such as fault-tree analysis (see [24, 25]) and stochastic simulation. The simulation described in the following is a stochastic simulation, as many of the subprocesses involved include distributed parameters, which are hard to account for, e.g., in a fault-tree analysis. 

All relevant processes are modeled and linked with realistic probability distributions. Each parameter is drawn randomly with respect to its probability distribution and possible dependencies on other factors in the simulation. The implemented scenario is an urban crossing scenario, as this is the most important one (see Sect. 3.2). The pedestrian crosses the street (straight road) from the right to the left from the view of the driver in the middle of a block. From the pedestrian s point of view, the traffic comes from the left. Scenario parameters include, for example, the geometry of the sidewalk, speed limit of the street or visibility restrictions. The traffic on the road itself is implemented as an exposure model depending on time of day and day 

3 Approach to Integrated Safety Evaluation Preventive Pedestrian Protection 

The vehicles in the simulation move on a straight street and have dimensions typical for mid-size vehicles. The braking capabilities are typical for up-to-date vehicles and typical road surfaces. The implemented preventive pedestrian protection system is thus modeled as part of the vehicle. Once the pedestrian is visible for the system, the probability per unit time that the pedestrian is detected by the system is modeled as a constant. The algorithm of the system includes a prediction of the vehicle s movement and the pedestrian s movement as well as the calculation of a collision probability as basis for a system action. The system itself has various stochastic components, e.g., inaccuracies regarding position and speed of the pedestrian. Depending 

---