Estimating the benefits of advanced safety systems before introducing to markets is useful to develop and enhance the systems effectively. Several estimation methods have been proposed to date. Some are based on comprehensive accident data such as those of NASS-CDS. Others are based on proving-ground test results. However, actual accidents present much more permutations and configurations of striking and struck vehicles than those. Furthermore, driver behavior varies among cases.

This paper presents a proposal of a novel method that addresses the issues described above. First, a virtual traffic flow that represents an actual one is created. Then, the way in which an active safety system is expected to play its role in accidents happening in traffic is measured. The Advanced Safety System & Traffic REaltime Evaluation Tool (ASSTREET) was used to generate road environment, vehicle movements, and driver behavior. In order to show the usefulness of the method, a pre-collision system (PCS) with forward collision warning (FCW), pre-collision brake assist (PBA), and pre-collision brake (PB) functions were applied as the active safety system. The procedure is the following. A virtual traffic flow was created. On a simple road environment with intersections and traffic signals, numerous vehicles run under traffic rules on ASSTREET. The vehicles' speed distribution and the duration of the stopping period were adjusted to match realistic driving data measured on roadways, by the road parameters such as speed limits and the distance between intersections.

Next, rear-end collisions in the virtual traffic flow were created. Driver errors and braking reaction after noticing the collision danger were incorporated into the virtual driver behavior. Because most of the driver errors in rear-end collisions are attributable to inattention, the inattention period and the brake reaction time with a convincing distribution were given to the virtual drivers. The braking deceleration distribution, which is also necessary characteristics for pre-collision reconstruction, was obtained using our driving simulator through the ACAT (Advanced Collision Avoidance Technology) program with NHTSA (National Highway Traffic Safety Agency). The distribution of the combination of striking vehicle speed and struck vehicle speed agreed well with actual data. Consequently, rear-end collisions in the simulation were regarded as representing actual ones.

Finally, the benefit of PCS was estimated. Rear-end collisions in the virtual traffic flow were generated by vehicles with no active safety systems. After collecting all rear-end collision pairs of striking and struck vehicles, a PCS was installed in striking vehicles. Then the simulation was repeated. Comparing the results obtained with and without use of the system clarifies the PCS benefit.

The advantage of this method is that a mass of rear-end collisions enables evaluation of PCS' specification differences quantitatively. Results clearly indicate circumstances in which the system is expected to function effectively.

Although the current simulation is considered as covering most of rear-end collisions that people might happen to encounter, such scenarios as avoidance by steering, collision during negotiation of a curve, and collision with a cutting-in vehicle have not been simulated yet. Those will be addressed in the near future.