Safety

Frontal collision warning is considered one of the most critical applications within the intelligent vehicle initiative (IVI) program (www.its.dot.gov/ivi/ivi.htm). California Partners for Advanced Transit and Highways (PATH, www.path.berkeley.edu) program started a project under the sponsorship of Federal Transit Administration (FTA, www.fta.dot.gov) to investigate the appropriate specifications for frontal collision warning systems for transit buses. The project is being conducted with the cooperation of several partners, including San Mateo Transit Agency (Samtrans), Gillig Corporation, and California Department of Transportation (Caltrans). [1]

The distinct nature of frontal collision warning systems in transit buses lies in their operating environment, which differs significantly from those in highway applications. First of all, transit buses operate in local streets with frequent stops, compared to high-speed cruising scenarios on highways. Secondly, a more diverse variety of obstacles are present on bus routes. Furthermore, from an analysis of accident data, it was noticed that transit buses often encounter situations where the front end or corners of a bus may make contacts with vehicles or obstacles at close range during their maneuvers at intersections, turns, or stops. All these factors make the implementation of collision warning systems for transit buses much more complicated than those designed for highway applications.

In this paper, we provided an analysis of real-world collision scenarios by reviewing an extensive list of accident situations from transit agencies. This analysis helped us identify certain aspects of bus operating environment that are unique and significant. We also laid out the foundation of scenario parsing by identifying the expected data patterns to be acquired by sensors placed on test vehicles. The collection of field data and the subsequent data dissection will serve as the basis for threat assessment in various traffic conditions and algorithm design for collision warning.

California PATH, Headquarters

Presented at the 11th ITS Annual Conference and Exposition, June 4-7, 2001 Miami Beach, Florida

We are developing a traffic simulator called  STREET, which is intended to enable us to
evaluate safety systems. To evaluate safety systems, however, we have to be able to faithfully
reproduce traffic accidents that are caused by interactive events between vehicles and between
vehicles and pedestrians. To this end, it is  necessary to construct  a driver model that
incorporates cognition, decision-making, and operation behaviors.  STREET simulates the
behavior of drivers and traffic  accidents that are caused by driver errors. In this paper, we
describe a driver model for  STREET, in which a driver recognizes the environment
surrounding his or her vehicle by using their abilities, and then decides the most appropriate
driving maneuver as affected by the driver’s characteristics. Also, we propose a method of
reproducing traffic accidents that are caused by a driver’s ability to perform non-driving tasks
being adversely affected. In addition, we talk about evaluating the ratio of accidents
with/without active safety systems.

TOYOTA CENTRAL R&D LABS., INC.

Presented at the ITS America Annual Conference and Exposition, November 16-20, 2008, New York, New York

We are developing a traffic simulator called STREET, which is intended to enable us to
evaluate safety systems. In this paper, we propose a new structure for the pedestrian model used
by STREET, which will give us the means to evaluate vehicle-pedestrian accidents and safety
systems. To simulate the sequence of a pedestrian’s behavior leading up to an accident, we
assume that they determine their next behavior from their cognition of objects such as vehicles.
Accidents occur as a result of pedestrians engaging in behaviors unrelated to walking such as,
in our model, using a mobile phone and generally being inattentive. The model also
incorporates behaviors resulting from pedestrian-unique accident factors (such as a “priority
sense”) which were identified by analyzing actual accident data. Finally, we confirm that
STREET not only lets us evaluate how safety systems reduce the occurrence of accidents, but
also analyze the factors contributing to accidents and the effectiveness of some safety systems.

TOYOTA CENTRAL R&D LABS., INC.

Presented at the ITS America Annual Conference and Exposition, November 16-20, 2008, New York, New York

We are developing a traffic simulator called STREET which is intended to enable us to evaluate
safety systems. To evaluate safety systems, however, we have to be able to faithfully reproduce
accidents that are caused by interactive events between vehicles and pedestrians. To this end, it
is necessary to construct a driver model that incorporates cognition, decision-making, and operation
behaviors. In this paper, we focus on a driver’s visual behavior, which is the first phase
of the cognition process, with the goal of solving some assignments in STREET. We propose
a novel behavior model by employing a model-based approach. First, we construct a hypothetical
model whereby the driver estimates the risks in his/her surroundings through a recording
process. Second, we measure the unknown parameters within the model by way of experiment.
Finally, we confirm that the proposed model allows STREET to faithfully reproduce the driver’s
visual behaviors.

Toyota Central R&D Labs., Inc. Vehicle Safety Research Center

Presented at the ITS America Annual Conference and Exposition, November 16-20, 2008, New York, New York

The increasing in-vehicle information and safety systems tend to confuse and distract the
driver from his/her primary driving task. This paper develops algorithms for the real-time
supervision of the traffic and environmental scenario around the vehicle for the optimization
of the Human Machine Interaction. The proposed algorithms reconstruct the scenario using
stochastic motion models and Kalman filters, predict the intention of the driver using
Demspter-Shafer decision fusion and calculate the level of risk in using fuzzy logic. The
algorithms will be part of the Driver – Environment – vehicle state estimation in AIDE
Integrated project.

Institute of Communications and Computer Systems, I-SENSE Group

Presented at the 12th World Congress on Intelligent Transport Systems, November 6-10, 2005, San Francisco, California