Summary

Proposed Research

Vehicle safety is one of the most important concerns of transit authorities and automotive manufacturers. The goal of transit authorities in this regard must be zero accidents, which implies the loss of no human life and the removal of a major cause of traffic congestion on highways. Vehicle crashes cause more than 41,000 fatalities per year, as well as 5.3 million injuries, and cost our society more than $230 billion per year. A large portion of these accidents is due to drivers' inattention or inability to perform driving tasks. Automotive manufacturers have introduced driver assistant systems such as collision warning systems (CWS) and road departure warning systems (RDW) for enhanced safety. CWS/RDW systems observe the environment around the vehicle (such as the distance from the preceding vehicle, and obstacles in the lane), monitor vehicle states (e.g. lateral position in the lane, and longitudinal velocity), predict possible collision and abnormal driving patterns, and alert the driver to take action to maintain the safety of the vehicle. Most CWS/RDW systems do not directly observe the driver even though the driver is the major cause of highway accidents: the driver makes wrong decisions, fails to pay attention at critical moments and/or gets sleepy. An implicit assumption is that any of the problems about the driver may be reflected in and can be identified from the vehicle behavior and that the driver may take a necessary measure once a warning message is generated. The indirect observation of the driver state via vehicle dynamics sets a limit to the performance of the current driver assistant systems. In particular, if the driver is getting drowsy, it is critically important to detect it as early as possible. With direct monitoring and observation of the driver state from available data and incorporation of it in the driver assistant systems, it becomes possible to alert the driver in a timely and effective manner and increase the safety of highway driving.

Most of commercial CWS/RDW systems have been developed for passenger vehicles. Such systems for transit buses must address unique aspects of bus driving; for example the bus driver must deal with the wide range of blind spots due to the dimension of the bus and long periods of driving. Since public transportation should prevent driver-caused accidents, buses may represent a natural place to develop advanced driver assistance/warning systems, which incorporate monitoring/observation of the driver. In view of this background, we propose a two-year research effort aimed towards the design and implementation of a reliable vehicle/driver monitoring system with aim to enhance driving safety of transit buses.

Research Plan, Deliverables and Contributions

To develop an effective and reliable vehicle/driver monitoring system, we will utilize all relevant signals available on the PATH test vehicle. In particular, a vision-based sensing system is a key element in the vehicle/driver monitoring system. The driver is modeled as a dynamic process; the primary input and output of the model are the look-ahead error that the driver visually sees through the front windshield, and the angle of the steering hand wheel, respectively. The look-ahead error is synthesized from the image obtained by the vision camera. The input and output are processed by the on-line parameter identification algorithm with an ARMAX (Auto-Regressive, Moving Average with eXogenous input) model, and key parameters to indicate the driver state are extracted from the identified parameters. If the driver state deviates from the normal state in a significant manner, a warning/message signal is given to driver. At the same time, the driver state is continuously combined with CWS and RDW such that false alarms and missed alarms are minimized. The major contributions of the proposed research are:

1. The research will develop a simple yet effective method to monitor the driver from the view point of driving skills and their variation due to inattentiveness and drowsiness.
2. The resulting vehicle/driver monitoring system will be able to give an early warning to the driver to alert his changing driving behavior. It will improve the effectiveness and reliability of current CWS/RDW systems, and enhance the safety of highway traffic.
3. The research is aimed to apply the system to transit buses, but it will be equally applicable to freight trucks and passenger vehicles. Thus, it will contribute to reduce accidents and congestions on California Highways.