Stephen H. Fairclough and Robert Graham The HUSAT Research Institute Loughborough University, UK
The related areas of real-time monitoring and diagnosis are central to the development of Intelligent Transport Systems (ITS). The provision of timely and relevant traffic information is dependent on monitoring the drivers' activities relative to an information infrastructure. For example, an Autonomous Intelligent Cruise Control (AICC) system must be capable of adjusting host vehicle speed by monitoring and predicting variations in the speed of other vehicles. Similarly, the provision of collision avoidance warnings is reliant on the accurate monitoring and diagnosis of an potential rear-end collision situation. The capability of technological systems to adapt appropriately and in real time is restricted by the proficiency of monitoring and diagnosis sub-systems.
The evolution of telematic systems which monitor and diagnose the presence of driver impairment (i.e. degraded driving performance due to the influence of fatigue, alcohol, drugs or illness) represents the most ambitious research and development programme in this field. The monitoring system under development on the current CEC-funded project SAVE ( Brookhuis, De Waard, & Bekiaris, 1997) employs a multi-sensor approach, where data from an array of vehicle sensors are combined with psychophysiological information on eye closure and the presence of long duration eye blinks (collected via remote means using a machine vision algorithm). These systems must be capable of a high degree of accuracy if they are to function as accident countermeasures ( Fairclough, 1997). In practical terms, system accuracy must be achieved against a background of high variability, both on an intra- and inter-driver basis and within a fluctuating driving environment. Impairment monitoring system must be capable of producing highly accurate performance on the basis of a low signal-to-noise ratio.
The aim of the SAVE project is to produce a prototype telematic system capable of diagnosing and discriminating between impairment due to fatigue, alcohol and sudden illness. The system functions on a cybernetic basis. If the performance of the driver deviates from a normative template, the system provides a warning message. The driver receives visual and verbal feedback at two levels of differentiation, i.e. a standard warning and a severe warning. If the driver is assessed to be dangerously impaired, the SAVE system activates a longitudinal and lateral control device (known as the Automatic Control Device or ACD). The ACD takes vehicular control from the impaired driver, provides an alert to other drivers in the vicinity and attempts to bring the vehicle to a safe and stationary position, The monitoring system is the decision-making component of the SAVE system architecture, providing warning feedback or automatic control on the basis of a real-time diagnosis.
It is essential that the decision whether or to evoke automatic intervention or warning feedback is based on reliable diagnostic criteria. The dilemma for system designers and researchers alike is to establish the criteria that define impaired driving in both quantitative and qualitative terms. In mental workload research, the estimation of unacceptable levels of workload is known as determination of the workload redline. It is proposed that there is little conceptual difference between the assessment of a workload redline and an impairment redline. Both are concerned with the setting of a lower limit for acceptable performance and conditions of operation.
The derivation of criteria for warnings and interventions into a classification system is a complicated affair. Under extreme circumstances where the vehicle weaves out of lane, steering input ceases and the driver slumps in the seat - it is obvious that the system should intervene. In fact, the SAVE system includes sensors to detect driver head position and grip force on the steering wheel to cover this eventuality. However, it is