The objective of Task 6C (Telematics Demand) is to develop a means to identify cognitive demand resulting from in-vehicle device use while driving. This information will complement measures of driver state as an input to moderate information interaction with the IVIS and mitigate distraction.
Task 6 focused on developing a model to predict IVIS demand in real-time, providing input for the distraction mitigation module being developed for Task 11 that uses the strategies identified in Task 4. The model integrates information describing the state of the IVIS and current glance location to calculate current demand and estimate future demand over the coming 3-5 seconds. This project involved three elements, first was a simulator experiment to assess the demand associated with long and short IVIS tasks. The second was the development of dynamic programming algorithm of IVIS messages, with the goal of creating an optimal sequence of IVIS messages to manage the flow of messages to the driver. The third element developed an initial model of IVIS demand.
The simulator experiment examined the effects of cognitive load on driving performance for interactions with an in-vehicle information system (IVIS) that varied in duration from one to four minutes. Twelve participants drove in a simulator while intermittently performing the IVIS task. There were three IVIS conditions: interacting with the IVIS, non-IVIS periods between IVIS interactions, and baseline driving without the IVIS task. Contrary to our hypothesis, driver response to lead vehicle braking was surprisingly uniform across IVIS conditions. IVIS interaction did undermine driver ability to detect the bicyclist along the side of the road, and some of these performance decrements persisted after the IVIS interaction had ended. Reaction time for bicyclist detection increased from the first to the subsequent minutes of the interaction. Eye movements were influenced by the IVIS conditions but not by task duration. Both ANOVA and factor analyses revealed that some of the changes in eye movements were concurrent with IVIS interaction while others persisted after IVIS the driver completed the interaction. Overall, the findings suggest that two mechanisms might account for the distraction-related performance decrements in this study: competition for processing resources and interference due to activation of competing goals
The dynamic programming algorithm addresses the issue of how to schedule the demands associated with messages over time. IVIS present an array of messages that range from collision warnings and navigation instructions to tire pressure and e-mail alerts. Currently the number of messages is modest, but as in-vehicle technology become more mature the number could grow substantially. If these messages are not managed properly, the IVIS might fail to provide the driver with critical information, which could undermine safety. In addition, if the IVIS presents multiple messages simultaneously, the driver may fail to attend to the most critical information. To date, only simple algorithms that use priority-based filters have been developed to address this problem. This paper presents a dynamic programming model that goes beyond the immediate relevance and urgency parameters of the current SAE message scheduling algorithm. The resulting algorithm considers the variation of message value over time, which extends the planning horizon and creates a more valuable stream of messages than one based only on the instantaneous message priority. This method has the potential to improve road safety because the most relevant information is displayed to drivers across time, not just the highest priority at any given instant. Applying this algorithm to messages sets shows that scheduling that considers the time-based message value, in addition to priority, results in substantially different and potentially better message sequences compared with those based only on message priority. This method can be extended to manage driver workload by adjusting message timing relative to demanding driving maneuvers.
The final element of this project involves creating a preliminary model of IVIS demand. The approach differs from the typical consideration of IVIS interactions, which considers IVIS tasks as secondary tasks that interfere with the primary task of driving. With our approach, driving and IVIS tasks are considered mutually interrupting, whereby either suffers from shift of attention to the other. To predict IVIS demand, the resource demands of the IVIS task, both while it is being performed and when it is interrupted, must be viewed as competing for the resources needed for driving.
The benefit of the model is particularly great in situations when sensor data is imperfect, such as when eye position data is noisy or missing. Estimating IVIS demand based only on driver state depends on the availability of eye and sensor data. Sensor noise or intermittent failure can greatly undermine estimates of distraction based solely on driver state information. Another benefit of the IVIS demand model is that it can produce leading indicators of distraction rather than producing lagging indicators., For example, a combining eye movements over time will necessarily produced delayed indicators of driver state, whereas the IVIS demand model predicts IVIS demands for several seconds into the future at the time the driver first interacts with the IVIS..
The preliminary validation of the model show promise for the approach. The simplicity of the model makes it a viable candidate for future production vehicles. All of the input data are readily available through current CAN systems and eye monitoring system and the computational requirements are modest.