Conclusion

A novel approach of renewal cycle was proposed in this study as the basic element for analyzing attention allocation. This method provide in-depth insight and different view of attention allocation. The contributions of this method are described as follow.

(1) In previous studies, the path related approaches extracted the sequence of drivers transiting vision between two or three focal points. However, such a way may represent only a partial process of drivers allocating attention to multiple areas.

Moreover, it is important to identify the focal point where drivers start and end a sequence of vision transition. That is, the basic component of attention allocation must be identified. In this study, the concept of renewal cycle is proposed by anchoring the glance to forward. Considering that the vehicle were moving forward in most of time, the frontal side is the area where drivers must constantly glance. Glancing at the frontal side may also be area where drivers can most comfortably looking at. Therefore, using the concept of renewal cycle can be utilized for distinguishing the forward and non-forward glances. This approach can also transform the sequence of vision transition into several components for analysis.

(2) In previous studies, the forward and non-forward glances were either analyzed as the same focal point, or separately without considering the relation between them.

Since the drivers spent most of time looking forward, such ways may obscure the characteristics of non-forward focal points owing to the dominant forward area.

Additionally, glancing away from the frontal side may increase the unawareness of leading conditions, and thus, urge drivers to transit vision back to the front. It implies the importance of analyzing the process of drivers shifting vision away from and back to the frontal side. Since each renewal cycle contains a glance to

forward side, using it as the basic component prevents analysis being dominated by the forward glance. It also presents not only the distinction between on- and off-road glances, but also enable the analyses to observe the interaction between these two types of focal points.

(3) Based on the scan path approach, another advantages of using renewal cycle is to provide a clearer understanding of the visual transitions among focal points. The most significant paths comprise visual transitions to or from the frontal area, the most dominant focal point. This method cannot thoroughly reflect all visual transits around the vehicle. By using renewal cycles as the basic component of attention allocation, two seemingly distinct scan paths can be combined in an attention allocation pattern that illustrates the chain processes of drivers glancing at sequential focal points. In other words, a paths containing more than three focal points can be observed, such transiting vision from one non-forward focal point, to forward side, then to another non-forward focal point and finally back to the frontal side again.

(4) Drivers do not always finish the information from a focal point in a glance. As an alternative, they sometimes repeatedly transit vision between an intended non-forward focal point and the forward side for preventing long glances off-road. Using renewal cycle as the basic component of analyzing attention allocation can observe such a characteristics of repeated renewal cycle. Moreover, the duration of repeated renewal cycles may indicate the investment of mental resources in an information source. Because drivers separate lengthy glances on a focal point into several shorter successive renewal cycles, the traditional methods for analyzing the duration of each glance may underestimate the total effort expended on certain focal points. Analyzing the total duration of glances over repeated renewal cycles provides vital insight into the manner in which drivers manage information perception and/or reconfirmation of traffic conditions.

(5) Finally, modeling the process of attention allocation microscopically provided the probability of drivers choosing specific type of renewal cycle or a specific focal point. A two-layer MNL model is proposed based on the concept of renewal cycles, which are the types of renewal cycle choices in layer 1 and the focal point choices in layer 2. Particularly, this study focus on the transition between two non-forward focal points under varying driving tasks or environmental conditions. The contribution of each individual factor affecting attention demand of each focal point can be derived.

Using the naturalists glance data of 100-car dataset, this study explored and modeled drivers’ attention allocation. Some patterns of vision transition were found in this study.

(1) More than 90% of the renewal cycles identified in this study contained only one glance away from the forward direction. Moreover, the probability of transiting vision from one non-forward focal point directly to another decreases with the time that drivers has glanced off-road. Particularly in the complex tasks and worsen environmental conditions, drivers tended not to glance at more than one non-forward focal point in a sequence. This result suggests a cautious behavior that drivers would prevent long glance off-road by glancing only one non-forward focal point before transiting vision back to the front. Additionally, transiting vision away from the frontal side would urge drivers to transit vision back to the front for compensating the lost awareness against leading traffic.

(2) As expected, maneuvers that entail different task loads create distinct patterns of attention allocation. Moreover, the drivers exhibited patterns of transiting vision to the roadside or to in-vehicle devices to gain non-driving related information less frequently when they were busy performing maneuvers with higher task loads. This finding suggests the existence of compensatory behavior to prevent crashes by allocating increased attention to where the risk is increased (Liu and Lee 2005, Törnros and Bolling 2006).

(3) Nevertheless, in some risky situations, such as driving under LOS D/E, InvD were found to occur most frequently among all non-forward focal points. Drivers who overestimated their ability to handle both distraction activities and driving tasks placed themselves at increased risk of having a crash, especially under poor driving conditions. Hence, managing distraction is clearly vital for improving driving safety. Detailed analysis of distracted behaviors and their implications for designing effective information systems warrant further research.

(4) A large proportion of these cycles occurred successively and repeatedly; that is, the drivers may separate a long glance at one focal point into several repeated short renewal cycles. This finding supports the hypothesis that shifting attention away from the forward area decreases the driver’s awareness of the traffic ahead.

Thus, the sample drivers generally avoided looking away from the forward area for lengthy durations.

(5) Among all non-forward focal points, in-vehicle distraction and rearview mirror related renewal cycles are the two that drivers would more likely to repeat. It supports the previous findings that the in-vehicle distraction contained complex

information needing to be processed, and the rearview mirror represented an usual mean for constantly checking the rear side traffic (Metz et al. 2011). These repeated renewal cycles would be more evident when driving in high speed.

Again, the higher speed may represent shorter time to collision against the leading obstacles, which induce higher expectancy. Drivers would be more eager for transiting vision back to the frontal side when driving fast. Therefore, repeatedly transiting vision between rearview mirror and in-vehicle distraction may be a compensatory strategy for balancing the attention resources allocating to on- and off-road areas.

(6) The renewal cycles of rearview mirror and in-vehicle distraction are repeated for different purpose. This study found that the drivers would tended to glance these two non-forward focal points longer in repeated renewal cycles. The duration of forward glance in rearview mirror does not vary between individual and repeated renewal cycle. However, drivers would decrease the duration glancing on-road when repeated transiting vision between in-vehicle distraction and forward side.

This distinction implies different purpose of renewal cycles. One is for constantly checking the specific area, such as rear side through rearview mirror, where potential threats may appear. Drivers’ main effort is still invested on the frontal side. On the other hand, the other type of repetition is for dividing a long glance on certain information source into several shorter glances. The in-vehicle distraction is fall into this category. Drivers would spent more mental resource for gathering and comprehending the information from the distraction.

Meanwhile, to compensate the lost awareness against leading traffic, they would transit vision back to the frontal side shortly.

(7) Regarding the different types of renewal cycle connecting two non-forward focal points, the probability of drivers transiting vision directly to from one non-forward focal point another was the lowest and decreased with the duration being glanced off-road. This result supports the notion of drivers being more stressful when glancing at more than one non-forward focal point in a renewal cycle.

(8) Drivers barely transited vision directly to other non-forward focal points after glancing at rearview mirror and in-vehicle distraction. It may be related to the heightened level of mental resource required for identifying the image through a small mirror reflection or reading the information on the in-vehicle distraction.

Such a tense glance may increase drivers’ uneasiness more evidently than glancing elsewhere. By contrast, after glancing at left or right side, drivers may feel comfortable to glance another non-forward focal point. Yet, in the

conditions of high density or high speed, the probability of transiting vision directly to other non-forward focal point would decrease, which is related to higher expectancy level of traffic on the frontal side.

(9) Through modeling the attention allocation process, this research is able to extract the path of shifting vision from one non-forward focal point, to forward side, and then to another non-forward focal points. In this type of vision transition, the most frequent occurred path observed in this study is transiting vision from one side of a vehicle to another side through a glance to forward area. Drivers would offset the effort of transiting vision across the vehicle by connecting two renewal cycles. As stated in Underwood et al. (2003b), such a combination can be observed in an experienced driver’s scan path, while a novice may directly transit vision from right to left. The renewal cycle concept can help in grouping these distinct paths for better interpreting the pattern of attention allocation. Yet, even though this type of vision transition requires less effort than the one of directly transiting vision from one side to another, it is still a driving tasks with higher effort. While driving in a mentally demanding scenario, such as heavy traffic, drivers would decrease the vision transition across the vehicle, no matter in the form of a direct path or a path containing two renewal cycles.

(10) This study analyzed the duration of each forward and non-forward focal point based on the numbers of glances in a renewal cycle. Applying such a method could identify the duration changes against when drivers are aware of the possible long glance off-road. It is found that drivers would alter the duration glanced at forward and non-forward focal points based on the clues obtained from driving tasks and environment conditions, especially the duration of forward glances in 3-glance renewal cycles. Drivers would glance at forward side shorter in 3-glance renewal cycles than the 2-glance ones, suggesting the drivers’ alertness against possible long off-road glances. However, the duration of off-forward glances did not decrease with the number of off-road glances. It shows that the off-road glances would increase dramatically when drivers intended to look at more than one non-forward focal point in a sequence.

(11) This study, incorporating the potential time loss of drivers not gazing forward, developed a desired PRT for examining the robustness of the current PRT rule.

The results revealed that drivers in certain conditions probably have insufficient time to perceive information, form decisions, and initiate reactions. As expected, attributes including maneuver intentions, distraction, and traffic density were found to have significant effects on the durations of forward or non-forward glances. Degradation of safety resulting from attention allocation with two or

more non-forward glances in a renewal cycle is substantial. Seeing that adopting inappropriate renewal cycles evidently deteriorates safety, improving drivers’

information searching skills can be beneficial in crash prevention. Drivers must develop defensive driving skills to observe surroundings properly and efficiently, especially when driving in high risk conditions.