Department of Mechanical and Industrial Engineering
University of Massachusetts
The forty-six “glance points” represented in this graph illustrate eye gaze tracking during driving. (Now!) Dr Anuj Pradhan has been crucial in co-developing the RAPT novice driver training in risk perception over the course of a six-year doctorate degree and four experiments. Risk Perception and Awareness Training combines simulation and field techniques for assessing new drivers’ scope and skill in anticipating potential risks while driving.
Did you know?
- Car accidents are the leading cause of death for teens in US
- Teenagers, during the first six months of driving, have an eightfold increase in the risk of dying in a car crash
- Teenagers, in general, are four times more likely than older drivers to die in a car crash
- In numbers: teenagers are involved in 4.7% of the six million crashes annually in the US but compose 13% of the fatalities
Previous research has identified three main causes of teenage accidents, including failure to adjust speed appropriately to conditions (20.8%), failure to maintain attention to the task (23%), and – the biggest – failure to conduct an appropriate search of the driving environment (42.7%).
After his presentation, Dr Pradhan’s Dissertation Committee gave him some grief about the distinction he wants to draw between “tactical scanning” and “strategic scanning.” (They also asked him, right at the beginning, to take off his suit jacket and relax. This may have been the signal that they planned to heat up the room…!) The first question, however, came from one of the faculty during the presentation, and it involved clarifying the dependent variable of eye movement. Dr. Pradhan’s first experiment established a correlation between the recognition of risk (seeing it) and the knowledge that risks may be present (use of eye gaze to scan in order to identify (i.e. see) them if they are present).
Two more experiments refined the technique for linking eye movement with perception and recognition of risk. Results from the three experiments indicate improvements in visual search behavior in all driving situations, from the benign – when no risks are present, to situations with a minimal possibility of risk, and on up to situations with obvious dangers.
In other words, the students and volunteer test subjects who participated in these experiments learned about the strategic need for constant maintenance of visual attention across the broad driving environment which might require the driver (i.e., me – or you!) to engage in specific tactical behaviors in order to reduce risk – or be able to implement evasive action should a risk materialize because one has seen it in time! My contribution came with the fourth experiment, I got to test out the version in development – my experience (as an “older driver,” grin) may or may not have aided in refining the program, but it certainly reinforced for me that there is a purpose to where, when, and why I look and watch in the ways that I do while driving. (I learned that I could still do better!)
The need for this kind of training tool in driver’s education programs everywhere is immediately and obviously apparent. I was also fascinated by the application of temporal and spatial algorithms to the eye movements captured by the Mobile Eye movement tracker. Time and space coordinates for every eye movement had to be combined and crossreferenced in a Fixation Identification Algorithm with prior and subsequent eye movements in order to define a glance. These glances are then superimposed on the objects in the driver’s visual range, and categorized as on-road or off-road. In this way, the Mobile Eye Tracker pinpoints whether the driver’s eye looked directly at the truck parked on the side of the road in front of a passenger crosswalk, when (from near or far), and for how long. Does the gaze return or simply pass on to other objects?
In other words, the direction of eye gaze can indicate the driver’s perception of risk – or lack of it. Once a driver is informed of their own eye movement behavior, then their awareness of risk is enhanced (or should be, I think the larger research program of the Human Performance Lab is lacking a necessary qualitative element). In fact, after training in the tactics of using visual scanning to perceive the possibility of risk, Dr. Pradhan shows that drivers improve risk awareness in four significant ways:
- Trained drivers maintain a wider horizontal range of vision
- Trained drivers shift half their glances offroad, more trained looking to right – where more risks presumedly originate (compared with the untrained who look left & right more-or-less evenly)
- Trained drivers glance off-road for slightly longer times (presumedly considering the extent to which the conditions in sight compose/obscure a risk or not)
- Trained drivers learn not only to transfer recognition of risk types between similar scenarios, but also transfer the skill of tactical scanning to different scenarios than those they were exposed to during training
Throughout the presentation, I kept thinking, “if only” – if only I had had this knowledge five years ago — the language of “visual scanning,” “risk perception,” and “risk awareness” — then Hunju’s driving practice might have gone more smoothly for both of us!
Anyway, Anuj’s defense rolled along. Dr Krishnamurty pressed him on the relevance or distinction between top-down and perspective views, which Dr. Pradhan handled with aplomb: “I got you, excellent answer.” No wonder Jeff calls Anuj, “my Yoda.” The (self-named) Curmudgeon wouldn’t let go of the tactical/strategic distinction but I wager this is merely ground for the next stage of hypothesis testing and theory building. The Committee Chair, Dr Fisher, supported Anuj throughout. They grilled him for a mere quarter of an hour after kicking out us observers (selected members of the fan club). And then they only made him wait for about that much longer (or less) before Dr Fisher came out and ushered him back in with a handshake and announcement:
The Younger Driver: Risk Awareness and Perception Training, Human Performance Laboratory, UMASS Amherst
Using Eye Movements To Evaluate Effects of Driver Age on
Risk Perception in a Driving Simulator by Anuj Kumar Pradhan and five others
glance, Merriam-Webster Online Dictionary
Fixation-identification in dynamic scenes: comparing an automated algorithm to manual coding, Proceedings of the 5th symposium on Applied perception in graphics and visualization
Driver’s License, Reflexivity