Every day, automobiles equipped with increasingly sophisticated autonomous driving technologies are coursing onto the world’s roadways. There is a remarkable amount already known about the capabilities of these technologies, even in the face of unpredictable driving situations. But in addition to the road, these systems must interact with something far more complex: the driver, who is rapidly becoming a part-time passenger in his or her own car.
Since its founding in 1999, the MIT AgeLab, part of the Center for Transportation and Logistics at MIT, has studied how humans of all ages interact with in-vehicle user-interfaces and other safety technologies. Today, when an automaker's software update can populate highways with new semiautonomous technology in the blink of an eye, that work has never been more important.
Tesla’s and Volvo’s semiautonomous systems, which allow drivers to relinquish control of their vehicles for periods of time, represent perhaps the leading wave of the single greatest shift in vehicle automation since the introduction of the automatic transmission. Other key advanced driver assistance systems such as Automatic Emergency Braking, Adaptive Cruse Control and Lane Keeping Assistance are rapidly becoming standard features. Soon, such technology will likely become the norm across the vehicle fleet. What’s not yet clear, however, is how people are using these systems, whether they're using them correctly, and if there are better ways of implementing the technologies to improve safety based upon a study of actual behavior in drivers’ self-selected use contexts.
To answer questions surrounding drivers’ decisions for use of such advanced technologies, the Advanced Vehicle Technology (AVT) Consortium was launched in September 2015 in collaboration with Touchstone Evaluations, Inc. and Agero, a leader in vehicle and driver safety services. This consortium aims to develop a new and deeper understanding of how drivers leverage vehicle automation, driver assistance technologies, and the range of in-vehicle and portable technologies for connectivity and infotainment. The AVT Consortium’s membership now includes Delphi, Liberty Mutual Insurance, Jaguar Land Rover, Autoliv, and Toyota.
The AVT Consortium, which began to collect data in the Boston area this past January, is currently recruiting drivers of Tesla models S and X vehicles who are willing to have data recording technology installed in their own vehicles. This recruitment effort was recently captured in a Reddit post and has since been featured in a set of articles including Yahoo News, Tech Times and others. Present plans call for expanding the scope of the program to recruit owners of 2017 Volvo XC90s and S90s later this summer and it is expected that other makes and models of semiautonomous cars will be considered as they become available. The research effort also involves loaning MIT purchased vehicles to participants to study technology adoption, survey efforts assessing consumer preferences for technologies, and a consideration of today’s automotive dealer delivery experience. Future work may broaden the effort beyond Boston to other areas of the United States and indeed the globe.
In the study, technicians outfit participants’ cars with a variety of sensors, including three small cameras, which record the forward-facing roadway, vehicle’s instrument cluster, and driver’s face, hands and body positioning, respectively. These devices are unobtrusive and tend to blend into the cockpit environment for most drivers, becoming a part of the ‘vehicle’s background’ after a few hours of driving – allowing drivers to behave as they naturally would. Using advanced computer-vision software, researchers are then able to quantify drivers’ actions, such as how they respond to various driving situations, as well as when they eat, operate phones, have conversations, and perform other actions behind the wheel. Great care is taken to protect personally identifying information in this analysis process, and to ensure that only the meaningful patterns are extracted from the data (while protecting individual identities).
Of special interest are the fleeting, yet critical, moments when control transfers from driver to car and back again. By assessing the driver’s gaze, hand placement, body posture and position, drowsiness, emotional state, and more; and then combining those data with vehicle telemetry and secure geographical location information; researchers can assemble a more complete picture of how people and semiautonomous vehicles work together—or don’t.
The work does not stop there. The AVT Consortium is interested in generating a deeper understanding of how drivers respond to alarms (lane keeping, forward collision, proximity detectors, etc.), leverage technologies such as semi-automated parking assistance, assisted cruise control, vehicle infotainment and communication systems, smartphones and more. An important aspect of the work focuses on utilizing a fleet of MIT owned vehicles (starting with 2016 Land Rover Evoques & 2017 Volvo S90s) to assess how education can play a role in successful adoption of new vehicle systems, and how the automotive ecosystem (manufacturers, suppliers, dealers, insurance providers, etc.) work together to support drivers’ acquisition of the knowledge needed to maximize the potential safety benefits offered by many new vehicle systems.
"Our goal is to contextually understand when key AVTs are engaged, how they impact driver behavior, and their effect on vehicle control and safety. Our hope is that vehicle manufacturers, suppliers and insurance providers can utilize the data-driven insight generated to more effectively engineer products that closely consider users' actual behavior with systems in the wild, while fueling the development of more effective driver education,” says lead researcher Bryan Reimer.
“Our aim is to follow over time the technological trends of the industry, quickly gaining critical insight on the most promising vehicle technologies asthey enter the marketplace."