Driverless cars. It’s the hot topic now, a media darling. Any online search of the keyword “autonomous cars” will reveal a slew of online articles that could take a good week to sort through, from Wired and Jalopnik to this. Such autonomous vehicles seem within our grasp…until you stop to think of the myriad issues related to near-term use of such systems. For instance, if some autonomous systems are programmed to relinquish control to the driver at the touch of a steering wheel, how much force must be placed on the wheel to deactivate the autonomous mode? Could the driver find himself/herself in a surprise situation during which an unintentional force is placed upon the steering wheel, leading to an accidental deactivation of the autonomous system? On top of the technical aspects and human factors issues related to autonomous vehicles, what are the chances that all drivers will want such a vehicle? It’s doubtful most Top Gear enthusiasts will want to cede driving to an autonomous system. That’s why, as I wrote in Part 1 of this blog series, manual driving will not be going away anytime soon. It’s more realistic to acknowledge that manual driving will continue to be a part of the transportation system. What we need to do, then, is learn how to safely and effectively incorporate automated vehicles into the existing transportation system. This is what several VTTI researchers are working on today.
Because we recognize that vehicle automation is a viable opportunity to further enhance driver safety, mobility, and environmental sustainability, we have a dedicated team of researchers within our new Center for Automated Vehicle Systems actively working with such transportation industry leaders as GM, Google, Meritor WABCO, Bendix, and other major automotive companies and suppliers to assess driver acceptance of and reaction to automated systems. (These VTTI researchers are also teaming up with Virginia Tech-based initiatives, including the Hume Center and the Virginia Center for Autonomous Systems, to study other automation issues, such as cybersecurity.)
As I’ve pointed out, automation itself presents a number of human factors challenges. In short, the drivers need to understand the capabilities and limitations of an automated system and their roles and responsibilities as the drivers of such vehicles. For example, how will a driver assume control from an automated system if there is a failure (this handoff from the system to the driver and back to the system, as feasible, is what we call the “transition”)? Will the driver really monitor the situation with enough vigilance to even know that the system has failed? The answer is a qualified “maybe not.” Our early work into partial automation shows that the driver is not nearly as vigilant as he/she would be in manual driving, often taking his or her eyes off the road for 10 seconds or more at highway speeds. These results lead us to conclude that the automated systems have to be robust enough to handle all failure modes without driver intervention until the drivers have enough information and time to make that transition. Let’s face it, that’s a large hurdle to overcome.
VTTI believes that, to surmount these human factors issues, there must be onboard sensors, vehicle connectivity, and driver state monitoring. Such a combination will help achieve the necessary levels of robustness and redundancy required to successfully integrate automated vehicles into today’s transportation system. Institute researchers are assessing each of these factors using our naturalistic driving study technique and our substantial work into connected vehicles as led by the Center for Advanced Automotive Research and the Connected Vehicle/Infrastructure University Transportation Center. In fact, connected-vehicle technologies naturally enable automation.
Connected vehicles are equipped with the ability to instantaneously share information between themselves, other devices, and the infrastructure. The same technology can provide enough real-time data culled from the surrounding roadway environment to help inform vehicle systems to perform driving tasks in an automated system. Automated features such as lane-keeping assist and adaptive cruise control are already available in newer vehicle models and are based off of sensors such as radars and in-vehicle cameras that assist in “reading” the driving environment. However, as all of us who have been working in advanced automotive technology know, these sensors are far from perfect. They can miss critical obstacles due to factors such as lighting, weather, or radar signature. (Have you ever considered the radar signature of a couch, concrete block, or mattress on the freeway?)
Connected vehicles provide both redundant and independent information. Using the above example to clarify, vehicles braking or swerving to miss an obstacle, or a traffic operations center broadcasting the presence of an obstacle in a certain location, adds robustness and redundancy to the overall system. This process is independent of the more obvious redundancy present in vehicle-to-vehicle applications where an imminent crash warning can be verified between the connected communications and other sensors. Essentially, it is our opinion that automated-vehicle systems will be dependent upon, or at the very least much improved by, connected-vehicle technology.
The work VTTI is undertaking into next-generation vehicular technology is allowing the transportation industry to ensure that the integration of connected and automated systems is relatively seamless and safe. Because we are working with local, state, and national agencies and private companies—including auto manufacturers and suppliers—to enhance such technology while simultaneously building a knowledge base about next-generation vehicles and driver behavior, we are laying the foundation for a stronger and safer future that addresses our overall goals: saving lives, saving time, saving money, and protecting the environment.