In the fall of 2016, the US Department of Transportation announced new policies and initiatives for autonomous vehicles (AV) and AV research. The new Federal Automated Vehicles Policy is based on the US DOT’s view that automated, autonomous vehicles can help promote safety, mobility, sustainability. With the increase use of AVs and semi-autonomous vehicles, there are some potential safety concerns as well, including relating to the ability of people using such vehicles to respond to potential hazards and potentially hazardous situations.
Siby Samuel, PhD, a UMTC Research Affiliate in Industrial Engineering at the University of Massachusetts-Amherst, and colleagues, including Shlomo Zilberstein in the Computer Science Department, have been studying the topic of semi-autonomous vehicles and safety for a number of years. Their research has focused on situations where the control of driving transfers to the vehicle in uncomplicated driving environments (such as a limited access highway), but where drivers still need to be prepared to take back control of the vehicle to address potential hazards that arise. This level of driving automation is known as Level 3 automation. Zilberstein and two of his graduate assistants, Kyle Wray and Luis Pineda, are researching how to transfer control “quickly, safely and smoothly back and forth” between the system and the person operating it. All of these studies were conducted on UMass’s Advanced Driving Simulator ( http://www.ecs.umass.edu/hpl/ ).
“The real trend in artificial intelligence is to build systems that can collaborate with people,” Zilberstein said. (Daily Hampshire Gazette)
At the Transportation Research Board (TRB) Annual Meeting in January 2017 Dr. Samuel’s team presented two recent studies on Level 3 driving automation and the time it takes for drivers to be able to respond to potential hazards when the driving control of the vehicle needs to switch from the automated system to the driver.
An earlier study by Samuel and Zilberstein also looked at this transfer of control on the driving simulator. Participants were instructed to transfer control to automation upon hearing an audio alert “transfer control”, and then later they were told with another audio alert “take over control” when they were to resume manual control of the vehicle. During the automated driving phase, participants were instructed to do tasks on a computer tablet. This study found that the minimum transfer of control altering time required for drivers in a Level 3 driving environment to respond to a potential hazard was 8 seconds when the hazard was expected, when the roadway environment was not changing during the transfer of control process, and when they were doing tasks on a computer tablet during the automation part of the drive. In other words, it took 8 seconds for these drivers to anticipate hazards at a rate equivalent to that of drivers who were manually driving their vehicles and weren’t distracted with in-vehicle tasks. In one study presented at this year’s TRB meeting, Samuel and colleagues found that more informative audio alerts, for example a message telling a participant about at at-grade rail crossing or a lane reduction ahead could reduce the needed time for participants to respond to a potential hazard by as much as 40% or 4 seconds.
UMass Affiliate Researchers make headlines on driverless cars:
http://www.gazettenet.com/University-of-Massachusetts-researchers-study-how-to-make-self-driving-cars-safer-3711488 (Daily Hampshire Gazette)
Written by Tracy Zafian, UMTC Research Fellow