A new paper presented by two University of Massachusetts Amherst computer science graduate students and their professor might hold the key to utilizing the strengths of both semi-autonomous vehicles and humans for a safe traveling experience. 



In a presentation at the International Joint Conference on Artificial Intelligence in New York on July 12, Prof. Shlomo Zilberstein and his students Kyle Wray and Luis Pineda presented an idea on how to seamlessly transfer control from a semi-autonomous vehicle to a human driver. This transfer of control is meant to be done during instances where the vehicle may not be able to properly judge a difficult set of circumstances as well as a human being could. 



Based off tests conducted through a driving simulator, Zilberstein, Wray and Pineda suggest high-level route planning that includes the occasional need for a transfer of control from the system to its driver. A “high-fidelity” model manages the transfer of controls, the awareness of the driver and will allow for stoppages of a vehicle in situations where a driver does not respond to the request to take over control. 



Through their experiments, Zilberstein, Wray and Pineda say that there will always be a “live state” for vehicles and also make sure that a system will never be put under the control of an individual that can’t handle the situation at hand. The three will work in collaboration with other professors and a postdoctoral fellow to integrate their findings using a large-scale realistic driving simulator. So far, their work has garnered support from both the National Science Foundation and the auto industry. 



The research presented in this paper was sparked by a reality that most autonomous systems aren’t yet capable of handling certain real-life situations. For Zilberstein, the days of a completely autonomous system are not in the near future.  



“Self-driving cars are coming,” says Zilberstein, “but the world is fairly chaotic and not many autonomous systems can cope with that yet. My sense is that we’re pretty far from having fully autonomous systems in cars.”



While the findings in this paper were used while experimenting with vehicles, they do have the potential to be applied to all autonomous systems going forward. 

<< Back to the News