NJIT Expert Evaluating Self-Driving Car Behavior at Yellow Lights

There's Musk-level hype about autonomous vehicles, and then there's NJIT Associate Professor Joyoung Lee working through painstaking and vital research of how self-driving systems should behave at stop lights.

He studies transportation systems as a whole, not self-driving cars specifically, but his current research makes traffic lights broadcast their intentions so artificially intelligent vehicles can decide whether to brake, maintain speed or accelerate — as the function of a yellow light is to clear the intersection, not make you screech to a halt.

"That's our low-hanging fruit," said Lee, from the Department of Civil and Environmental Engineering. "I'm a transportation engineer. I'm not interested in self-driving cars. I'm interested in what if X percent of cars have self-driving technology, then what will happen in a transportation system?"

Lee started by leading an effort in 2019 for the Department of Civil and Environmental Engineering to acquire a hybrid Ford Fusion and to have Detroit-area Dataspeed Inc. install a self-driving kit, while also writing his own software and obtaining permission from the Newark government to install wireless data technology at traffic lights around University Heights.

He explained that the Society of Automotive Engineers rates self-driving cars on a scale from zero to five. A car without any autonomous technology, not even basic cruise control, would be a zero. Common modern features such as adaptive cruise control would rate at 1.5. A full-featured Tesla would be at 2.5, as is the Dataspeed-equipped NJIT sedan. Higher-rated vehicles aren't yet for sale in the U.S., while a car that rates 5 would be barely recognizable, sans a steering wheel or pedals.

Currently, the NJIT car uses 4G cellular technology and an automotive connection called Dedicated Short Range Communication, but Lee said his team is planning more tests with wideband 5G connections using equipment provided by Verizon that is not yet widely available. Faster wireless connections will have less delay, which is urgent when there's a two-ton moving object at stake. (The Ford Fusion was selected because hybrid cars have more precise braking control than cars with conventional systems, Lee noted.)

Of course, just because a car is smart doesn't mean it is sensible. One school of thought is that self-driving cars are safer than human-driven cars, because self-driving cars are not distracted by listening to the radio, reading messages or checking themselves out in the mirror. Another is that artificial intelligence will never master human common sense, such as not confusing a tipped-over traffic cone for a stop sign, which Lee said actually happened once and led to a car stopping in the middle of a highway.