Monday, April 18, 2016

Liang Shi: Will Driverless Cars Improve Traffic Congestion?

Many colleges and universities have a quick but tough competition among their graduate students called the 3MT or the Three Minute Thesis. Students stand up and deliver, in three minutes or less, a coherent and interesting summary of their research leading to their Masters or Doctorate degree.

I am very pleased that my PhD student Liang Shi competed among ~60 University of Hawaii at Manoa graduate students and got 2nd place.  Congratulations!


"You may have seen the news that Google is testing their Driverless cars in California. The car was designed to drive itself without any human intervention. It uses a rotating roof top light radar system (aka LIDAR) that maps critical components of the road environment, including curves, signals, traffic lanes, vehicles, pedestrians and bicycles. It also maps cones and construction zones for avoidance. It has no driver controls, such as steering wheel or pedals. It is believed that driverless cars can reduce traffic accidents and congestion. But last November, a Google car was pulled over by the police for driving too slowly, which generated my research question ”will driverless cars improve traffic congestion?” My research is to estimate how driverless cars affect the traffic flow of freeways and city streets.

For example, take an intersection controlled by a traffic light , when the traffic light turns from red to green, it takes 1 second to 3 seconds for human drivers to perceive the change of the signal and react accordingly by switching pedals. For a driverless car, it only takes 0.3 seconds to do so. Driverless cars use sensors and cameras to detect the traffic lights.  They can communicate with the local controller that operates the traffic light. For advanced driverless cars, called connected vehicles, they can talk to each other exchanging information of location, speed and other parameters. Imagine if all the cars at the signalized intersection were driverless cars: They would all start at the same time without the human response delays. So the road can serve a larger amount of traffic in the same amount of time. In addition, driverless cars are equipped with collision avoidance technology, so if a regular motorist violated the red light, the driverless car would not enter the intersection and it would avoid the collision.

Based on my analyses, if all vehicles are driverless cars, capacity of the road will be doubled. This means that every city intersection would have twice as many lanes at no cost to the city.  But if in the future we only have 1% of driverless cars in traffic, traffic flow will not improve. If we have 5% driverless cars, congestion will improve by about 5% if they are regular driverless cars, or by 12% if they are connected driverless cars. So with 5% driverless cars, traffic flow will begin to improve in a noticeable way. However, if the cars are designed to drive conservatively, they will cause more delays than humans do.

To answer my research question, which is “will driverless cars improve traffic congestion?” the answer is that if driverless cars become dominant in the market and if they are designed to drive very closely to each other, then they will reduce traffic congestion, possibly by a lot. "