This conclusion was drawn from research presented by Professor Alexandre Bayen from U.C. Berkeley. Professor Bayen demonstrated how a small percentage of connected autonomous vehicles (CAVs) programmed with machine learning could be added to existing traffic to help optimize traffic flows. This concept is labeled as a new traffic control framework called Flow. The CAVs would act as traffic control devices and could essentially replace traditional ramp meters and signals. One of the more impressive examples shared in Professor Bayen’s presentation was a simulation of how Flow could improve traffic operations through the San Francisco Bay Bridge in California (see the 2 minute mark in the video below). He was confident enough in Flow that he predicted this will be the primary method of traffic control within ten years. Of course, a part of the ‘if’ is whether researchers can convince decisions makers (and the public) that adding more vehicles to existing queues will reduce their traffic delays.

As AV manufacturers share more information about readiness and researchers investigate AV pilot performance, it is becoming clearer that hyped expectations about AVs operating in ‘revenue services’ (taxis and transportation network companies) by 2020 will be the exception and not the rule. Many challenges exist and researchers at TRB cautioned that AV expectations could quickly fall into the trough of disillusionment this year. This is all part of the hype cycle and the sooner it happens, the sooner more reasonable expectations can be established with regard to where AVs will operate, what they will carry (goods versus passengers – goods are replaceable while passengers are not), how they will influence traffic operations and safety, and when.