How Self-Driving Cars Will Reduce Traffic Congestion
For companies such as Cruise, Waymo and others developing the latest self-driving tech, developing cooperative frameworks will have social and financial benefits
The onset of autonomous vehicles will introduce many benefits to existing road systems. Some of these will eventually include the ability to optimize traffic flows, reduce the number of overall vehicles needed, and reduce road violence. Many of these benefits, however, will require major updates in transportation infrastructure and a sea change in how we utilize automobiles. In the meantime, many are concerned that the transition will be a messy congestion-filled nightmare. While concerns are valid, there is a lot of low hanging fruit where autonomous vehicles can reduce congestion and make driving easier for everyone.
The basic problems of traffic congestion are primarily human centered. Regardless of road design, the payoff principles described in economic game theory predict, rather accurately, that car drivers will generally make choices that cause traffic congestion in “free” driving conditions. Adam Millard-Ball associate professor in the Environmental Studies Department at the University of California, Santa Cruz has produced some in-depth research into this area however, simple observations for many of us, is ufficient to see the outcome of these interactions: traffic congestion.
For example, not allowing other drivers to change lanes, or when construction and accidents reduce road capacity, and when drivers prevent cars from merging on highways, is like throwing a stone into a glass-smooth pond and watching the water ripple out. These uncooperative actions manifest as traffic congestion — most of which are the result of bad driver behavior.
For AVs the lowest hanging fruit on the tree is rule based frameworks that prevent them from obvious bad behavior such as blocking crosswalks or intersections. However, on freeways and highways properly developed cooperative interaction frameworks have the potential to significantly reduce a great deal of visible congestion.
Merging is the single most stressful activity we face in everyday driving, according to a survey by the Texas Transportation Institute(TTI). In his book aptly named “Traffic,” Tom Vanderbilt notes, even casual observation shows merges cause a great deal of traffic congestion. The zipper merge, where cars adjust speed to space evenly as the merge approaches was developed by Delft University in the 19080’s and has been proven to be an effective solution to reduce congestion, stress, and increase road safety. “The more equal use of both lanes on the approach for queuing shortened the physical extent of the queues and, in turn, reduced delay,” according to a study completed by the Transport Research Laboratory, in conjunction with the TTI.
There have been numerous challenges associated with actually getting drivers to “zipper.” One Seattle driver outlined his personal mission to reduce traffic — one of the strategies was to allow other vehicles space to merge. Another San Francisco driver attempted to decorate his vehicle to get the message out (image above). However, aside from a few good Samaritans, as Millard-Ball (and many economists) might point out, there exist very little incentive for drivers to participate. However, as AVs become more prevalent, this becomes a much more realistic task.
To get AVs to exhibit a consistent zipper merge is not all that simple. The payoff matrix described by game theory requires that an AV be able to predict other vehicle intentions, especially tricky if we are discussing a mixed AV and self-driven vehicle environment. In their research, Autonomous Vehicle Social Behavior for Highway Entrance Ramp Management, Junqing Wei, John M. Dolan and Bakhtiar Litkouhi, working with GM at Carnegie Mellon, have developed an intention Prediction and Cost function-Based algorithm (iPCB) to help AVs ‘understand’ human driver behavior.
Simply, it has the ability to both understand, and signal, intention and dynamically develop an optimal output for merge behavior. And, while it is difficult to communicate optimal value to individual human drivers, AVs can be taught that it is optimal to act in a way that is better for everyone.
The results are rather stunning. Compared to other approaches, “the safety and acceleration costs are reduced considerably, meaning the control of the vehicle is smoother and safer. The number of potentially unsafe scenarios is also greatly reduced.” Merges had significantly less speed volitility — resulting in less “ripples in the pond.” Overall implementation of the iPCB algorithm produced a 41.7% performance over other methods in simulated environments. In short, implementing frameworks like this, AVs will be considerably more effective at reducing highway congestion during merges.
For companies such as Cruise, Waymo and others developing the latest AV tech, adding socialy cooperative behaviors creates a huge social good. And, while policy makers could consider forcing these types of AVs frameworks, it would behest the companies to do it themselves to smooth out policy friction, regulatory risk, and convince the purchasing and voting public of the benefits of an AV vehicle future.
