Michael Morscher

December 8, 2017

Don't Get Your Signals Crossed

Whenever it is that self-driving cars begin to roll out to the mass market, there will be a period of time when traditional and autonomous vehicles share the road together. In certain applications such as highway driving, there probably won't be too much of a clash. We can look to the presence of cars such as the Tesla Model 3 currently on highways today, which for all intents and purposes in this narrow use case, behave just the same as higher-level automation would. However, one clear case in which there could literally be a clash between self- and human-driven cars is intersections.

We can imagine what intersections would be like with 100% of vehicles automated, with cars zipping past each other with centimeters to spare (check out these cool simulation videos).  But with unpredictable traditional cars in the mix, this cannot become a reality. One short-term solution to this problem that can be implemented with minimal infrastructure change is the use of "priority traffic signals"—traditional stoplights that communicate with vehicles deemed to be of high priority and are accordingly granted green lights.

‍Active priority traffic signals are already used in some cities to ensure green lights for bus rapid transit.

Priority signals have already been implemented around the world for use along bus rapid transit routes without a dedicated right-of-way, having to share the road with vehicles. Take for example the gradual deployments of priority signals to the Silver Line in Boston and Select Bus Service in New York City. Of course this prioritization comes at the expense of people who have to wait a few extra seconds in their cars, but to the benefit of the many more people that can fit on a bus.

Surprisingly, there hasn't been much of anything published regarding the extension of priority signals to self-driving cars. Any analysis out there escalates immediately to the futuristic hyper-speed intersections that make for those cool simulation videos. Yet, priority signals would allow higher vehicle throughput by prioritizing cars that are able to safely go faster and coordinate to use more intersection surface area at a given time, reducing traffic jams for everyone. And, it would also help obviate the problem of autonomous vehicles battling with human drivers.

As outlined by the Federal Transit Administration, there are two types of priority traffic signals. "Active priority" occurs when a transit vehicle communicates with a light to ensure it is green when arriving at an intersection, as in the image above. "Passive priority" however is easier to implement, and simply refers to setting light timers to be green more frequently for continuous movement of slower traffic, such as buses. Take this visual for example, showing a smoother flow of prioritized traffic.

‍Passive priority signals manipulate the timing of traffic lights to favor continuous movement of "prioritized traffic" with a different average speed than other traffic.

Theoretically, passive priority could work for autonomous cars if timers are programmed to favor faster traffic, but active priority signals taking advantage of the autonomous technology is the optimal solution.

There is one final concept in existing priority signal setups that would be highly extensible to self-driving cars: queue jumping lanes. This infrastructure modification at intersections creates a dedicated lane with dedicated traffic signals to prioritize eligible vehicles at a red light before other vehicles can start moving again. Creating queue jumpers for self-driving cars would allow them to get ahead of traffic and reduce battling with traditional cars.

‍Queue jumping lanes used in coordination with priority signals help promote traffic throughput.

These simple retrofits of existing infrastructure would enhance the flow of mixed traffic in a theoretical sense, and perhaps make it possible at all in a practical sense. Best of all, the costs to implement this concept would be an affordable alternative to the inconceivably immense costs to build an entire parallel infrastructure only for self-driving cars. A DOT estimate from 2002  claims the cost is as low as $8000 per intersection, quite a reasonable distributed cost for the trillions (yes, trillions) of miles driven in the US every year.

So, what do you think? Should the industry visionaries and hype men take notice of the "here and now" and explore priority traffic signals as a means to permit mixed traffic? Or should they keep dreaming, hoping that this problem will resolve itself in time, and leaving us to deal with the mess when it doesn't? I think I've made my point to them clear: Don't get your signals crossed.

 News: A few weeks old at this point, but tangentially related. Who would have ever thought hacking a self-driving car could be done with stickers on traffic signs?

https://spectrum.ieee.org/cars-that-think/transportation/self-driving/the-scary-efficiency-of-autonomous-intersections https://medium.com/self-driving-cars/adversarial-traffic-signs-fd16b7171906 http://www.nyc.gov/html/brt/downloads/pdf/brt-transit-signal-priority-july2017.pdf https://www.transit.dot.gov/research-innovation/signal-priority http://www.itscosts.its.dot.gov/its/benecost.nsf/ID/478B21EDD18C9EAE85256DB100458929?OpenDocument&Query=CApp https://www.npr.org/sections/thetwo-way/2017/02/21/516512439/record-number-of-miles-driven-in-u-s-last-year
Michael Morscher
Engineering Psychology and Computer Science