Monday, June 26, 2017

Progress on Autonomous Vehicles

By Baruch Feigenbaum, as reported in the June 2017 issue no 164 of Robert Poole's SURFACE TRANSPORTATION INNOVATIONS

Last month I attended an automated vehicle conference at Princeton and read the Eno Center for Transportation's new AV policy paper. Both make excellent contributions.

The 2017 Smart Driving Car Summit in Princeton was created and organized by AV researcher and professor Alain Kornhauser. Kornhauser has been researching the intersection between technology and transportation for 50 years; he also writes a useful and amusing weekly AV newsletter. The summit was divided into two days. The first day featured presentations on AV technology, safety, and insurance while the second consisted of workshops on AV planning, insurance and artificial intelligence.

On the first day, AV researcher Bern Grush gave a very good presentation on how AVs could affect communities, highlighting that while most experts expect AVs to increase VMT, uncertainty about a transformative technology makes planning challenging. Adriano Allesandrini from the University of Rome gave a forceful presentation on how automated vehicles are available and on the road today; they are called buses. CityMobil 2 has demonstrated automated slow-speed buses in many European cities. The challenge is that buses traveling at 8 mph have limited ridership potential, so the technology needs to be improved before these buses are viable. He also challenged conference attendees to think about what is best for urban areas rather than what is best for car companies.

Danny Shapiro from NVIDIA presented a fascinating discussion of machine learning. He discussed how, using CAD simulation software and a graphics processing unit, researchers have been able to build a car's brain. It is the same technology used for facial recognition software, and the improvement in machine learning over the past five years has been revolutionary.

In the afternoon we had several presentations on the availability/effectiveness of AV features. Auto dealer Sheldon Sandler revealed that the two most valued AV features are a rearview camera and blind-spot warning. They are the only features that more than 60% of buyers request. Manufacturers offer advanced AV features on a limited number of models only, and they often require customers to buy a package of features that they don't want. For example, the Hyundai Sonata offers a safety package. But it costs more than $5,000 because in order to get the safety features, customers have to order the Limited Tech package with a premium audio system and heated rear seats. Manufacturers offer these packages because most consumers don't put a high value on safety features. So despite what buyers claim in stated preference surveys, car buyers are not willing to pay much for advanced technology. More depressing was a presentation from the Insurance Institute for Highway Safety that showed many of today's Level 2 AV features, such as automatic braking, don't work in many real-world situations. In almost 50% of cases, automated braking did not engage fast enough to avoid hitting a car or a pedestrian.

The Eno policy paper, "Adopting and Adapting: States and Automated Vehicle Policy," is a 30-page report worth reading. It provides an overview of AVs and offers recommendations to states. My focus here is on the recommendations section that is broken into three parts: Regulations, Infrastructure Investment and Funding, and Research and Workforce Training.

The paper suggests relatively relaxed regulations, including not overdoing reporting requirements, and creating non-binding statements or principles. It recommends states work with NHTSA on liability and safety issues. I believe this means that NHTSA should stick primarily to federal issues such as safety standards, and states should control licensing. Federalism suggests a certain balance between federal and state-level rules. I generally agree with these recommendations, as nothing that I have seen so far in AVs suggests that balance should be upset.

For state AV infrastructure investment, the report recommends focusing on keeping highways in a state of good repair. The report also recommends that automated vehicles pay mileage based user fees (MBUFs) instead of a gas tax. So far, MBUF acceptance has been slow due to the development of the technology and the political resistance to paying by the mile. Eno believes that MBUFs can be implemented more easily using vehicles with new features such as AVs.

The document also recommends funding research to explore how AVs may affect the broader economy. One recommendation suggests revising procurement processes so research keeps up with innovation. Governments seldom keep up with new technology because internal processes are designed to be slow and conservative. For research to be useful it cannot be developed at the typical government pace. Finally, Eno recommends investing in programs to train people for AV repair. Today's colleges and technical schools need to think about the skills tomorrow's workers need. And those are not the skills needed to fix a 2005 Dodge.

The Eno report has two weaknesses. The first is that it examines states only; it does not look at metropolitan planning organizations (MPOs) or cities. MPOs are the federally designated planning entity for regions; they are interested in AV guidance as well.

The second is the suggestion of digital short-range communications (DSRC) pilot projects. DSRC has a host of problems: it's outdated and inferior to 5G, it takes up valuable wireless spectrum, and it is really expensive to build the needed infrastructure. For the past 18 years, part of the 5.9 Ghz band has been reserved for DSRC, but the technology is still not ready. With the advent of 5G and the need for connected vehicles still some years off, there is simply no point to pursuing DSRC.


No comments: