June 19, 2019

As we near the end of this decade, the reality of autonomous vehicle (AV) has never felt more tangible. Low levels of autonomy (i.e. driver assistance) have been commercially available for over a decade and most vehicle manufacturers, tech companies, and market research firms are publicly predicting Level 4 autonomy within the next two to four years. There is certainly an argument on the veracity of those opinions, though taking advantage of inevitability of their introduction would be beneficial, especially in urban areas where conflicts with other modes of transportation will be commonplace.

Concurrently, people who walk and cycle are increasingly at risk of injury and death on roadways. Fatalities resulting from collisions between vehicles and people who walk or bicycle have increased by 32% from 2008 to 2017, meanwhile total traffic fatalities have decreased slightly by 0.8%. Non-motorist fatalities now account for over 18% of all traffic fatalities.

The introduction of autonomous vehicles puts a fresh perspective on a problem that has existed for over a century: how do we better allocate space in the transportation network that improves the efficiency and safety for all modal users? Investing in better infrastructure can reconciling this problem in a way that benefits today’s drivers, future AVs and vulnerable road users by. For example, clearly delineated separated infrastructure can provide clear right of way, safer separation, and uniform physical indicators for AV software to interpret. Mitigating these potential conflicts through urban design benefits both AV designers and the other users their product will share the cities with.

AV manufactures have said that uniform infrastructure will allow for speedier development. Clear lane markings, visible signals, and legible signs are top priorities that enable consistent coding and limit sensor confusion. Using a variety of cameras, sensors, and radar, automated driving systems use algorithms to gather information about its surroundings, interpret it, and then perform appropriate maneuvers based upon the data. For example, lane keeping assist features are heavily reliant on clear lane markings, which is why AVs perform poorly in inclement weather and construction zones. By standardizing lane marking, and in many cases install them in the first place, AVs can more successfully navigate cities.

AVs can further benefit by using permanent infrastructure components from dedicated separated infrastructure, rather than less permanent paint, to guide vehicles. The life span of a curb feature far outlasts that of road paint, and the raised geometry is easier to identify than lane markings. The separation of infrastructure also increases predictability, for both humans and AVs. By increasing predictability of behavior, AVs are less likely to have to interpret a complex situation which would require a corrective maneuver.

For people who cycle and people who walk, redesign of infrastructure can bring about changes in the built environment that will dramatically increase their safety, even while AVs are still being developed. The implementation of dedicated and separated infrastructure for non-automobile users of city throughways will diminish potential conflicts. Unsafe infrastructure continues to be an impediment to cycling modal shifts, for long term viability of both AVs and sustainability, infrastructure deficiencies must be addressed.

Meanwhile, uniform safety standards and liability protection creates precedent for AV software developers to prioritize avoiding collisions with people who cycle and walk. Issues of liability and insurance requirements with AVs is an ongoing area of debate, though this presents an opportunity for fixing outdated and ineffective laws regarding collisions between vehicles and people who walk or bicycle. AV software will be programmed in accordance to adhering with the law as it is currently written and interpreted. Transitioning to laws rooted in vulnerability, which is more in line with international best practices, would shift the way that AVs treat vulnerable road users.

For example, current US laws generally treat bicycles as vehicles and treat collisions much in the same way as two cars colliding. While AV manufacturers certainly do not want their vehicles to collide with humans, as it stands they have a fair amount of legal protection if they do. Instead, foreign traffic regulations such as Section 193(1) of Ontario’s Highway Traffic Act or Article 185 of the Dutch Road Safety Act of 1994 require that, “As a driver you are liable when you crash into a cyclist.” Putting the onus on the vehicle driver, whether that is a human or machine, creates clear precedent that the safety of vulnerable road users is paramount and AV software can be coded appropriately to meet this directive.

The opinions expressed above are those of the author and do not necessarily reflect those of the Eno Center for Transportation.