Thinking Beyond the (Autonomous) Vehicle: The Promise of Saved Lives
Colleagues Michael R. Boswell and William Riggs continue the "Autonomous Future" series, exploring the issues and opportunities presented by of autonomous vehicles (AVs).
At a number of recent industry meetings we have attended regarding autonomous vehicle (AV) technology, one trend is clear—most discussions focus on the technology of the vehicle but not the factors outside the vehicle. Factors outside the vehicle include: pedestrians, bicyclists, roads and intersections, signals and signage, and the city or countryside. The U.S. federal government's recently released policy statement on AVs provided some broad policy on autonomous vehicles, but did not link to ideas about smart cities, transportation infrastructure, and built environment policies. Such policies are important, and largely controlled at the local level. The highlights of the federal policy statement include clarity on definitions, information on safety and security, and standards for operation, but the statement does not move beyond the vehicle itself.
While vehicle technology is important (given the complexities of high definition mapping, digital detection, vehicle decision-making and control, and the system redundancies that make autonomous driving possible), focusing on technology in isolation omits how transportation and land use policy can and will change in response to this new form of driving. We believe these topics are important and that they need parallel consideration to the technical design and adoption of AV technology. Without that kind of holistic approach, all the talking points of the current discussion about AVs miss a critical point: the promise of saved lives.
The State of AV Policy
As mentioned above, so far there is limited guidance on AV policy. The recent federal guidance primarily focused on establishing standards for technological advancement without stifling innovation. Most significantly, the policy formally adopts a common language, establishing SAE International levels of automation 1-5 for discussing AV technology. The policy adopts these definitions with a condition that clusters levels 0-2 and 3-5 based on level of automation. The rationale is that in levels 3-5, the automated system "is primarily responsible for monitoring the driving environment (10)," which the statement defines as a "highly automated vehicle" (HAV).
In should be noted that this policy does not address level 3 systems, which require the human driver to take back control in certain situations. For example, the expectation that an HAV can "perform the complete driving task… without any expectation of involvement by a human driver" is hotly debated by industry professionals, as many believe level 3 technology does not have this capacity except for in a very limited operational domain. Problems such as "failed handoffs" and disrupted situational awareness make it likely that this intervening step should be skipped (Lipson & Kurman 2016).
This idea of the operation domain is also formally defined in new policy—in that AVs need to have an Operational Design Domain (ODD) that defines conditions for operations (e.g., roadway types, geographical location, speed range, lighting conditions, day and/or night, weather conditions, and other operational domain constraints"), as well as an Object and Event Detection and Response (OEDR) system that defines operation under normal driving scenarios (e.g., unexpected events, hazards, etc.) While these policy statements are important, there is very little information or policy related to the built environment in the work that has been completed thus far—offering opportunities for exploration and policy development in the future. For example, bicycles are not mentioned in the current guidance on detection, underscoring the importance of our recent manifesto calling for consideration of bicyclists.
Another key area of existing AV policy is safety. The recent federal guidance extends the reporting expectations for collisions and NHTSA crashworthiness standards to HAVs "regardless of the effectiveness of crash avoidance." Linked to this notion of crashworthiness is an emphasis on consumer and operator education to cover the responsibilities of HAV occupants, particularly in emergency situations (like a Zombie apocalypse, of course). The report recommends consideration of on-road or on-track hands-on experience demonstrating HAV operations as well as "innovative approaches," such as virtual reality simulation. Likewise, there is ample focus on minimizing risk of cybersecurity threats—although with limited operational definitions about how that should occur, other than being systematic and nimble.
Last, but definitely not least, current guidance outlines state responsibilities for testing, but more importantly for aspects related to the handling of motor vehicle regulations. Most succinctly these include: licensing; law enforcement and safety; and liability and insurance. In the realm of licensing:
'Licensed drivers are necessary to perform the driving functions for motor vehicles equipped with automated safety technologies that are less than fully automated (SAE Levels 3 and lower)…. (and that) Fully automated vehicles are driven entirely by the vehicle itself and require no licensed human driver (SAE levels 4 and 5)'. This is clearly linked to law enforcement since many safety-related or preventative enforcement measures are no longer as relevant—however responding to eventual crashes is emphasized as a key state / local law enforcement policy. The report suggests that adequate education be provided to support first responders since they may encounter new hazards related to these vehicles which 'may include, for example, silent operation, self-initiated or remote ignition, high voltage, and unexpected movement.'
Aspects of liability are currently held by the states, but there is a policy recommendation to consider how liability is allocated among "owners, operators, passengers, manufacturers, and others when a crash occurs." That recommendation is linked, in part, to ethical considerations of how life and death outcomes may be determined by programming rules (an area other Planetizen bloggers, and our academic colleagues Patrick Lin and Ryan Jenkins, have been giving attention). Given the complexity of those ethical quandaries, and the desire for such decisions to be made "consciously and intentionally," current policy suggests the creation of a commission to study liability and insurance that can make recommendations to states.
So where does this leave us with regard to policy? Arguably, we're left with a lot of work to do. Rules about roadway design and engineering, transportation economics and roadway pricing, streetscape standards, and the land use and environmental implications are still absent. Such rules are of interest to planners and engineers because cities are becoming smarter. As exemplified by the U.S. Department of Transportation's Smart Cities Challenge, efforts are afoot to make roads more dynamic—to disseminate better pricing and information from the built environment. This parallels massive open data movements in many cities across the country, where in-ground sensors, meters, gates, and mobile applications can inform our knowledge of transportation systems. One example is the massive parking sensor project in San Francisco [pdf] to provide dynamic price and availability information for parking spaces to users. Another example is lane technology pricing and capacity and flow tracking, currently in use by cities like London for congestion pricing. A final example is the outright banning of cars from urban centers.
Similarly, Silicon Valley companies like Lyft and Uber (i.e., with UberPool and LyftLine, respectively) are competing to harness the power of mobile phones to connect travelers to carpooling options. The byproduct of that competition is cheaper and more convenient travel and carpooling options for consumers in previously untapped markets (for example, Las Vegas). Conceivably, these networked transportation options provide a window into smarter commutes and have the potential to reduce the margin cost of driving, yet at the same time smarter homes and workplaces are also beginning to gain the capacity to communicate with vehicles. Amazon's artificial intelligence in the Echo can already order an Uber from inside the home. In an AV future, it's clear that homes and vehicles will be able to communicate with one another. If a home is “smart” or “connected,” it will be able to communicate with an AV, anticipating when it might need to be available and ready to travel. (The notion of availability relates to the fact that there may be some weather or other conditions where autonomous vehicles may be inoperable due to its Operational Design Domain (ODD). Messaging to a home or personal technology interface might be important for information dissemination as consumer will need to plan travel anyway.) Since AVs will be electric, the home can also communicate with the car regarding best charging times and the potential to supply supplemental power from the car's battery.
These futuristic visions do not relate to the way that autonomous vehicles are currently being designed. In simple terms, most AVs use the following "redundant" technology to operate: a combination of LIDAR infrared based spatial detection tools, sensors, cameras, and algorithms that make constant decisions about surroundings. Working simultaneously, these tools get someone from point A to point B with the capacity to interact with roads, buildings, and other vehicles if needed. Yet some AVs do not rely all of these "redundant" systems (for example, the Tesla autopilot feature). As a result, these vehicles cannot communicate from one to another or with other features around them (for example, roads or buildings).
The idea that the car need not be connected presents an interesting dichotomy. Although it may not be a necessity for HAV to connect to other vehicles (V2V) or to infrastructure (V2I), much of the potential of the AV may be found in the ability to link infrastructure and the wider transportation network (sometimes referred to as vehicle to everything, or V2X). While a lack of connectivity may not be common to all vehicles (for example, Ford and Google plan for these redundant systems, even those that can read bicycle signals), the opportunity to change the dialogue to consider factors outside the vehicle and the individuals the vehicle serves is important from a public policy standpoint. The government is investing millions into design of smarter energy systems, buildings, transit, roads, signals, and related pricing efforts. It would be a shame if the development of automated vehicle technology happened in isolation of these investments and discussions. For example, assuming that an AV needs to park (a large assumption and something we will discuss in future articles), will it be able to access data on available spaces? Likewise, roadway-sensor-based volume information may be essential for AV platooning on the freeways to mitigate peak traffic events—so shouldn't the cars be able to link to this?
From the Vehicle to the Individual: Saving Lives with AVs
While it is clear that some companies like Google are working on interactions outside of the vehicle, it is not clear that there is a shared technological perspective given the current market competition to build viable AVs. Developing a shared technology perspective and outcome for AVs should be particularly compelling as it relates to one of the key of promises of automated driving: saved lives. We have written that collisions with vehicles are one of the leading causes of death for bicyclists. What if bicycle frames could be equipped with sensors? Could we eliminate bicycle and automobile collisions and move closer to the goals of the zero-accident goals of the Vision Zero initiative? Likewise, a disturbing number of child fatalities relate to automobiles. According to the U.S. Centers for Disease Control and Prevention, on average a pedestrian fatality occurs every two hours, one in five of which is a child under 14. This statistic is frightening, but what if we could implement a new safety standard that put a chip in every backpack or lunchbox? It might raise costs, but if it could save lives it might be worth it.
In a recent keynote at Stanford University, White House Innovation Fellow Eric Daimler extolled this saving of lives as, "the promise of AVs." He referred to it as the "killer app," in that the promise of AVs might not be to increase mobility or productivity, but to save thousands of lives every year. This tug at the heart is important to keep in context during the discussion of AV technology, but the industry and federal policy are not yet discussing the technology as a part of a larger system. We challenge automotive manufacturers and designers of autonomous systems to think beyond the vehicle. Given the potential lives saved, the benefits of doing so may far outweigh the costs.
William (Billy) Riggs, PhD is an Assistant Professor of City & Regional Planning and a leader in the area of transportation planning and technology, having worked as a practicing planner and published widely in the area. He has over 50 publications and has had his work featured nationally by Dr. Richard Florida in The Atlantic. He is also the principal author of Planetizen's Planning Web Technology Benchmarking Project. He can be found on Twitter @williamwriggs.
Michael R. Boswell, PhD is Department Head & Professor of City & Regional Planning at Cal Poly, San Luis Obispo and is an expert on strategies to reduce greenhouse emissions and increase community resilience to climate change. He is lead author of the book Local Climate Action Planning and most recently advised UN-Habitat on climate planning as a part of COP 21. His Twitter handle is @mboswell.