NCHRP 20-24(128)

We as a country have never really had a national vision for AV. Perhaps if we had a national- scale deployment plan and associated business case investment it would usher in more coordinated policies? Several people have suggested a comparison to the "moon-shot" mission to land a man on the moon.

Challenge #1: Complexity. One challenge in resolving this deficiency is complexity. There are many different stakeholders involved in automating transportation, more than there were for the moon-shot. Hundreds (thousands?) of state, regional, and municipal departments of transportation, transit operating agencies, metropolitan planning agencies (MPO’s), toll agencies, and more make up the public sector. There are likewise hundreds (or more?) of private companies engaged currently in some aspect of developing AV.  Many if not not most have different approaches, business models, and goals. It’s incredibly complex.

Challenge #2: Scope. There are so many different aspects of transportation that are evolving at the same time, and not always on the same path. Connectivity, automation, electrification, shared mobility - it’s an extremely broad scope of changes all going on, and impacted in different ways depending on the specific use-case you’re looking to explore.  When mapped across the complexity of stakeholders - and mixed with ever-changing politics and socio-economic factors - the scope is very difficult to narrow in focus.

Federal leadership in setting a national vision, and in providing clear and consistent regulation that is appropriate to the federal level would be beneficial. States will continue to set their own goals, potentially including economic growth and attracting private sector testing/business operations. But having a “North Star” or “moon-shot” from the USDOT would help guide states toward “doing the same thing” In a manner that would enable more private sector engagement with the broad understanding that their actions benefit many, and not individual states.

Regardless of how a vision is created, cooperation rather than competition in key areas is recognized as preferred, or simply more rational. For example, we heard from several organizations about how "...it's not healthy to compete on safety..."

Cavnue noted that "we need to meet people where they live now, not in 20 years."  

Example. The technology installed during pilot projects should also be designed to provide enhanced benefits for human drivers - maybe through extra data collection, information dissemination, or system design.

[updated] Aurora mentioned that they are frequently asked by IOOs what do you need from us?  Their answer is consistent: "what is good for drivers today will be good for us tomorrow."

States should invest in good repair while technology is being developed (e.g., if signals are inoperable, V2X doesn’t matter; “good striping, good signage, good hardware/back to basics”). The ongoing challenge of directing funds away from physical infrastructure repair (e.g., potholes) or construction (new lanes) toward technology remains, and investment decisions will continue to be based on sound cost/benefit analysis - but done so through a longer-term lens in some instances.

Organizational readiness is a significant challenge within CAT because, unlike some past technological revolutions or endeavors, a diverse range of stakeholders and diffused or uneven leadership exists. The same complexity that has challenged the development of a national vision also challenges the notion of readiness.

What Is Your Focus? One solution to improving mobility might be a low-speed AV shuttle. We met with Local Motors who wants to specialize in building the vehicles, but they have other private partners that focus on the artificial intelligence and other needs. We have since met with other companies that don’t want to focus on the vehicles at all. In fact, when we met with Waymo they made it clear - they want to build the world’s most experienced driver through software, they don’t want to build a driverless vehicle. And more recently with Cruise, they are focused on both the vehicle and the AI/Driver. Entirely different points of focus, even though we tend to “lump them all together” in the AV category of companies.

Not only are companies taking a different focus, but the vehicle "type" could influence the complexity.  Low-speed shuttle vs ride-hailing passenger vehicle - there are dramatically different needs, issues, operational domains, and risks. The needs for an organization to be “ready” are dramatically different.

Complexity Shifts. Just as politics can shift an agency's focus, the marketplace can shift a company’s focus over time, either by necessity or deliberate strategy. The AV truck company TuSimple that’s testing and operating right now sees itself evolving into a software company eventually offering subscription-based automated driving software and mapping packages in the future.  But today they are operating trucks and maintaining vehicles.

One company noted that “Tech can happen to you or happen with you.” The message was clear: engage in dialogue now, don’t wait until it’s already starting to deploy.

During a conversation with another company, it was clear that early dialogue will help uncover inconsistent expectations - such as a private company assuming it will get a dedicated lane for CAV while the public agency doesn't see that to be financially, politically, or operationally feasible anytime soon.

State legislation or other regulatory or policy guidance on the testing and deployment of AVs may take on growing significance as federal and state roles begin to blur.  But for now, such policy must still work within a federal regulatory framework that is based on human drivers.

Many companies we heard from recognize that you can go farther by partnering compared to going fast by going alone - but not all follow this philosophy. How will that impact long-term evolution of AV?

However, expectations of equal level of benefits are more difficult to have as a specific target. State and local governments have several decades’ worth of experience with transportation planning, and it is worthwhile to hear from them what cities could look like if things such as curb management, parking and EV infrastructure, shared street designs, and sidewalk usage were re-imagined.

• Both public sector and private sector entities in the Phoenix region emphatically point to partnership and cooperation as critical to advancing CAT applications in the best interests of all stakeholders and end-users.
• This point was corroborated by Robotic Research’s movement to serving the commercial sector, which was the result of partnerships with the federal government, academia, and other companies such as Local Motors. The partnerships enabled Robotic Research to understand the commercial sector and pursue the development of a low-speed AV shuttle.
• The point was made that some companies (e.g. Robotic Research) have decades of experience with the public sector and have worked in risky environments, whereas other companies (e.g. Cruise) don’t have much experience working with public agencies and are asking for very little from the infrastructure, except for permission to operate its vehicles on their roadways.
• The City of Chandler has built a successful relationship with Waymo that balances its built- environment considerations with Waymo’s business goals for a scalable and profitable AV rideshare service.
• Ambivalence toward the public sector still exists, with AV companies that view themselves as technology companies. In general, their outlook is to have a business model predicated on the free use of the public right-of-way and that state DOT’s are to conform the public right-of-way via major investments to optimize these companies’ services.
• Sometimes one partner can be the catalyst, even if they aren’t the lead funding authority. The Road Commission for Oakland County (MI) (RCOC) used its operations and maintenance experience to leverage CV partnerships with FHWA, Michigan DOT, Big Three auto companies, top-tier auto suppliers, CAV companies, academic institutions, and the Oakland County general government. This has allowed them to partner on more than 20 CV projects.
• RCOC also worked to define new contributions from its partnerships - such as encouraging industry to develop more field-hardened equipment, more thoughtful access to field equipment for maintenance, remote software updates, status monitoring software, remote hard reset of equipment, more up-front planning for field installations, and building as much as possible in a controlled environment first.

Not all partnerships are simply public-private ventures.

Partnerships between State and Local DOTs are critical. RCOC has had a successful partnership with Michigan DOT that has been mutually beneficial (e.g., each benefits from the other’s different areas of expertise) and has learned that state DOT’s can be quite helpful in the purchasing of equipment (e.g. thermal cameras) and in the standardization of CV applications (e.g. security credential management system) and interoperability.

Partnerships with academia are significant to help develop the talent and workforce required of CAT technology—both developers and application users.

And partnerships with Public Safety are critical to ensuring successful pilot demonstrations, such as Waymo’s outreach efforts with the City of Chandler police/fire ahead of beginning operations. Argo AI and Beep both recently confirmed this by noting their strong relationships with police, fire, and EMS in all cities in which it is testing - providing briefings on how to interact with its vehicles, perform extractions, and the general capabilities of the vehicles they are running.

A newly created consortium of public-private-academia in Arizona suggested that their formation was based upon “cooperation is the new competitive advantage.” The Institute of Automated Mobility (IAM) is a public-private-academic partnership model that facilitates the sharing of research resources and results, data sharing methods that protect proprietary interests, testing facilities and tools, and traffic incident management training for autonomous mobility technology.

The CEO of Beep in central Florida underscored the importance of cooperation across all partnerships.  He noted that "we need everyone to succeed....one failure sets everyone back." 

[updated] In a conversation with Cavnue, they too noted that several years ago the landscape was highly competitive and not at all collaborative.  While the competition level still exists, there is a lot more collaboration, particularly in testing and development.

Early AV testing and operations focused predominantly on transit substitutes or transit augmentation (e.g. AV rideshare, low-speed AV shuttle applications) and long-distance freight movement (e.g., TuSimple automated truck operations) use-cases. The majority of these activities have been led by the private sector. The public sector is both encouraging this private- sector development by enabling permissive regulatory environments and piloting several AV use-cases themselves, especially the feasibility and public reaction to applications of low-speed AV shuttles.

Example. Robotic Research provided a good example of this diversity of use cases - they develop AV technology for the military, they are working with low-speed AV shuttles, and they are working on automating full-size buses.

[updated] Example. Cavnue also pointed out that some freight companies are jumping right to Level 4, while many passenger cars are starting with ADAS and incrementally building up toward ADS.  The differences are varied and stakeholders diverse.

This is not to say the private sector doesn't care about connectivity, but a significant number of companies we met with suggested that the ongoing regulatory uncertainty and challenges toward ubiquitous deployment of CV technology has led them to "accept it if available, but not rely on it solely."

Example. In the Phoenix region CV use cases focus mostly on arterial safety and mobility applications for public fleets and pedestrians, including emergency vehicle, transit, freight, and pedestrian signal priority applications. These pilot deployments have been led exclusively by the public sector.

Example. Utah is aggressively pursuing traffic signal priority CV applications for bus transit and snowplows, along with other safety-related CV applications such as curve speed warnings and road weather warnings. Utah DOT has partnered with Panasonic to deploy the necessary physical and digital infrastructure for these future deployments.

Example. The City of Marysville, Ohio has chosen to focus on CV technology, specifically providing the infrastructure necessary to enable CV’s to operate with V2I technology.

Much of the technology is still immature, such as Utah’s experience with low speed AV shuttles experiencing problems with weather tolerance, sensor “false positives,” and mechanical issues; or RCOC’s experience with DSRC implementations where basic device design didn’t incorporate enough weather-hardened considerations and had to be adjusted after the initial deployments.

As mentioned previously, not all private companies felt roadside infrastructure was a necessary component for pilot testing, but some suggested that “if there is advanced technology on the infrastructure, we’ll use it – but if it’s not everywhere we can’t plan on it.”

Example: infrastructure providing connectivity like signal phase and timing broadcasts at arterial intersections would have to be present across an entire network or system and provide guaranteed service for some companies to be compelled to capitalize on its availability.

Example: Argo AI suggested that V2X technology could be "additive" to their current efforts, but remain focused on designing its system to be able to operate safely on its own capabilities.

Some public agency deployers have openly suggested we should select one technology and proceed, even if it "scares away" some parties that have already made their choice.

Example. The City of Marysville, due to their close proximity to the largest Honda manufacturing facility in North America, decided to go with DSRC for now, even if it meant Ford might not be interested.

Related to this was the question of how jurisdictions should proceed with their investments (i.e., either commit to one technology or a combination). The recommendation was that swapping of one technology for the other often involves "unplugging one and plugging into the other."

Example. Georgia invested in both technologies with the understanding that one will win out and found that the costs to deploy both were not significantly more than investing in just one.

When pilot testing or developing new solutions, it’s important to remember that perfect can be the enemy of good. We spoke to one private company who noted that we don’t have to solve every problem for every scenario - perhaps the use case is narrow, but testing & deployment is thorough within that narrow environment.

Example. TuSimple suggested they had trouble solving AI issues in ice/snow conditions, so their initial testing and deployment will be focused along portions of I-10 that don’t encounter ice/snow conditions. There are no written rules than an application must be applicable to every possible domain.

Policy development with DOTs often focus on safety - and most companies agree that safety needs to be number one. BUT, there is an economic imperative associated with CAT deployment policy too, and that can sometimes get overlooked.

Having a statewide AV workgroup has proven beneficial in several locations.

Methods to enable AV testing and operations on public roads include governor’s executive order (e.g. Arizona) and state legislation (e.g. Utah).

The executive order approach tends to focus implementation on the priorities of the executive office (e.g. positioning the state as a “leader” in attracting private-sector AV deployment and investment).

By contrast, the legislative approach can represent more of a compromise among various stakeholder interests across considerations such as AV registration requirements or protocols for accidents involving an AV.

• Companies that are currently testing recognize that relationships with government are important – but if they feel they are being inhibited or held back they’re going to find other communities to test in.
• Factors influencing potential testing locations (for Cruise and in regard to testing outside of California and Arizona) consist of a friendly regulatory environment along with an openness to having conversations.

If testing is permitted through legislation of some kind, your agency should actively participate in the legislative process and be knowledgeable of the implications of issues or processes that affect safety—for example, definition of terms, enforcement mechanisms, ability to know or track who is testing what, and crash responsibility/liability. These issues may require compromise among the different interests at the table, both public and private.

The lack of regulation at the national level is presenting some challenges. For instance, low-speed AV shuttle pilots that require a NHTSA regulatory process has been described as very cumbersome (written approval for routes, 30-page submission, 3-month turnaround times) and reportedly favor foreign manufacturers.

Cruise has identified several principles to guide the AV regulatory framework:

(1) have a clear path to deployment (laws and regulations should be explicit in authorizing AV deployments and remove existing roadblocks that assume a human driver).

(2) real-world environment testing (ensure testing is permitted in areas that replicate the locations in which deployments will occur).

(3) statewide applicability (have a single unified, statewide framework instead of a patchwork of regulations).

(4) service neutrality (ensure the framework permits both transport of people and goods in AV’s).

(5) technology neutrality (encourage governments to permit new technology to mature and avoid overly prescriptive approaches that may constrain progress, innovation, and effect).

(6) insurance (insurance requirements should reflect a state’s requirements for other vehicles in the same category).

Several panelists agreed that public agencies don’t always do a good job identifying what information we need to know - complicated by different ways to talk to the tech industry (through a standard vs through an opportunity) - too often we just say "give us everything and we’ll figure it out later," which can have negative implications for private industry.

Public agencies should focus more on the desired outcomes as opposed to the inputs, how to achieve those outcomes, and what can be achieved together in terms of safety and efficiency by identifying the necessary information to achieve those outcomes. To do so would require being able to partner with the engineers and data analysts of private companies to collaboratively solve problems rather than by just asking for the data.

Example. The Road Commission for Oakland County (RCOC) recognized that the use of V2X data will likely require a major paradigm shift that changes the way transportation agencies do business - "Can there be too much data? How and why we collect data would be different."

Example. The Univ of Florida has been running a lot of simulation and modeling to predict future impacts of AVs, but lacks sufficient data on how AV's will "drive" the vehicles.  There are many years of data that can replicate how humans drive and simulations are based on that - but do we truly know how a computer will do it differently?"

Private-sector stakeholders are highly reluctant to share “all” their data, particularly in a competitive business environment. They need to see the clear value for sharing their (often proprietary) data.

Example. Regarding mobility data specifications, Cruise has not taken a position, adding that there are areas in which Cruise and the public sector can function as partners, but there are also privacy concerns that will need to be addressed so that private data will not be inadvertently disclosed.

Example.  Beep suggested that telemetry data is already openly shared to help improve learning - but private data collected on the rider app contains several limitations.  Video data is also tricky - while external video for the most part is okay, they want to share as much as possible but not make customers uncomfortable.

A bigger conversation regarding the standardization of what DOT’s need (in terms of data) should take place (perhaps a cause that AASHTO can take up) and establish a national platform requiring that every state and local DOT get "XYZ data" so that there can be a single dataset.

For example, data on the use of physical infrastructure (e.g. travel data or intersection data), however in other areas, private companies may collect their own (e.g. LiDAR) because they don’t trust or want to wait for what is available from the public agency.

It was suggested that if AV companies can provide state DOT’s general information, such as on work zones, it will allow them to better manage the roadway system and lead to greater efficiency. The possibility of putting value on the data coming into state DOT’s and having the AV company pay for it was also recommended.

Another suggestion was that state DOT’s could function as data clearinghouses to provide predictive analytics to other providers and that the better information state DOT’s have, the more reliable the system will be and that there will be less congestion.

A symbiotic relationship looks likely in the future in that DOT’s will receive the data for use in their operations centers and feed it back into the system, which ultimately results in a better service with the private companies helping with the data management.

The public-private-academic forum IAM (from AZ) demonstrated a great example on how to focus on anonymizing data through a third party and sharing lessons learned more openly. In a broader sense, the forum also represents a promising model for bringing key players to the table in a collaborative environment that recognizes the complexity of the CAT endeavor, and allows each organization to contribute and benefit in line with their goals.

Cybersecurity risks could potentially limit private sector willingness to rely on government data. It is important that agencies have internal conversations with IT departments in advance, so that future dialogue with private partners isn't impacted by concerns or risks that could be mitigated through proper planning and communication.

Local, county, and regional governments and MPO’s see a partnership with state DOT’s regarding the secure access to data, data storage and governance in which state DOT’s help to develop policies and regulations to create equity between jurisdictions. There also needs to be standards regarding data sharing so that state and local DOT’s “can speak to each other” (i.e. achieve interoperability).

Google suggested to the panel that efforts toward a national vision should also include an emphasis on national data standards.  They suggested three ways to view data standardization: (1) how to ingest data into the cloud, (2) how to share data internally to your organization, and (3) how to share data externally with other partners, developers, and consumers.

The public has to be comfortable with the technology and see the benefits.  At the same time, there also needs to be trust in public institutions that policies are in place to, among other things, preserve safety and promote an equitable distribution of benefits.

The public has to be comfortable with the technology and see the benefits. At the same time, there also needs to be trust in public institutions that policies are in place to, among other things, preserve safety and promote an equitable distribution of benefits.

Estimate Benefits. Agencies and/or private companies looking to pilot CAT technology should make the case through economic studies that show cost reduction, increased transit usage, the better use of public funds, etc. We often focus only on the high- level safety case, when additional justification might be of benefit to all parties.

Rely on Partnerships. Pilot programs can also open the door to new investment opportunities. RCOC saw small cell/5G coming and recognized the need for a CV business model, resulting in a RCOC public-private partnership. Its goal was to raise money to support the deployment of CV including on-going operations and maintenance.  This required the hiring of a third-party to broker agreements with cellular companies, and revenue was shared with the broker. As of now, there are 267 sites. In the future, RCOC plans to have cellular companies install DSRC as part of small cell and use cellular companies’ communication backbone. Argo AI also suggested that consumer engagement could be more robust, and they welcome help from DOTs in that regard.

We don't have to wait until all the results are in to share possible benefits.  In fact, some benefits might surprise even the scientists and are worthy of sharing.

Example: according to an industry expert, fuel savings from truck platooning has long been an accepted benefit; but recent testing and simulation has confirmed that automated driving systems for individual trucks (solo AVs) can also result in fuel savings. The ADS can be more consistent, careful, and economic as a driver - which results in some fuel savings as well.

Florida DOT has instituted 4 hours of training for all TSMO program engineers in each District.  There are 13 modules in the comprehensive course meant to not only build awareness, but identify gaps and opportunities for future growth.

"Health is now a transportation factor" (e.g., having to quarantine for inter-state travel and commerce and the effect on how things are delivered).

The pandemic slowed some testing temporarily, but also allowed many companies to look inward and continue development

Example. Robotic Research mentioned that COVID-19 forced them to develop new methods and they now have two patents pending for ways to sanitize AV’s that scan for living organisms, self-clean, and monitor conditions afterward.

Example. Cruise and others have relied heavily on testing via simulation due to the pandemic. AV development hasn’t slowed down, but the testing angle has adjusted.  In response to the COVID-19 pandemic, Cruise has also been collaborating with the SF-Marin Food Bank and SF New Deal to deliver meals and groceries to those in need.