Federal Highway Administration National Dialogue on Highway Automation: November 14-15, 2018 Infrastructure Design and Safety Workshop Summary

Collaboration Corner

Format

The Collaboration Corner consisted of a career-fair-style setup with seven stations for collecting different types of information from stakeholders. This setup encouraged a highly interactive session, with participants on their feet and moving from station to station. USDOT staff members were located at each station to encourage participation, clarify the exercise, engage in discussion, and ask follow-up questions. Participants were allowed to move at their own pace but were provided with informal prompts to move to a new station every 15 minutes. Information was collected at each station through two methods:

• Sticky note exercise—Attendees used sticky notes to respond to a specific prompt, which was presented on posters at each station. This was a public form of communication that allowed attendees to view and engage with their colleagues' suggestions.
• Suggestion box—Participants wrote their questions, suggestions, or other input on an index card and placed it into a suggestion box. This was a more private form of communication that allowed attendees to provide information that they may not have been comfortable sharing in a public forum.

Modernizing Infrastructure for AVs

At the Modernizing Infrastructure for AVs station, participants discussed the following three questions:

1. How do physical infrastructure and AVs support and challenge one another?
2. How do digital infrastructure and AVs support and challenge one another?
3. How can existing manuals and standards be updated to address AVs?

Table 1 summarizes their responses.

Table 1. Participant Input: Modernizing Infrastructure

Physical Infrastructure
Connectivity dependencies	Connectivity will depend on power supply and the smooth operation of specific technologies such as 5G.
Data collection using vehicles	AVs can potentially support infrastructure maintenance and the overall system by collecting and communicating specific data (e.g., sending real time data on roadway conditions to IOOs).
Design goals	Various stakeholders need clarification on the expectations of the transportation system they are designing.
Funding concerns	Concerns spanned a range of topics, such as: Who pays for vehicles to communicate to infrastructure? How will the future system be financed? Most States and communities currently struggle to maintain existing assets to the desired standard.
Helpfulness of infrastructure	Physical infrastructure can be helpful to AVs by providing visual landmarks for navigation and real time communications about the state of the system (e.g., work zones, signals, incidents, etc.).
Lack of stakeholder guidance	IOOs and OEMs were identified as stakeholder groups that need further guidance about standards, installation, and cost recovery plans specific to AV infrastructure.
Lack of uniformity	The patchwork of infrastructure conditions poses problems as vehicles move from one jurisdiction to the next.
Digital Infrastructure
Limitations of digital infrastructure	Need to address the accuracy and consistency of digital maps and the coverage and reliability with respect to communication systems.
Mapping methods	Thoughts and questions about how maps might be generated and implemented, included: Should there be one true digital map all companies use, or should there be redundancy? Private sector does all the heavy lifting on digital mapping (private sector easily adapts) Crowdsource mapping system may be possible.
Purpose	Is the purpose of digital infrastructure to be a redundant failsafe?
Manuals and Standards
Acknowledge both the human and the machine	Develop manuals and standards that clarify the similarities and differences between the machine driver and the human driver (e.g., machine vision versus human) and support infrastructure systems that meet the needs of both.
Crashworthiness	Protocol are needed to establish the crashworthiness of AVs and their components.
Intermodal and multimodal	Address the interfaces that enable a system that is both intermodal and multimodal (e.g., connectivity of automated driving systems with vulnerable road users).
Minimum requirements	Provide the minimum requirements for AVs to operate in a variety of different conditions (e.g., minimum visibility requirements for road markings in rain).
More frequent updates	The update cycles for manuals should be shorter.

FHWA AV Research Showcase

At the FHWA AV Research Showcase, a team of directors from the FHWA Office of Operations Research and Development (R&D) presented videos and fact sheets about their current research initiatives. The showcase focused on their Cooperative Automation research, which investigates platooning, speed harmonization, lane changing, and other capabilities of vehicles that can communicate with other vehicles and infrastructure. Special emphasis was placed on CARMA: the Cooperative Automation Research Mobility Applications³ technology. CARMA is an open-source software platform that has enabled the testing of use cases for cooperative automated driving systems (CADS). The platform is an example of an innovative and collaborative project built using an agile software development approach within a diverse community of stakeholders. CARMA is available for download on the software development platform GitHub. FHWA anticipates that this open-access tool will continue to support industry collaboration and CADS testing to improve the safety and efficiency of transportation systems.

Participants provided feedback on FHWA's existing research portfolio as well as suggestions for future research. A summary of their input is in Table 2.

Table 2. Participant Input: Research Suggestions

Participant Input on Research Suggestions

How automation will change land use patterns and community development. How to best educate various stakeholders about AVs (e.g., drivers, all other road users, IOOs, regulators, elected officials, etc.). How to establish tools that enable peer exchange (e.g., universal code for digital platform, clearly defined terminology). How automation and platooning potentially impact human factors (e.g., level of exhaustion of a human driver whose truck is part of platoon). How to test AV detection of and interaction with motorcycles. How AVs might interact with rail.

Roadway Safety Challenges for AVs

At the Roadway Safety Challenges for AVs station, participants discussed the following two questions:

1. What are safety challenges with AVs?
2. What tools, strategies, analysis should be explored to address safety challenges?

Table 3 summarizes their responses.

Table 3. Participant Input: Roadway Safety Challenges

Safety Challenges
Communicating roadway changes	How are updates made to an area that has already been mapped for AVs (e.g., day-to-day changes such as work zones or new turn restrictions)?
Difference between human and machine driver	Relative rigidity of the machine The behavior of humans in traditional vehicles has influenced safety countermeasures up until this point; how will machine driving behavior differ, and what will this mean for safety countermeasures?
Health and safety related to AVs	What are the health and safety risks that accompany: Regular, everyday exposure of passengers to AVs? Exposure of fire/emergency medical service (EMS) personnel to AVs that have been involved in an incident?
Incidents and special situations	Interaction with first response Communication networks during power loss Weather and other special events Work zones Dark intersections
Mixed fleet composition	How to ensure that AVs safely interact with a diverse and changing composition of other vehicles; developers might work with traffic compositions that vary by the percentage of AVs.
Safety Benefits	What's good enough? Questions were posed regarding what improvement in safety is good enough to allow AVs on the road. Is it simply a reduction in total vehicle crashes?
Tools & Strategies
Backup power	Provide backup power systems to mitigate the risk of power loss.
Incident data	Real-time and robust incident data Work zone data Incident data captured by an AV's "black box" (obtained post-incident)
Researching safety countermeasures	These countermeasures can be further researched through new safety data and potentially through a cost/benefit analysis.
Standardization	As a strategy to improve several areas, including: Performance metrics Submission of infrastructure changes to cloud Physical infrastructure requirements Road readiness of AV system
V2I	Connectivity as a tool to improve system safety

Workforce and Organizational Development

At the Workforce and Organizational Development station, participants discussed the following questions:

1. What are your workforce development needs?
2. What kind of technical assistance is needed to address AVs?
3. What guidance would be useful to address AVs?

Table 4 summarizes their responses.

Table 4. Participant Input: Workforce and Organizational Development

Workforce Development
Preparing workforce for change	Train the employees enabling the current transportation system about the new capabilities required to enable the future system.
Cultivate expertise	Human factors in context of AI (artificial intelligence) Data management and analysis Virtual simulation Sensors Communication of science and technology Internet of things
Technical Assistance
Computer vision/ programming transparency	Transportation officials want to understand how computers see and make decisions so they can design and manage infrastructure better.
Training for emergency responders	What to do and not to do responding to AVs Health and safety risks (e.g., potential exposures) Necessary personal protective equipment
Guidance
Customized for various organizational levels	Requests that guidance be customized for specific audiences (e.g., National strategy, metropolitan planning organizations (MPOs)/local planning, State infrastructure, EMS training).
Knowledge exchange	Need help seeking perspectives from a variety of groups and cultivating relationships between various groups (e.g., State/MPOs/local officials and AV developers).
Planning and design	Help design roadways to accommodate AV requirements and behavior.

Terminology

Participants shared the most common terminology that they hear when discussing AVs and indicated which terms are helpful and which are confusing. They placed these terms along two axes to show how these terms are used. The vertical axis represented the frequency with which these terms are used, and the horizontal axis represented the level of confusion surrounding the use of these terms. Table 5 illustrates the terms placed into each quadrant.

Table 5. Participant Input: Terminology

More Confusing
5G (fifth-generation network for cellular mobile communications)	Connected vehicle
ADAS (advanced driver assistance systems)	Disengagements
AI (artificial intelligence)	Fallback conditions
Alphabet soup of agencies	Machine vision
Automated	Robots
Autonomous	RSM (roadside safety message)
Autonomous is wrong	SPaT (signal phase and timing)
BSM (basic safety message)	SAE (Society of Automotive Engineers) levels
Chaperone	Operational design domain
Connected
Clear
ADS (automated driving systems)	SAE levels
C-V2X (cellular vehicle to everything)	Safety driver
DSRC (Digital short-range communications)	SCMS (smart city management system)
I2V (infrastructure to vehicle)	Simulation
J2735 (SAE standard for dedicated short-range communications)	SPaT and Map Data (MAP)
Mapping	TOSCo (traffic optimization for signalized corridors)
MUTCD (Manual on Uniform Traffic Control Devices)	V2I (vehicle to infrastructure)
OEM (original equipment manufacturer)	V2V (vehicle to vehicle)
OBU (onboard unit)	V2X (vehicle to everything)
RSU (roadside unit)

Parking Lot

Any remaining questions and comments that did not cleanly fit into the other topic areas were included in this topic area. Topics included:

• How legal and insurance industries should collaborate.
• Collaboration between agencies of the USDOT to provide optimal solutions.
• How the airline industry can guide planning and deployment for AVs.
• Need for an FHWA-led vision for planning near-term implementation.
• Need for considering train detection at passive crossings.
• Consideration of whether AVs may need to be marked or identified in a way that is obvious to other human drivers.
• Assessing how to keep the driver engaged at appropriate levels of automation.