Automation lives: transportation agencies can spark AV advancement

What if infrastructure owners and operators and the automated vehicle industry could invest together to identify a set of minimal functional requirements for automation, better accelerating the safety frameworks for deployments, and thereby support automation developers and operators during a time when their funding is tight?

Looking back at 2022, we saw the wrap of several organizations focused on the deployment of automated vehicles (AVs) with the closing of Argo.ai being the most recent. The combined forces of inflation and potential recession have driven up the cost of investment funding, and the United States is facing intense global pressure related to the development of automation. Vehicle automation also has the additional burden of being a safety-forward technology and safety solutions have traditionally shown a lower return on investment. This economic environment is leaving a strain on start-ups entering the market as well as organizations currently operating in the AV market.

Where does this leave automation? How can AVs prove their safety reputation while delivering on their promising investment to market? More importantly how do we continue to advance a technology and strategy that can help us tackle the significant loss of life caused by our existing modes of transportation that connectivity and automation could help solve.

Continued investments from global giants like Google’s Waymo which recently launched full-fledged robotaxi services in Phoenix, GM and Cruise which launched commercial services this last fall, or Baidu and Pony.ai which have won the right to deploy automated taxis in Beijing point towards a continued growth in a tightening global market.

Infrastructure owners and operators (IOOs) could equally have a significant investment in how to accelerate automation. Over the past few years, IOOs have worked to find their role in the deployment of automated vehicles. For the past decade, AV manufacturers have consistently messaged to IOOs that their vehicles are able to function in environments built for human drivers; however, minor adjustments to the infrastructure, particularly in the form of extra-vehicle situational awareness provided via communications, would allow for AVs to function more optimally. AV operators have explained that infrastructure consistency is important providing an environment that minimizes conflict with other road users. Until now, IOOs have not had a significant role in the deployment of AVs.

What if IOO’s and the AV industry could invest together to identify a set of minimal functional requirements for automation, better accelerating the safety frameworks for deployments, and thereby support automation developers and operators during a time when their funding is tight?

There are a handful of AV operational needs that are common across most platforms and approaches. If IOOs could develop some of these common factors, AV developers may be able to use their limited funding for other automation development. Equally IOOs would be playing an investment role in accelerating the deployment of safety benefits brought by the automation technology.

Some of the key areas of cooperation and support from the IOOs may include:

• Localization support
• Object detection and classification
• Common elements of path planning – such as sparse high-precision GPS waypoints, and high definition (HD) mapping.

If the AV industry can harmonize on these attributes across operational design domains (ODDs), AV developers may use investments to support more specific automation capabilities required for that developer’s specific business needs.

Localization and mapping

Many IOOs are considering creating high-definition maps of their geographies and several are considering integrating these with digital twins that also allow the IOO’s to convene digital policy, rules of the road and insights. As part of their efforts to improve safety in Utah, Utah Department of Transportation has already created HD maps of the entire state. Not all AV operators or designers use mapping the same way, and most AV OEMs create their own maps.  If IOOs were to undertake an effort to understand the minimal set of data attributes needed for these maps, there may be opportunity to provide some harmonized basic mapping protocols that could be used by AV operators.  If IOOs can increase safety with an investment in mapping, that may also allow AV operators to invest in other areas of operation, thereby proliferating safety and mobility improvements and improving automation technology.

IOOs should also work with the AV industry to determine what information can be shared from the industry back to IOOs if, for example, minimal map data is generated and shared with the AV industry, perhaps the industry could reciprocate with high-precision GPS corrections to the position of map elements. A thorough understanding of potential shared data needed to support automation could also be a part of an IOO effort to create digital twins of infrastructure.  If an IOO can work with AV operators to understand data needs, digital twin design can be harmonized to accept and use data from vehicles.

Harmonized asset data

Roadway assets, specifically lane markings, signage, and traffic control devices, are not the same throughout the world.  If IOOs can work together on developing and approving a harmonized dictionary for roadway assets and create a data exchange for this information, this could enable safety capabilities of AVs. This concept is already being pursued in the Department of Transportation Work Zone Data Exchange (WZDx), and for other infrastructure-based information such as signal phase and timing (SPaT) through the USDOT Joint Program Office (JPO) Operational Data Environment (ODE). Expanding on the WZDx idea, AV truck operators have also requested a Weigh Station Data Exchange (WSDx), which is another area where IOOs could add a spark.

Likewise, precise localization is a challenge for both automated and connected vehicles, specifically in “urban canyon” areas where tall buildings inhibit direct line of sight to GPS satellites and the GPS signals are reflected.  Tunnels also provide a specific challenge for automated and connected vehicles for blocking GPS entirely, and due to the extreme lighting contrast for machine vision systems when entering or exiting a tunnel. Even in the complete absence of GPS information, AVs have the benefit of numerous onboard sensors, which are used to provide precise localization data to AV systems, such as the path planner. However, the effectiveness of this is tightly coupled to the algorithms used within the AV software stack. The USDOT-sponsored connected vehicle deployments have shown significant challenges in urban areas with tall buildings, specifically in New York City.  In that pilot, the noisy GPS data was addressed using a novel method of measuring time-of-flight from Roadside Unit (RSU) to offset GPS signal error and verified using a vehicle mounted laser pointer. This would be possible by precisely measuring the GPS position of the RSU, which can be transmitted to On-Board Unit (OBUs) or stored in an onboard map of the AV or CV system. This is another example for how automated and connected vehicle systems can inherently improve each other, and how IOOs may be able to better support automation.

Like the RSU solution NYC used, GPS corrections can also be provided using a technology called real time kinematics (RTK), which uses a precisely positioned base station and broadcasts a correction that devices can use to overcome the error in the GPS signal. IOOs could provide something similar as a service to augment GPS precision equipment, which may include any kind of roadside equipment that is able to be precisely located and transmit a simple message with its location and a timestamp; however, regardless of the information an IOO is able to provide, the automated or connected vehicle devices will still need a minimum level of capability to process the data available effectively.

Path planning

Path planning is one of the fundamental components of an AV. In essence, this function is responsible for evaluating all available paths the vehicle could take in both the short-term and long-term planning horizons, and then selecting the “best” path.  This occurs many times each second for short-term planning, which allows a vehicle to correct for small deviations in the vehicle’s position versus its previously planned position, and to react to immediate or predicted hazards that have been detected by the AV’s perception pipeline. Long-term path planning is akin to route planning and may never be revaluated once a route is set; however, a flexible path planning architecture will have the ability to replan a route based on unforeseen circumstances.  To the extent an IOO can support fundamental path-planning which is something AV developers could potentially share.

One of the limitations of today’s AV systems is their inability to drive on roadways that have not been previously mapped by the AV developer using proprietary methods and data structures.  This limits scalability and operational flexibility, but according to the previous mention on minimal map data requirements, IOOs could provide a sparse GPS waypoint data layer, accessible through a permissioned API for example, that would provide the AV developer with an idea of the contours of the roadways that a new route could be created within their system.  The first time a vehicle travels on a new roadway using only the sparse GPS waypoints, it could proceed more cautiously relying more on its onboard sensors to navigate the environment, but as its traversing this new route, it can be recording all the data needed for the AV developer to create its own version of a map.  The AV industry could then contribute back to the IOO information such as corrections to the sparse waypoints, further improving the accuracy of these, and expediting the use of the roadway for others in the future.

Together IOOs and the AV industry have an opportunity to use ingenuity and transferable solutions-thinking to integrate data, systems and mapping that can improve the safety ROI needed to ensure the livelihood of the AV market. Investors, developers, public and private organizations all should be working together to enable the future of automated transportation.