Henry Ford invented the Model T in 1908, making motorised transport accessible to all. The only issue was that, at the time, even if people could afford a car, they couldn’t drive it very fast or far (only 4% of roads were paved). Roads at the time “were often little more than trails that were muddy in the rain and dusty the rest of the time. Any long trip by automobile required not only time, patience, and ingenuity, but tire-patching equipment, tools, spare parts, and emergency food and fuel”. Of course, people did not buy a car to be stuck in the mud, they bought a car to get from A to B faster. Thus, the introduction of the vehicle created a need for better roads and, appropriately, better roads were built. In 1916 and 1921 two separate federal bills were passed to begin the mass pavement of roads throughout the US, enabling cars to operate more smoothly and thus also increasing their appeal and market. Just as dirt roads were inadequate to support the Model T, today’s roads are not equipped to support autonomous vehicles.
Indeed, a century later DARPA’s autonomous vehicle challenge brought a first-of-a-kind unmanned vehicle race across the California desert. This challenge launched the automotive and tech industries into an arms race to build the first fully automated vehicles. Full of ambition and hope, executives promised fully autonomous (L5) vehicles by 2017. However, as we came to understand and appreciate the challenges and complexities involved, the tone has changed. Waymo CEO no longer aspires for full autonomy, rather focusing on developing L4 autonomy: vehicles that can operate autonomously within a specific set of parameters (time, conditions, geographical areas, scenarios, types of road, etc) known as Operational Design Domains (ODD).
Despite this limitation, many companies are investing heavily in developing autonomous capabilities. Even with the more narrowly-defined ODD, a vehicle would need to manage a huge amount of complex and varying scenarios. This requires additional expensive sensors and systems that drastically change the economics of the vehicle. As Harvard Business Review put it: “While self-driving technology is maturing quickly, realising its benefits… requires that the technology be affordable. This prospect is, in our view, far from certain”. The key challenge here is that vehicle developers are assuming the vehicles would need to be able to operate completely autonomously (in the purer sense), without any support.
“Autonomy always will have some constraints” — Waymo CEO, John Krafcik
This is where roads could come in. If vehicles had access to comprehensive information about their surroundings from roads, it would create an opportunity to reduce vehicle unit economics, by reducing the required complexity of the on-board systems. This is currently purely theoretical — roads today do not have the infrastructure or data capabilities to support such use cases. Accordingly, vehicle developers are not developing vehicles to rely on roads for support.
However, this argument forgets the new targets the industry has made for itself. The auto industry is no longer aiming for L5. Vehicle developers are planning for L4. That means we have accepted the vehicles will not be able to operate driverless in all roads, scenarios, conditions, and times. There will be times when a human driver is required. Herein lies the opportunity for roads to offer the right connectivity and data support. This will expand the ODD of L4 autonomous vehicles, creating value for the vehicle operator and the passenger. Practically speaking, it would mean that if you are on your commute to work and suddenly a hail storm hits, or a vehicle suddenly stops, the data from the road would provide the data redundancy needed for the vehicle for it to maintain its autonomy and not require you to take back control of the wheel.
Of course, vehicles today cannot use data from the road, predominantly because this data does not yet exist. If we want vehicles to start using data from our infrastructure, we (the road industry) need to be able to offer reliable, standardised, unique, and value-adding data. Specifically, we should start offering data that supports vehicles in the ‘edge-cases’ of autonomy; scenarios where they would otherwise need to cede control. Once roads begin to offer this data, vehicles have a clear incentive to use it, and therefore automotive OEMs would learn how to receive, interpret and leverage this data. Once a critical mass of vehicles use, and at times rely on, this data road operators could start to offer the data commercially, allowing them to recoup their initial investment. We are unlikely to see all roads offering such services, but if we get our major highways, motorways, and interstates connected, we can easily serve the vast majority of traffic (over 66% of cars and 93% of trucks).
This is a simple win-win. Road operators want to be able to add value to the future of autonomy (and they will likely invest in monitoring capabilities anyway), and auto developers want to maximise autonomy while reducing their BoM. Most importantly, this is a win for the customers of both the road operators and the auto makers: the drivers. Just as drivers in 1908 did not want to patch their tires on every trip, so do tomorrow’s owners of autonomous vehicles want to work, read, or nap rather than constantly monitoring their car’s behaviour. For this to be possible, roads need to take the first step. This infrastructure needs to be available for the virtuous cycle to start. It is a classic case of: ‘if you build it, they will come’.
Learn more about Valerann’s dedication to support the future of CAVs and Valerann’s experience of partnering with road operators to “build it” at valerann.com