Autonomous Vehicles in Slow Lane as Robots Accelerate

While robotics advances at pace in many industrial settings these days, the robotic or autonomous vehicle remains a nebulous prospect for many.

Emily Shao, partner at McKinsey & Co, says the reality is that the widespread adoption of Level 5 autonomous vehicles is decades away from becoming a common reality.  She also clarifies the prevalence of robots as being defined by those that already inhabit and automate warehouses.

Before analyzing the industry trends and the current forecasts more closely for Level 5, fully autonomous vehicles, she asks us to take a step back. That’s because Level 5 autonomy is very challenging – even for humans to achieve anywhere. For this reason, the industry is focused on Level 4 of autonomy.  So, the focus is on building capabilities on specific operational design domains (ODDs) such as weather, road conditions, specific types of regions, or the time of day.

Shao explains: “An example of an ODD can range from operating on one street from 9am to 5pm when it’s sunny or be as broad as operating any time in a particular city. Humans often don’t wish to drive in particular conditions, and many of these conditions are also difficult for autonomous systems, so it’s also a challenging technical problem to solve to enable the vehicles to operate autonomously anytime, anywhere.”

As well as the need to consider a diverse range of ODDs, she believes commercialization of autonomous vehicles operating in dense, urban environments is a long way off because “the AI, the software and the valuation requires an unlocking of the different ODDs”. This requires the development and evaluation of the hardware, the software and of the safety case to ensure that any specific ODD is safe for the vehicle to drive in.

History of robots

Drew Winter, principal analyst, Informa Tech Automotive Group, and director of content at WardsAuto, then talks about the history behind robots themselves: “The automotive industry has been trying to replace humans with robots for decades. In the 1980s and 1990s automakers such as General Motors spent billions on robots and machine vision systems in an attempt to improve productivity (robots don’t take bathroom breaks or go on vacation) and replace expensive union labor.

“These grand efforts were largely failures for obvious and not-so-obvious reasons. The obvious reasons were robots represented a huge initial investment but also massive maintenance and tech support and programming expenses. Malfunctions such as robots painting each other were instant headlines that clobbered stock prices. Instead of making automakers look high-tech, they looked foolish.”

“Robots were also dangerous in the workplace to humans and had to be fenced off. Even then, horrible deaths and injuries have occurred and resulted in lawsuits and bad headlines. A less obvious problem with robots is that when tough times come, you can’t lay them off.”

Robots: finding their place

He explains that robots eventually found their niche in the factory by being tasked to complete difficult, dangerous and complex tasks. For example, “Robots are do body and chassis welding, which is a difficult and dangerous job.” However, they don’t tend to assembly work where humans can do a better job. He therefore foresees similar scenarios occurring with autonomous vehicles.

He asks: “How horrible would Uber’s balance sheet look if it had purchased tens of thousands of expensive self-driving taxis right before the pandemic?” He then highlights the key problem with autonomous vehicles – that being they will have to coexist with unpredictable, drunk, and sometimes hostile human drivers for at least the next 30 years. The challenge will be about how to manage autonomous vehicles alongside human-driven ones to avoid any collisions within a shared space.

Following the same path

Nevertheless, he sees autonomous vehicles following the same path as industrial robots. To begin with there will be much hype about them – just like there was initially with robots. This will then settle down into “less complex and highly specific commercial roles where they will shine,” he says. Low-speed delivery vans, parking lot shuttles are already here today.

As for personal cars that can “drive on narrow mountain roads in a snowstorm, not going see them for a very long time,” he predicts. As for whether robots are more ubiquitous the autonomous vehicles, like Shao he suggests that it depends on how robots are defined. However, with the many uses robots already have, such as for surgery, he would agree that robots are more widespread.

Shao comments: “Robots are actually autonomous vehicles operating in constrained ODDs, such as a warehouse with clear lane markings. They don’t have the additional noise that occurs in the street, and so the challenges (such as roundabouts, jaywalkers, construction zones, and other obstacles) are fewer.”

“Companies are prioritizing use cases such as robo-taxis, which was the first because of the very attractive market size and we still see players in this, and then there are now offroad use cases in agriculture and in long haul trucking given driver shortages or to optimize fuel to avoid shipment delays. These areas have the opportunity to commercialize autonomous vehicles faster because the use cases can have more constrained conditions to solve for.”

Use cases for trucking

Shao also thinks that there are some pure on-highway use cases for trucking. It may not occur on a large scale for the meantime. However, the opportunity is still there. A key issue is that the autonomous trucks don’t have their own dedicated lanes – despite the fact that having them would accelerate their adoption.

She explains: “For example, if you had a dedicated lanes for just autonomous vehicles, where human drivers couldn’t create obstacles, you would be able to better control the environment. However, you can’t expect every city and state to do that. So, companies are either trying to find simpler ODDs to solve or doubling down on solving the harder ODDs.”

M&As and partnerships

As well as technical and infrastructural constraints, there are also regulatory constraints to consider. She finds that over the last 4-5 years there has been a lot of sand shifting, with mergers and acquisitions taking place where one player within the autonomous vehicle ecosystem market acquires others, and then there has been the forming of partnerships.

Essentially, automakers, regulators, and others within the AV ecosystem find that they need to work together to overcome any technical or regulatory challenges they face. The primary catalyst for this is the need to prove the safety case for autonomous vehicle technologies. Different approaches are nevertheless taken to achieve this outcome. For example, some players may manufacture their own sensors while other may outsource this to other companies or buy them from other firms.

Shao says some players may choose to work closely with suppliers to provide the parts and materials they need, while designing their own autonomous vehicles. There are others that will focus just focus on the vehicle, while outsourcing the entire autonomous vehicle stack, and then there are others that integrate it all.  So, there are different strategies in how they think about partnerships.

So, in conclusion, she reiterates that autonomous vehicles will take decades to achieve ubiquity everywhere and anywhere on public roads. Meanwhile, warehouse robots will continue to become more widespread and used within their constrained ODD.

There will nevertheless be incremental progress in having autonomous vehicles on the world’s highways. This will largely be concentrated on pilots as autonomous taxis in the “anytime, anywhere” scenario will be decades off. At present they are also operating in pilots, within very limited ODDs, which over time may need to expand. So, yes, robots are closer to ubiquity than autonomous vehicles.