Self-Driving Cars: A Multi-Billion Dollar Science Fair Project

We were told that self-driving cars were the next phase of mobility. Almost a decade later, the promise of driverless vehicles has not been kept. What happened?

It wasn’t supposed to be this way.

As recently as 2017, automotive and technology industry leaders were touting the disruptive potential of autonomous vehicles (AVs), forecasting an upside economic opportunity of up to $7 trillion by 2050. Major automakers such as Ford and GM placed AVs at the forefront of their technology portfolios, investing billions every year to create and test AV technologies with the intention of introducing them to the public.

Startup companies were popping up everywhere, with more than $60 billion invested in 2018 alone. This led to partnerships, such as that between Lyft and Aptiv, which created AV shuttle services in Las Vegas. Waymo, a division of Google, launched a commercial AV taxi service in Phoenix, AZ. And Michigan-based May Mobility recently completed 10,000 shuttle trips in the greater Grand Rapids area.

But common to all of these taxi and shuttle services was the presence of a human behind the wheel, plus the vehicles were traveling along specifically designated routes in carefully controlled conditions in specific cities. These services were not actual services, rather they were perpetual testing grounds for technologies that have yet to deliver on the promise of the full self-driving experience.

Our Robotaxi Future has been Postponed

Not long ago, we heard of bold predictions by automakers and tech developers of our robotaxi future, where on-demand mobility would be handled by driverless vehicles offering a variety of voice-activated and touchscreen commands as well as door-to-door destination choices. Level 5 autonomy would be required to deliver on the entire robotaxi experience, meaning the technology must enable full vehicle automation when navigating streets, assessing and reacting to pedestrians or obstacles, and sharing the road safely with human drivers, all without human assistance. Level 5, which the auto and tech industries were hoping to achieve by 2020, is now postponed indefinitely.

Anyone with a working knowledge of the myriad of AV technologies and the infrastructure required to make it all work could have predicted this postponement. Building a prototype vehicle that can drive itself on a closed course (or along a carefully selected stretch of public roadway) is not a proof of concept; it is a laboratory experiment. With enough money and skill, a self-driving vehicle can be built by just about anyone. No, the fact that AVs can be built is not indicative of our AV future, any more than someone building a robot chef is indicative of everyone having automated meal service in our homes by next Thanksgiving.

Autopilot is Not the Same as Self Driving

In early 2019, when Tesla forecasted a million robotaxis without human back-up drivers in service by mid-2020, financial analysts were not convinced. They noted the erratic behavior of the AVs, how the cars failed to recognize obstacles, hesitated to change lanes and missed turning onto a ramp.

Tesla calls its driver assist feature AutoPilot, which is quite misleading as it has nothing to do with how actual autopilot technology is used in aviation. This has led to a growing public mistrust of self-driving vehicles as reports emerge of serious and fatal crashes to passengers while using the AutoPilot feature. To be fair, AutoPilot is not positioned as a self-driving feature (it augments the on-board driver safety systems and allows for a Level 3-type of autonomy) however it is not as reliable as portrayed, which in turn further damages public perception.

Technical, Infrastructure and Legal Obstacles

To state that we currently have usable technology today for AV adoption on a mass scale is to be naïve to the complexities and interoperability of on-board and off-board technologies and systems required to make self-driving cars a reality. These vehicles not only need to see the environment around them, they must also understand what is happening, what is about to happen and perform the correct tasks in a reliable and robust manner. That requires a high degree of artificial intelligence and the ability of the system to “learn” from the driving experience. That technology, dubbed “machine learning,” is still in its infant stage.

AVs must also rely on a connected infrastructure that does not exist today. Technological and operational standards for the AV connected infrastructure have not been established, costs have not been determined and laws have not been written to regulate the rollout and usage of AVs to any degree of scale.

While there are practical, controlled uses of AV technology having limited physical boundaries (e.g. agriculture, mining, airports, military bases, etc.), commercial taxi and personal use cases are mired in unworkable technological, infrastructure and legal challenges that have no solution for the foreseeable future.

There is also a growing concern that AV technologies used on a scale required for commercial taxi use will be a huge financial drain on technology and automotive partners that would take many years from which to recover, if ever. As a result, many developers and automakers have either scaled back their AV development or put the program on hold indefinitely.

The promise of self-driving vehicles was made by auto and tech executives to boost their capabilities to shareholders and to grab venture capital. These “visionaries” did not recognize – or they ignored outright – the real-world challenges faced by AV delivery and consumer adoption. Billions of dollars later, their little science fair project did not get the blue ribbon they were expecting.