3 Reasons We Haven’t Reached Level 5 AV Deployment

By Tiera Oliver

Associate Editor

Embedded Computing Design

April 08, 2021


3 Reasons We Haven’t Reached Level 5 AV Deployment
Image Courtesy of IStock

Where are we on the road to Level 5 autonomous vehicles (AVs)? We’ve all seen fully autonomous Ubers and test vehicles driving around.

But when it comes to widespread deployment of fully autonomous vehicles, the truth is, we’re not anywhere close.

The Society of Automotive Engineers (SAE) uses a six-level system, from 0 to 5, to describe the features and capabilities of vehicle autonomy. It’s important to understand these before we can understand the technical (and other) reasons Level 5 autonomous vehicle technology has not reached critical mass. Each of the six levels is defined below.

  • Level 0 – No automation. Pertains to your everyday, traditional car with no automated or active safety features and full driver involvement.
  • Level 1 – Driver assistance. Consists of radars and cameras for distancing, automatic braking, lane assistance, and adaptive cruise control.
  • Level 2 – Partial automation. The car has the ability to accelerate, brake, and even steer under certain circumstances. Still needs driver involvement for the majority of tasks.
  • Level 3 – Conditional automation. The car is able to drive itself but only under the right conditions and with certain limitations. Driver involvement is still ideal as automation can be halted at any time.
  • Level 4 – High automation. Vehicles can drive themselves without driver interaction, but can still stop under certain circumstances. Regulations and legal obstacles are heavily involved here.
  • Level 5 – Full automation. A true driverless car that can operate on any road and under any conditions with no driver involvement needed.

Of course, some of these capabilities have made their way into mainstream production vehicles. On the contrary, many modern cars now include features that lie between 0 and 3.

At Level 4, the complexities of handing off full control between the vehicle and a human driver, often at a moment’s notice, are so immense it often seems designing to this level is altogether unrealistic for most passenger vehicle use cases. In some ways, Level 5 can be viewed as the challenges of Level 4 extrapolated out over the entire road environment, at least in the eyes of the general public.

And the challenges keeping industry from these higher levels of driving automation are not purely technical implementation challenges. They include public perception, legislation, and simply figuring out what we don’t know.

Reason 1: We Haven’t Found All the "What Ifs”

From the perspective of Chad Chesney, General Manager of NI’s transportation business, “the main challenge that we face today is really a problem of not knowing what we don’t know. The world is infinitely variable, and humans are able to quickly adapt to that variability. We need to be able to provide evidence that our AVs are as safe or safer than human drivers, and that requires us to demonstrate vehicle safety in an unbounded number of scenarios.”

The scenarios Chesney describes aren’t just related to AVs themselves, but anything and everything they could come in contact within the driving environment. This includes outside influences such as pedestrians, road signs, and one of the biggest wild cards – inclement weather that could potentially disrupt the vehicle’s ability to follow the rules of the road altogether.

Chesney believes that in order for automotive vehicles to be safely deployed on public roads, AV testing needs to reach the point of ultimate redundancy where vehicle data is instantaneously processed and uploaded to the cloud so it can be shared.

By leveraging data that is already useful, we can reduce the amount of time and effort that is dedicated to developing test cases for the same scenarios. Test data sharing would also promote collaboration between automotive engineers across the AV supply chain who are all working to overcome the myriad corner cases driving environments present. And eventually, together, it is possible that the industry could reach a point where there just aren’t any more scenarios to test.

For example, NI and Konrad Technologies have together developed high-bandwidth recording systems to collect and store high-fidelity data from vehicles on the road today. The solution consists of executing real-world sensor data on an ECU and testing the ECU’s path planning and control system algorithms.

“We need to combine all of these systemic methodology improvements with guidelines that help us understand when we are done,” said Chesney. “We need to develop an infrastructure that defines what it means to be safe, and we need the people to believe us when we say we have achieved safety because consumer trust is required.”

Test is how we increase public confidence in the technology,” he added.

Reason 2: Complex Supply Chains Complicate AV Development

Like Chesney, Pedro López Estepa, Director of Automotive at Real-Time Innovations (RTI), acknowledges that there is still a great deal of work required before we can get to widespread deployment of high-level autonomous driving technology in passenger vehicles. But there’s another sector making faster strides towards autonomy – the long-haul trucking industry.

According to Estepa, the trucking industry has a few advantages that allowed it to gain a head start on the implementation, testing, and deployment of autonomous drive technology. Most of them center around the use case, which happens to be narrower than that of consumer vehicles. The long, straight highway transport scenarios associated with long-haul freight present a comparatively static driving environment to the hustle and bustle of urban and suburban driving, and occur almost exclusively on public roads. These factors helped minimize some of the legal headaches associated with getting semi-autonomous fleets into the field and generating data.

As Chesney pointed out above, the greater availability of test data helped autonomous trucking technology providers improve their systems to the point that they have nearly perfected the use of lower-level autonomy technologies such as auto pilot and truck platooning, the process of linking multiple autonomously-driven trucks en route using connectivity. And it was easier to share knowledge and experience across the sector because most trucking fleets are so similar that totally unique autonomous system designs were not required.

“The trucking industry is really well-defined and consistent. The supply chain is simpler than the traditional OEM,” said Estepa. “The amount of Tier 1s that dedicate designs for trucking makes it simpler to adopt new business models without the complexity of the traditional tiered supply chain that the traditional automotive OEMs have.”

The good news for the rest of the automotive industry is that those reduced-complexity supply chains are now making their way into emerging vehicle designs.

“We see the same in many of the new EV startups,” Estepa said.

But another hidden advantage of the trucking industry is the payoff. There is direct ROI associated with commercial autonomous trucking that isn’t so readily apparent in consumer transportation, such as increased drive time, better fuel economy, and so on. Without an obvious near-term economic driver, the consumer AV technology market may not have the urgency of the commercial trucking market.

Reason #3: No Eyes Inside

That being said, there continues to be an expansion of completely autonomous robo taxi and shuttle pilots in highly populated, public areas. According to Vayyar’s Head of Automotive, Ian Podkamien, aside from the monetary potential a big reason these deployments are taking off is that they are being conducted in controlled areas, with controlled speeds, and following specific driving routes. Because the environment is more contained for these vehicles than it would be for automated consumer vehicles, the regulations surrounding these vehicles are limited, quite similar to trucking fleets.

Despite this, we still haven’t addressed the challenges of Level 4 systems that require a handoff of control from vehicle-to-driver and driver-to-vehicle during operation. Here, Podkamien raises the question of whether there should be increased safety and monitoring technologies placed within the cabin of the vehicles, which can be achieved with Vayyar’s Radar-on-Chip. And those can be leveraged to ensure a driver is ready, willing, and able to assume control of a moving, autonomously driving vehicle or in public transportation settings to assess passenger activity, behavior, and wellness.

“The vehicle itself needs to start developing some awareness about what is going on inside the car,” Podkamien explained. “You need to be able to monitor what’s going on.”

Full AV Deployment: The Road Ahead

With all these factors in mind, Podkamien believes that deploying fully autonomous cars right now sounds like more of a wish than something that will be achieved any time soon. There is just too much information that is still not well defined.

For instance, who would be liable if accidents were to occur involving autonomous cars? How would insurance handle those claims?

OEMs, their suppliers, insurance companies, drivers, and everyone else who plays a part in the automotive value chain have a seat at the table of the AV conversation. More discussion – and collaboration – is needed before we can start truly addressing the questions that have limited widespread Level 4 and 5 AV deployment.


Tiera Oliver, Associate Editor for Embedded Computing Design, is responsible for web content edits, product news, and constructing stories. She also assists with newsletter updates as well as contributing and editing content for ECD podcasts and the ECD YouTube channel. Before working at ECD, Tiera graduated from Northern Arizona University where she received her B.S. in journalism and political science and worked as a news reporter for the university’s student led newspaper, The Lumberjack.

More from Tiera