If you believe self-driving cars are already here, you would be wrong, especially as the number of fatalities from "driverless-car" accidents over the past couple of years seems to indicate. It’s far from an epidemic, but there have been enough fatalities related to semi-autonomous technology to make the case that we’re not yet ready to remove our hands from the wheel.
The confusion primarily arises from the loose usage of the term "autonomous" when applied to safety and driver-assist systems found in most of today’s new cars. Forward-emergency braking, lane-keeping assist and even adaptive or smart cruise control are such systems. Yes, much of this technology will appear in some form in future cars that don’t require a human driver. But in and of itself, this technology doesn’t make a new car today an autonomous vehicle, not even when several of these systems operate together. To most of us that seems fairly obvious, but based on the number of deaths to date, not everyone gets it.
There are those calling for a change in how we label the different stages of vehicle autonomy. The current SAE levels of autonomy followed by the industry and government were established years ago, before the realities of achieving complete autonomy were fully understood. Now that we are waist-deep in data and feedback from stakeholders and drivers, it may be time to reevaluate the labeling protocol.
A recent paper in the Field of View makes a pretty solid case for revisiting the SAE autonomy levels, and perhaps eliminating at least a couple of them.
Current Levels of Autonomy
Here are the current levels of autonomy recognized by the government:
Level 0 — No automation with the driver in total control.
Level 1 — An automated system, like stability control, that can sometimes assist the human driver.
Level 2 — An automated system, like adaptive cruise control, that can actually conduct some driving tasks with the human doing the rest.
Level 3 — In limited instances the automated system can monitor the environment and perform a driving task with the human prepared to retake control.
Level 4 — In specific situations the automated system can monitor the environment and perform driving tasks without human involvement.
Level 5 — All driving tasks in all situations and conditions are performed by the automated system.
The Argument for Change
Proposing a change, the Field of View paper takes the position that currently only the even-numbered levels have any real merit in terms of safety.
Level 0 — Such cars will always exist even if only in museums and private collections.
Level 1 — While systems like stability control have saved lives, level-1 systems don’t do enough.
Level 2 — It is a manageable and reasonable mix of human and machine involvement (for example, emergency braking or adaptive cruise control).
Level 3 — It puts too much confidence in the human behind the wheel to pay attention and regain control when needed. Tesla’s Autopilot approaches level 3.
Level 4 — It is currently achievable, but operational parameters must be established and adhered to, such as expressway-only or downtown corridors-only.
Level 5 — Although this remains the ultimate goal of most stakeholders, it is now recognized that complete autonomy, if even achievable, is years or decades away.
What It Means to You: Someday self-driving cars may be a reality, but no time soon. In the meantime, the technologies available in new cars are semi-autonomous. They aren’t meant to assume complete driving control. The human behind the wheel is still totally responsible.