Tesla FSD Milestones

Telsa’s Full Self-Driving feature has struggled for many years to provide the expected features of a self-driving car. FSD is awesome! FSD is way ahead of any competition in providing self-driving features on any road. But, there is also lots of misunderstanding of the feature.

Tesla FSD is currently an SAE Level-2, advanced driver assistance system.

“Autopilot, Enhanced Autopilot and Full Self-Driving Capability are intended for use with a fully attentive driver, who has their hands on the wheel and is prepared to take over at any moment. While these features are designed to become more capable over time, the currently enabled features do not make the vehicle autonomous.”

https://www.tesla.com/support/autopilot

Self-Driving is Difficult

I believe that Tesla’s FSD feature will continue to progress in capability, but the difficulty of providing a self-driving car on any street continues to be underestimated. I am especially dismayed by those who espouse that at some point self-driving capability will magically appear. For example, the term “Solve FSD” is often heard, but I find this meaningless, unuseful, and even detrimental to moving forward to the self-driving goals. Unfortunately, it seems that even Elon continues to promote this point. It is unclear to me, if he truly believes this, if he sees this as the correct company line, or if I’m just wrong.

The good news is that I have figured out a way to help dig into this difference of opinion. By looking at some milestones that must be achieved along the way to say an SAE L5 autonomous vehicle, we can start to understand and appreciate some of the difficult hurdles that must be cleared.

Of course, another problem is the lack of specificity around what it means to have a self-driving car.

If you view the discussion with this poll, you will see that the statement of “driverless-FSD” is so vague that asking a question like this has ZERO meaning, and PROMOTES the continued cloud of (mis)understanding around what it means to have a self-driving car.

SAE Levels are of limited use

The SAE 5-levels of auto driving are almost always used when trying to characterize a self-driving car. I see serious issues with this as it leaves out the most important aspect of a self-driving car, the liability for the actions of the car. Waymo and Cruise take responsibility for the actions of their cars. Telsa pushes the liability for the actions of its cars to the drivers (like any driver assistance system). I wrote a bit more about this here if you’d like more discussion on this.

The Milestones – Second Draft

May 25th, 2023, I tweeted my first draft of some milestones for getting to self-driving Tesla vehicles. I slightly update below and give a bit more definitions. I have added some guesses at dates. I did not really try to be exact as it is so difficult to predict the future. My primary purpose is to detail some of the steps along the way to a self-driving car future.

• 2024: Tesla Full Self-Driving City Streets Feature (often called FSD) out of beta. (SAE L2)
  • This seems to be a very specific milestone. The beta feature started going to select public users in Oct, 2020.
• 2025: Tesla-owned, geofenced, Driverless RoboTaxis are publicly tested. (SAE L4)
  • Telsa at one point said we don’t need a $25k car because we will have RoboTaxis. The “next gen, Model-2” platform seems to be the basis for this car and I believe it will initially have 2 variants. A public version will be a scaled-down less-expensive Model Y. And the second variant will be a Tesla-owned RoboTaxi version that will be used to compete with the likes of Waymo and Cruise in a similar fashion.
• 2026: Tesla AutoPilot enables unsupervised driving in personal cars. (SAE L3)
  • “Unsupervised AutoPilot” allows for geofenced unsupervised driving with a 5-second abort-to-human-driver warning window. This is similar to the Mercedes-Benz DrivePilot system. Of course, there will be a slow rollout with expanding users and road availability. SAE L3
• 2027: Tesla-owned, geofenced, Driverless RoboTaxis on 10,000 miles of roads. (SAE L4)
  • This is the expansion of Tesla-owned RoboTaxis making them available on more roads.
• 2028: Unsupervised AutoPilot is available on over 50,000 miles of roads. (SAE L3)
  • This is the expansion of consumer-facing advanced driver assistance systems (ADAS) to more roads. These will be interstates and limited-access freeways.
• 2029: Unsupervised AutoPilot is available on over 500,000 miles of roads. (SAE L3)
  • Half a million miles of roads represent about 10% of the roads in the US and this may be where the expansion of the Unsupervised AutoPilot feature starts to slow down.
• 2030: Tesla accepts immediate liability for accidents by Unsupervised AutoPilot. (SAE L3+)
  • Prior to this, Tesla would only selectively accept liability for accidents while Unsupervised AutoPilot was engaged. Now, Tesla accepts the initial responsibility and will selectively go after the driver if they feel the system was improperly used or abused.
• 2035: Tesla enables Personal Vehicle RoboTaxis feature. (SAE L4)
  • I think this is what most people think about when they say “Solve FSD.” They want their cars to be able to generate revenue when not otherwise used by the owner.
  • A subscription includes service will be provided by Tesla that includes:
    • Scheduling of rides and handling payment.
    • A human-lockout kit to prevent rider interference with driving.
    • Remote 24/7 support and service to handle rider issues: phone left in a vehicle, vehicle disabled (flat tire, accident, police, …), vehicle dirty, low on battery,…
    • Available in limited geofenced areas.
• 2040: Availability of Smart Summons beta.
  • This is a bit of tongue-in-cheek but does point to the difficulty of this Smart Summons feature which was promised long ago, but will have many possible issues of its own.