Hey,
I've been exploring FSD data from Tesla FSD Tracker and doing some curve fitting just for fun.
Steps:
Data Scraping: I scraped the data from the version table page.
Data Aggregation: I aggregated the data (city miles driven and critical disengagements) for each major release. For example, data from versions 12.5.1 and 12.5.2 were combined with 12.5.
Confidence Interval Calculation: I calculated the 80% confidence interval using the formula listed here.
Curve Fitting: I tried fitting an exponential curve using SciPy.
Future Prediction: I calculated when Tesla might reach 17,000 miles per Critical Disengagement Event (CDE).
To give Tesla some leeway, I assume they can work on other Level 4 (L4) issues like safe disengagement while improving MTTF. Authorities approve it on day one, and all other disengagements can be resolved in the meantime. Additionally, I assume Tesla won’t hit any hardware limitations before deploying RoboTaxis. The day FSD reaches 17,000 miles per CDE is what I refer to as "RoboTaxi Day."
Graphs:
If we fit the curve using the MTTF data, RoboTaxi Day is projected to be 09-03-2036.
If we fit the curve using the 80% CI upper-bound data, RoboTaxi Day is projected to be 06-06-2029.
Thoughts:
Even in the best-case scenario, Tesla's RoboTaxi is still about 5 years away.
The adoption of end-to-end FSD does not significantly change the MTTF trend. The MTTF has remained consistent from FSD 10 to FSD 12.
Interesting post! What definition did you use for "Critical Disengagement Event" and where did you get the 17k miles per CDE target?