Tesla’s auto pilot feature not only brings autonomous functionality to its vehicles but each vehicle’s experience is collected centrally and the cumulative experience of autonomous driving from all vehicles is passed back to each individual vehicle in a feedback learning loop. The result is continuous improvement in the vehicle’s autonomous behaviour.
The behavioural adjustments are sent over the air via scheduled updates. This shared experience amongst drivers is leading to incremental improvements in the autopilot features every day so that when one car learns something every car learns it.
Elon Musk referred to Model S drivers as “expert trainers” as every vehicle is training every other vehicle. Vehicle owners have already been reporting improvements on the autonomous technology in the Tesla forums.
Tesla Self Learning explained in a video:
This technology by the way is not exclusive to Tesla – Google and Audi also have machine learning algorithms for their vehicles, meaning that data collected from each vehicle is shared around the vehicle fleets.
Autopilot has already been credited with preventing a head on collision at 70kph in the US (night time and looks like some light rain on the windscreen). See the video below:
And from the Moscow Tesla Club comes this video of a Model S avoiding a collision with a taxi that cut it off.
Autopilot is still officially in ‘beta’ mode and is recommended at this stage for highway driving as it’s not fully capable of navigating suburban or city environments – yet!