Tesla could have done better with the naming of their driver assistance and autonomous driving features.
With AP2+ vehicles, Tesla really has 3 levels of features:
- Driver assisted operation on limited access highways and limited features on other roads. This is the standard AutoPilot in the latest vehicles or Enhanced AP in AP2 vehicles, and much of this available is becoming available from other manufacturers today. Limited access highways have the simplest environment - without the complications on urban streets (pedestrians, parked vehicles, intersections, traffic lights, stop signs, …).
- Driver assisted operation on almost all roads and under almost all conditions. This is Full Self Driving, before regulatory approval. This is significantly more complicated than the standard AutoPilot because the software has to operate safer than a human in many situations, especially complicated due to the lack of standardization and the unpredictable actions of the objects being tracked. Anyone purchasing FSD is primarily going to see this level of operation, and an expanding number of roads/conditions for most of the time they own their vehicle.
- Fully autonomous driving (no driver interaction) requiring regulatory approval. This is what everyone is shooting for. We will see FSD vehicles in limited conditions soon - such as trucks driving on limited access highways or delivery/taxi vehicles driving in limited areas (moving slowly or stopping when something unexpected happens).
With our 2017 S and 2018 X we purchased FSD, under the expectation that we'll likely achieve the FSD features in driver assistance mode for most (if not all) of our ownership period. If Tesla does receive regulatory approval for FSD, it may happen in stages - with approval first on limited access highways. FSD on urban streets is much harder - especially driving at the same speeds as a human driver - and will take longer to get approval for all roads.
Tesla has made two major design decisions not shared by the other manufacturers. They (Musk) believes they can implement FSD using less expensive 360 degree cameras, front facing radar and proximity sensors vs. using more expensive LIDAR. In theory, since humans don't have the equivalent of LIDAR, it should be possible to implement FSD without LIDAR, though using LIDAR should make object tracking (distance, movement direction, speed) much easier.
The other major difference is Tesla's AI approach with fleet learning. Musk believes they will be "feature complete" with the FSD software soon. This doesn't mean FSD will be ready - it means the major functionality will be implemented in the software, waiting for the AI rules to be developed that will actually drive the vehicle. Because Tesla has access to data from all vehicles produced since late 2016, including the huge number of Model 3's being delivered, Tesla has a gigantic advantage if they can utilize all of this data to develop and fine tune the driving rules. Other manufacturers are relying more on simulation and the use of a limited number of test vehicles; with Tesla's fleet learning driving their AI rules, it's possible that once the FSD software is "feature complete", Tesla may be able to move much more rapidly to safe operation (in driver assist mode) in more conditions than the other manufacturers.
Regulatory approval is an unknown - coupled with managing liability when accidents happen (which cannot be eliminated).
Rather than getting FSD approval for all roads and conditions, what we're likely to see with Tesla (and other manufacturers) is FSD approval in limited conditions - such as on limited access highways or operations on surface streets at reduced speed (which is likely what we'll see for FSD delivery/taxi applications initially).
Will Tesla deliver FSD approved to operate everywhere sometime next year - highly unlikely - though I do expect we'll start seeing more value in owning the "FSD" feature in the next 12 months - and with Tesla's fleet learning/AI strategy, new functionality could be added pretty quickly.