After a couple of yr of silence, Tesla launched a brand new model of its Car Security Report, offering attention-grabbing new information. The voluntarily printed quarterly experiences give us glimpses of how Autopilot improves automobile security.
The brand new report incorporates reviewed and up to date information concerning the variety of miles pushed per one accident registered when utilizing Autopilot expertise, not utilizing Autopilot expertise, and the US common (NHTSA and FHWA information) as a background.
New numbers for the interval 2018-2021 are totally different (greater) than earlier than as a result of Tesla found points within the previous information (some accidents have been counted greater than as soon as). With out the duplicated information, the outcomes are literally higher than beforehand (see full clarification on the backside).
An vital factor is that Tesla counts all accidents, whatever the trigger, regardless of that greater than 35 p.c of all Autopilot crashes happen when the Tesla automobile is rear-ended by one other automobile, the corporate says.
Let’s get into the main points.
Based on Tesla, the variety of miles pushed per one accident registered, when utilizing Autopilot expertise, is constantly rising, which signifies that vehicles are getting safer to make use of.
In Q3, the corporate famous one accident per 6.26 million miles pushed – a 13 p.c improve year-over-year. The document outcome was famous in Q1, at 6.57 million miles pushed.
What is absolutely vital is that utilizing the identical methodology, the result’s virtually two instances greater than 4 years in the past (3.35 million miles per one accident in Q3 2018).
With out utilizing Autopilot expertise, Tesla automobiles statistically observe extra accidents. In Q3 2022, it was one accident per 1.71 million miles pushed (8 p.c greater than a yr in the past).
There’s a vital distinction between the 2 modes of driving (with or with out Autopilot) and the distinction is rising, as there’s not that a lot progress within the case of driving with out Autopilot.
Nonetheless, we should observe that the 2 information units shouldn’t be in contrast towards one another, in addition to a quick look to have some thought of the numbers.
It is vital to notice that the outcomes are comparable solely inside a selected class (Autopilot or with out Autopilot), not between the classes because the enter information could be extensively totally different (like easy freeway driving or advanced metropolis driving). In different phrases, we will solely see whether or not the lively security programs are enhancing over time (and it is also solely a tough comparability), however we won’t evaluate Autopilot to non-Autopilot driving.
We assume that the right use of Autopilot improves security, however Tesla’s report doesn’t enable us to judge the distinction.
And beneath we hooked up a full chart with all numbers supplied by the producer plus NHTSA and FHWA information concerning the common distance between an car crash in america.
The distinction is fairly big, however as soon as once more, we should not actually evaluate the numbers immediately. The US common contains additionally a a lot older automobile fleet (with much less security programs), which alone is a giant issue. The US common presently seems to be virtually a 3rd higher than 4 years in the past (0.652 million miles vs. 0.492 million miles).
Tesla has its personal model of the chart:
- information for every setting could be collected at totally different driving eventualities (like easy freeway driving or advanced metropolis driving), which makes the outcomes incomparable between the classes
- outcomes could be affected by varied elements, together with seasonality (diminished daylight, climate situations), much less driving throughout lockdown
- NHTSA and FHWA common for the US (up to date not often) contains all vehicles, additionally previous
- Tesla’s data concerning the methodology of registering accidents:
“We accumulate the quantity of miles traveled by every automobile with Autopilot lively or in handbook driving, based mostly on accessible information we obtain from the fleet, and accomplish that with out figuring out particular automobiles to guard privateness. We additionally obtain a crash alert anytime a crash is reported to us from the fleet, which can embrace information about whether or not Autopilot was lively on the time of impression. To make sure our statistics are conservative, we rely any crash by which Autopilot was deactivated inside 5 seconds earlier than impression, and we rely all crashes by which the incident alert indicated an airbag or different lively restraint deployed. (Our crash statistics are usually not based mostly on pattern information units or estimates.) In observe, this correlates to just about any crash at about 12 mph (20 kph) or above, relying on the crash forces generated. We don’t differentiate based mostly on the kind of crash or fault (For instance, greater than 35% of all Autopilot crashes happen when the Tesla automobile is rear-ended by one other automobile). On this approach, we’re assured that the statistics we share unquestionably present the advantages of Autopilot.”
- assuming the methodology was not modified, we will see how every class improves over time
Tesla’s clarification of knowledge replace:
“…we found experiences of sure occasions the place no airbag or different lively restraint deployed, single occasions that have been counted greater than as soon as, and experiences of invalid or duplicated mileage data. Together with these occasions is inconsistent with our methodology for the Car Security Report and they are going to be excluded going ahead. These upgrades in information evaluation reinforce the constructive impression that Autopilot has on automobile security.
To make sure the accuracy of our reporting, we up to date all collision charges traditionally to account for these upgrades, together with the baseline collision charges for america based mostly on presently accessible NHTSA and FHWA information.
(Observe that for functions of the baseline collision charges in america, an car crash is one which entails at the very least one passenger automobile, gentle truck, SUV or van that’s 10,000 kilos or much less, as categorized by accessible federal information.) The top result’s that, when Autopilot is lively, the collision charges are even decrease than we beforehand reported.”