Tesla Full Self Driving Beta: How Close to Level 5 Autonomy?

Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!

Background: I purchased my Tesla Model 3 on October 22, 2019, and have 66,034 miles on the odometer. I’ve only had one significant repair, a squeaky left front suspension joint fixed under the warrantee. I have had to pay only one significant maintenance cost: two sets of tires for a total of $1143. I’ve driven my Tesla from the US East Coast to West Coast using the fabulous Tesla Supercharger network and I’m still just as crazy about the car as the day I bought it.

I’ve been using Tesla’s FSD Beta obsessively in my Model 3 now for 150 days.

On December 1, 2021, I passed Tesla’s safety test and was able to download FSD Beta V.10.5 to my car. Since then, I’ve downloaded three updates to the software: V10.8, V10.10, and recently V10.11.2. I use the software every day. The longer I use it, the more I can observe whether negative behavior is a unique event, a rare event, or a consistent failure.

It still blows my mind that FSD Beta will drive you from your current location to the place you enter into the navigation without intervention for some routes. It changes lanes, makes turns, adjusts speed, etc., etc. to navigate on city streets, onto Interstate highways, and off again. I feel very privileged to be one of ~100,000 drivers in the world to be testing a system which allows my car to drive totally autonomously, at least in some situations.

So, the $64,000 question is: What must Tesla change for FSD Beta to reach Level 5?

How close is FSD to being totally autonomous, requiring no interventions, and being significantly safer than human drivers? I am now using the third version of FSD Beta software that I have had access to. These are my observations of what Tesla needs to fix to get to Level 5:

1) Choosing the correct lane: FSD Beta V10.11.2 will choose the correct lane at an intersection most of the time. However, this is a serious problem because I have observed numerous times where it will put you in the wrong lane rather than the correct lane specified for your navigation route. Also, sometimes it will turn into a wide bike lane on the right instead of the correct traffic lane. This could be caused by errors in the map data, but as the human driver, I don’t need a map and I don’t have a problem reading the arrows on the pavement or signs that indicate the correct lane. Also, FSD Beta will sometimes fail to get into the turn lane in time to make the turn. As a human driver, I observe the traffic conditions and change to the correct lane very soon if the traffic is heavy. I almost never miss a turn because I didn’t get into the turn lane soon enough. In order to reach Level 5, FSD will have to select the correct lane more of the time.

2) Unprotected turns in heavy traffic: FSD Beta works extremely well making protected turns (turns at a traffic light). However, for unprotected turns, FSD Beta doesn’t adequately judge when there is a break in the traffic and then make an aggressive move to fit into the break smoothly. The problem could be ameliorated by working with the navigation to find a traffic light when turning onto a busy highway. In order to reach Level 5, FSD will need to be able to make unprotected turns automatically onto busy highways.

3) Speed bumps and dips: In my neighborhood in Utah, we have frequent speed bumps and dips. In this relatively dry climate, the sewers frequently run as a dip on top of the road instead of in a pipe under it. In both cases, if you don’t slow down to 15 mph or sometimes less, you will risk damaging your suspension, bottoming out, or jarring your teeth out. In order to reach Level 5, FSD will need to observe speed bumps and dips and slow down.

4) Phantom braking: Too often, FSD Beta will reduce speed or brake for no apparent reason. In order to reach Level 5, FSD will need to differentiate between legitimate dangers and those that are not.

5) Phantom swerving: Occasionally, FSD Beta will swerve for no apparent reason. In order to reach Level 5, FSD will need to differentiate between legitimate dangers and those that are not.

6) Premature stopping: At some stop signs, FSD Beta will cause your car to stop 15–20 feet early. In order to reach Level 5, FSD will need to stop at the correct place at every stop sign.

7) Running stop signs: There is one stop sign turning from 1650 W onto Snow Canyon Parkway in Saint George, Utah, where FSD Beta will always run the stop sign. In order to reach Level 5, FSD will need to stop at every stop sign.

8) Stopping at rotaries: FSD Beta will stop at the entrance of rotaries even when no cars are present. In order to reach Level 5, FSD will need to treat rotaries the same way it treats yield signs.

9) Timid behavior at stop signs: FSD Beta will proceed slowly after stopping at a stop sign. In order to reach Level 5, FSD will need to accelerate briskly after stopping at a stop sign.

Chip in a few dollars a month to help support independent cleantech coverage that helps to accelerate the cleantech revolution!

These are only my personal observations. Other drivers in other areas with different conditions may have other observations. Please contribute your observations in the comments section!

From the improvements I have seen in three versions of FSD Beta, do I see I see significant progress toward Level 5 autonomy? Short answer: I won’t be holding my breath.

What does Tesla FSD cost? Is it worth it?

If you want FSD right now, you must pay $12,000 for it, on top of the price of a Tesla. Elon Musk says the price will keep going up as FSD gets better. If that is too big a pill to swallow for an unknown capability, Tesla also lets you try out FSD Beta as a monthly subscription for $99 or $199 per month, depending on the level of driver assist you have already paid for.

The smart cruise and autosteer that come with every Tesla are incredibly worthwhile. Would I pay $12,000 for Full Self Driving Beta? If you are a techno nerd like me, it is great fun and very interesting to observe the state of autonomous driving now and watch improvements. Would I pay $12,000 out of a very tight budget for software which at this time has marginal utility. No!


Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.

Latest CleanTechnica.TV Video

Advertisement
 
CleanTechnica uses affiliate links. See our policy here.

Arthur Frederick (Fritz) Hasler

Arthur Frederick (Fritz) Hasler, PhD, former leader of NASA Goddard Space Flight Center Scientific Visualization & Analysis Laboratory (creator of this iconic image), and avid CleanTechnica reader. Also: Research Meteorologist (Emeritus) at NASA GSFC, Adjunct Professor at Viterbo University On-Line Studies, PSIA L2 Certified Alpine Ski Instructor at Brighton Utah Ski School.

Arthur Frederick (Fritz) Hasler has 121 posts and counting. See all posts by Arthur Frederick (Fritz) Hasler