Published on April 21st, 2019 | by Kyle Field0
Speed Dating: Getting To Know Tesla’s Autopilot
April 21st, 2019 by Kyle Field
Tesla’s Autopilot has been hailed as the ultimate platform for developing a fully autonomous driving solution. After 8 months of owning our Tesla Model 3, Tesla shrunk down the base Autopilot package by removing a few features, and at the same time dropping the after-purchase price from $7,000 for Enhanced Autopilot to $4,000 for the new Autopilot package. It felt like the right time to pull the trigger on the update, so we made the purchase on March 1st.
In the article below, I have attempted to summarize my experiences with the solution, but your mileage may vary, as Tesla’s vehicles are running a number of variations of hardware and its software is evolving very rapidly. Tesla takes the learnings from its fleet of deployed vehicles and turns those around in new revisions of software that build incrementally on past revisions in an incremental but steady march towards a more fully autonomous solution.
The Autopilot solution was supposed to come down to the car in 3–5 business days, but there was an issue on Tesla’s side that prevented the company from seeing the $4,300 payment (with tax). A phone call and a few screenshots confirming payment from my end resolved the issue and a few days later, with no celebratory popup or notification, Autopilot was available in my Model 3.
Yes, it arrived without a red carpet or even a popup notification. On a day like any other, we headed out in the Model 3 and I noticed that putting cruise control on resulted in the blue cruise control bubble, indicating Traffic Adaptive Cruise Control (TACC) was engaged. That blue color was instead of the normal black cruise control indicator. My ears perked up and I quickly attempted to engage Autosteer, only to find out that it did not work.
A quick flip into the settings revealed that the beta feature was disabled by default and can only be engaged when the vehicle is in Park. A quick pitstop on the side of the road and several quick changes in the options screen (Settings -> Autopilot), and Autosteer beta was enabled. As with many momentous occasions in life, it felt like there should have been more of a celebration — in recognition of leveling up in life, for the safety improvement our vehicle had just made, for upgrading to the car from the future — but nothing happened. It should be, or could be, more of a celebratory occasion, because the implications (and the purchase price) are huge. Pulling myself back down from the clouds, I pulled back into traffic and started playing with the system.
As a driver, Autosteer appears to do the majority of its side-to-side tracking based on lane lines. These are detected and displayed on the 15″ touchscreen regardless of whether the vehicle has Autopilot installed or not. On top of that base, it layers in objects like the cars, trucks, motorcycles, and bikes around the car, and even 2 cars away in some cases. The display proved to be a helpful foundation for me to understand what the vehicle saw and why it might be reacting the way it did as it took over steering for me. Elon noted in his recent interview with MIT’s Lex Fridman that Tesla developed the customer-facing display to show humans what Autopilot saw in a way that’s easy to interpret. The display is not comprehensive, but it helps to wrap your mind around what the car sees and what it is reacting to.
The art of keeping a hand on the wheel when Autosteer is doing the heavy lifting is not one that is easily mastered. When Autosteer first took over the steering of my car, it was a bit timid, as if it was still getting familiar with the vehicle. Seeing as how my car is virtually identical to nearly every other Tesla Model 3 out there, it came as a bit of a surprise to me. I was expecting the solution to have taken advantage of the much-touted fleet learning that the thousands if not tens of thousands of Teslas that had already driven my route would have learned. The car would follow the lines of the road as if they were the letter of the law. When a new lane merged with mine, the car would immediately recenter itself with a sharp, drastic turn. It was not reassuring, but I hoped and trusted that it would improve.
Give Up Control
Using Autosteer those first few times, I was probably more nervous than it was at letting the car do most of the heavy lifting when it came to steering the car along the road as I maintained a close view of the road and a tight grip on the wheel. My body was tense, my grip was firm, and my body was taught, ready to take over at any instant. Part of that was valid, but most of it was just my instinctual reaction to letting the car do most of the heavy lifting when it came to steering the car. After several decades of manual driving, not driving is a hard thing to adapt to. It’s not natural.
After a few trips with Autopilot engaged, I gradually started to realize what the solution was doing and why. I knew that it liked driving much closer to the right hand lane line, and that was okay. It was also much better than I was at holding a line in the lane and not accidentally drifting into the center divider or across a lane line. Autopilot also had those eyes in the back of its head that my mom had always threatened me with — it was able to not only drive, but to keep a watch on the cars around it for the potential bad drivers around me. That alone is far better than what any human can possibly do on their own and drastically improves the safety of the vehicle.
Learning to drive with Autopilot engaged is like learning to dance with a new partner. It takes some time, as you would expect, but once you get used to it, it can be a great thing. There are still a few missteps. I sometimes take over steering when I don’t need to or when it makes a turn that is a bit too abrupt for me. But we’re getting used to each other and I’m hopeful about the long-term prospects.
I’ve learned to keep a casual but firm hand on the wheel, with extra caution applied when coming up to any merges, tight lanes, or transitions. In the weeks following installation, Autosteer graciously mellowed out. The first few days were tense, much like driving with a teenager who you trust but who simply has to learn for themselves how to actually use the car that they are now in control of. Several days later, it matured into a 20-something year old driver with more experience under their belt. Still, it has not fully mastered the art of relaxed driving. The car approaches stopped traffic at a speed that is not conducive to trust, followed by a rather abrupt stop. When traffic gets moving again, Autopilot takes its sweet time getting the car rolling, leaving a larger than normal gap from the car in front.
Ideally, Autopilot would just drive like a “normal” driver, just safer. That’s what we all want, and as soon as we learn to let to go of our cash, we can start the process of learning to let go of the steering wheel for that. Tesla’s deployed vehicles have already racked up more miles than all of the other companies that are developing autonomous driving technologies combined, according to CEO Elon Musk. After speaking with Musk on a recent podcast, ARK Invest said that, “Tesla vehicles equipped with first- and second-generation Autopilot hardware, for example, have collected a cumulative 8 billion miles to date, while Waymo’s vehicles had collected roughly 10 million miles as of October of 2018.”
That’s a pretty significant moat that Tesla is already using to sell more electric vehicles, keeping our air cleaner and our roads safer at the same time. This latest pricing change advances both of those goals, and as such, is something to celebrate.
You can listen to CleanTechnica’s Zach Shahan talk at length with ARK Invest’s Tasha Keeney about Tesla Autopilot hardware, Tesla Autopilot software, and Elon Musk in the following interview (broken into two parts):
Tesla CEO Elon Musk is known for his overly optimistic timelines, but as the years go by (as they tend to do), Tesla appears to be narrowing down the timeline for deploying its Full Self Driving solution into production. He said on the ARK Invest podcast that the “feature complete” Full Self Driving solution will be pushed out by the end of 2019, which feels likely. He expected that the solution would be fully refined and ready for owners to “set it and forget it” (my words, not his) by the end of next year. Then owners can just let their cars drive them to their destinations — assuming regulations allow it.
Autopilot has improved in just the handful of weeks that it has been installed on the CleanTechnica Tesla Model 3 and each update brings the solution one step closer to that future where I can actually sleep on the way home from the airport instead of fighting to stay awake and on the road. It holds the promise of making transportation available to those who cannot drive due to physical disabilities, age, or intoxication.
From the early progress with Autopilot and from the more obvious progress that we have seen in recent weeks, I’m hopeful that Tesla’s Autopilot and Full Self Driving solutions will indeed change the way we get around and even how we live life in the years to come. It has me dreaming about new possibilities, about a future that is downloaded to my car, one over-the-air update at a time.
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.