It’s all over the Internet: Tesla’s “Dog Mode” has a serious bug; it doesn’t work if the AC fan is set to manual. One headline said “Musk vows to fix.” But is there really such a bug?
Dog Mode is a cool feature Tesla added to run the AC when no one is in the car. It displays a image of a dog on the touch screen along with the car’s interior temperature, so passers by won’t be worried about the pet. I’ve used it myself a couple of time and it worked great. Of course any responsible pet owner checks up on the temperature anyway with the Tesla App.
People love their pets, and any threat of them is big news.
Elon Musk replied with one word “fixing.” I don’t call that confirmation, but damage control, and a commitment to fix a bug if there is one, but here’s what’s on the Internet:
First it’s impossible. With the current version of Tesla firmware (I have 2019.24.4), you can’t put turn on Dog Mode with the AC fan set to Manual. When you turn on Dog Mode, the car automatically sets the fan to AUTO. I tried it. Further, if you do set the fan to Manual after turning on Dog Mode, it still works just fine. I turned on Dog Mode, set the fan to manual, exited the car and watched it cool to the set temperature on the Tesla App. I tested this on a 2018 Model 3 and a 2016 Model S.
None of those stories said anything about confirming it. I consider this highly irresponsible. Maybe there is a bug, but my cars don’t have it. Who knows? Certainly not all the writers of these articles.
Update: The “Fix”
Tesla is rolling out a “Dog Mode” improvement. I got mine today (Aug. 7) in firmware 2019.28.2 along with a Chess game, volume control improvements and game controller support for Beach Buggy Racing 2.
The “fix” is to disable the AC controls (except for temperature) when Dog Mode is enabled. I still haven’t figured out what the problem was is unless it was to prevent the driver from turning on Dog Mode, setting the AC fan to manual, and then lowering the fan speed so far that it couldn’t keep the car cool. I guess we could call this fix “idiot proof dog mode.”
I’m returned from my 5,000-mile Model 3 road trip, which gave me a great deal of time to think about Tesla Autopilot as I supervised it across the US and back.
Technically NoA works as designed most of the time. It changes lanes to follow the route, it exits, it merges and it passes slower traffic on controlled-access highways. In one instance it tried to take the wrong exit in a rather complicated interchange, and in another it failed to take an exit when it should have. I caught both of them before a mistake was completed.
What NoA fails to do is operate with “common sense”; for example, a human is not going to attempt to pass a car that has slowed slightly with its blinker on approaching an exit ramp. I know the car is going to take the exit and I don’t need to pass. I also know not to try to pass a car going one mph slower than me near the top of a hill; that car is going to speed up after cresting the hill and I’ll never get around it. I tend to drive near the posted speed limit and sometimes other cars go faster. NoA will put you in the passing lane sometimes and back up a lot of traffic; if I were driving myself, I wouldn’t have attempted to pass in those circumstances.
Sometimes I couldn’t figure out why NoA failed to start a passing maneuver when it should have. Sometimes NoA is almost prescient, predicting that it’s going to need to pass before I detect the distant slow down, but other times NoA will come up behind a truck traveling 10 mph slower them me and just sit there. Suffice it to say that there is room for improvement.
I set NoA to signal me by vibrating the wheel when it is about to pass. That way I know to tug the wheel to make sure the car knows I’m holding it. The worst thing about Autopilot is holding the wheel, but the car not sensing it. It’s a constant distraction and annoyance. You can’t hold tension on the wheel for 8 hours a day without relaxing. My wrist was getting sore.
I’d never taken a trip this long, over 5,000 miles across the United States from Virginia to Idaho, Yellowstone, Denver and back. But it seemed like an adventure, and I figured that Tesla Autopilot would do most of the driving. So here we go …
Day 1
Day 1 starts in Palmyra, Virginia where I live. I charged to 100% at home on my Tesla Model 3 Long Range RWD. The result was a surprising 308 miles of range. I expected closer to 325. I think a couple of things, including the fact that I never charge to 100%, nor let it go low, has has not allowed the battery management system to calibrate itself. The car supposedly had 310 miles to start with and a software update should have increased it to 325.
We drove to Louisville, KY on the first of our 400-mile segments. Autopilot did a great job on the Interstate highways. All I had to do was watch. Still it was a tiring day.
The second segment was from Louisville to Columbus, Missouri, the fastest growing city in the state. We had some seriously heavy rain on the trip, so heavy I could barely make out the lines on the highway. Fortunately Autopilot didn’t have that problem. Check out this video, focusing your eyes on the place a driver typically watches to find lane lines:
Some of the rain was so heavy I could hardly even see the road, much less the lines. Autopilot drove in much heavier rain that in that video clip.
When we arrived in Columbus I made my usual check for Tesla News at Electrek, and found that the latest firmware update for the car has increased Supercharging speed. That explains why the car keeps charging before we’re finished eating or even going to the restroom. This is not like the old days in our Tesla Model S 75 when we spent lots of time waiting. I considered 75 kW good for that car, but I was peaking at 146 kW today.
So this journal of the great American electric road adventure didn’t develop as I planned because there really isn’t much adventure. We finished the trip to Idaho. It was totally straightforward. Autopilot certainly took the strain out long drives.
One thing happened after our arrival. We met family and the went on a tour of Yellowstone National Park. My car doesn’t seat 6, so a van was rented and LastGas (what I named my Model 3) was parked at a relative’s house. That’s where I made my mistake. I left Sentry Mode on and that keeps a good part of the car’s electronics running. Over several days we lost a good deal of range, perhaps 80 miles. I should have turned Sentry Mode off. But not to worry: when I realized what had happened, I just plugged in to a 120V outlet and the car adds 5 miles an hour. If I wanted Sentry Mode on, I should have plugged in. It’s recovered all the lost mileage in less than a day.
While I didn’t have the opportunity to drive my Model 3 in Yellowstone, I thought about it. We saw three Teslas in the park. I was surprised to learn that there is a Tesla Supercharger station in West Yellowstone, at the park entrance. I also saw a ChargePoint charging station in the park somewhere. The park roads were perfect for Tesla Autopilot, and Autopilot would make it much easier for the driver to take in the scenery, and to deal with the inevitable backups at animal sightings. I’m not going to drive another 5,000 miles again just to visit Yellowstone in a Tesla, but it would certainly be fun.
Or not…
I want to thank the hypothetical reader of this post for causing me to write that previous sentence, because after writing it I had a realization. I really wasn’t that far from Yellowstone, staying with family. I had a free day, so why not take Model 3 to Yellowstone right then, so I did. I drove from Swan Valley, Idaho, to West Yellowstone, Wyoming, and charged up a little while I got breakfast at McDonald’s a block away in the cold and light rain and then headed into the park. It was right before the 4th of July, and the park was pretty crowded, so I avoided the big name attractions and visited some new things, Firehole Road, the Keppler Cascade, and the Left Thumb Geyser field. I did see Old Faithful again erupt as I was driving in for lunch. Autopilot worked great on the well-lined roads of Yellowstone and when traffic bunched up, it was so much more pleasant to let the car stop and go.
And back again
Generally I had zero problems plugging in. Only Teslas were parked at Superchargers, and I think 5 of 8 was the most occupancy I found, and usually I was alone. The one exception was in Jackson, Wyoming. All 8 Tesla charging spots were occupied at one point, mine the only Tesla. To be fair, 6 of the occupied parking places are shared spots. Only the slot on the left was parked inappropriately in a reserved spot.
8 Supercharging slots. One Tesla charging and 7 gas cars parked.
I drove back on my solo trip to Yellowstone through Jackson and stopped at the same location in the middle of the afternoon with a very different result:
Jackson, Wyoming, Supercharger — all by myself
In Rawlings, Wyoming, I found what I think was vandalism. Four Superchargers were damaged so that the charging handles couldn’t be reattached to the charger–leaving them on the ground. Rawlings is adjacent to Sinclair, an oil company town.
I-70 west of Denver was an awful road, with lines completely worn off on many curves, along with uneven pavement and construction. Autopilot, which worked amazingly across Nebraska, had rough going going east through Colorado.
It was great to visit my son in Denver, who also drives a Model 3.
One odd thing happened at one of the Superchargers–the bill was $0.00 for 30 kWh. The only mishap on the trip was two big rocks thrown at the windshield resulting in two large cracks and the necessity of replacing the glass, which I did the day after we returned.
Tesla Autopilot is still described as “Beta” because the human driver has to maintain vigilance to ensure safety. I don’t usually share my Autopilot outtake videos in public, but this one is interesting in pointing out one of the challenges for AI self-driving, handling the unusual.
I don’t think I’ve ever seen anything like the piece of farm equipment that pulled in front of my Model 3 today while Autopilot was enabled. The speed was 45 mph. I wondered what would happen, and at the last minute I slammed on my brakes to prevent a collision that would surely have occurred had I not done so.
I use Autopilot a lot and I will continue to use it. I will also continue to watch the road.
My insurance carrier, State Farm, offers the “Drive Safe and Save â„¢” program that rates driving, and offers discounts for good driving. It works using a “beacon” installed in the car, essentially an accelerometer paired with your mobile phone. It tracks location, speed, acceleration, braking, cornering and whether you’re using your mobile phone.
I’ve had the Drive Safe and Save beacon in my Model S for a couple of years. When I first got their app, the only information provided was a quarterly report. It said I accelerated too rapidly and braked too abruptly–that’s how Autopilot 1.0 worked back in 2017. Now the State Farm app is much more informative, allowing you to view your latest trip. It even pinpoints on a map when you do something it doesn’t approve of.
Just recently I installed the Drive Safe and Save beacon on my Model 3 with Autopilot 2.5 hardware. Today I took a 16-mile trip involving metropolitan city streets and very challenging twisty rural highways. Autopilot drove the entire way except when I had to manually stop for stop lights and make turns. When Drive Save and Save graded my driving, it was really grading Autopilot. Here is the result:
It gave my car’s driving 5 stars. That’s better than I usually get.
I’ve enjoyed Tesla Enhanced Autopilot on my Tesla Model 3, including the new Navigate on Autopilot feature. Autopilot requires human supervision, or else something bad is likely to happen. I believe that the combination of Autopilot and a human driver is better than a human driver alone. The human managing Autopilot has more bandwidth to monitor traffic and things approaching from the side. It reduces stress and helps the driver stay fresh.
With the current Autopilot the car sets a safe following distance with the car ahead, and does so superbly with extreme reliability. It also steers the car along the road just fine on controlled access highways, and with fair results on curvy rural highways. Basically, all the driver has to do us just watch for irregularities in the traffic and to pay attention to steering on tight curves. This is easy.
I have some reservations about supervising the Full Self-Driving Capability features slated for release this year. The car will do a lot more things to watch out for. It will change lanes on controlled-access highways. It will stop for traffic signals and stop signs. It will merge and exit roundabouts. Presumably it will handle city driving. That’s a lot to watch out for. It’s one thing to make sure the car keeps doing what it is doing, and another when the car starts making turns, yielding and switching lanes. Not only does the driver have to monitor the traffic, but also what the Autopilot is doing. Stay tuned.
MIT has been studying self-driving for some time, and they have just published a paper on the “Human Side of Tesla Autopilot.”
