Some weeks ago I posted a poll on how often people experience phantom braking, and the results were around 50% for "Almost never" and the rest spread out over "seldom" to "almost always". That's a really high number so I decided to test out different strategies to improve it (I was one of the "almost always" voters). This is the results from my testing.
Car: October build Model 3 Performance. Autopilot, no FSD. No vinyl, no PPF, no wax, no coating. 2020.12.6
Road: 50km of high speed (100-110km/h), 15km of medium speed (60-80km/h) two lane highway each way and physical separator between driving directions (by the book). In total 120km+ round trip every time on AP. Plenty of ramps, a lot of trailers, plenty of other cars.
Timing: Always middle of the day, no traffic going out, heavy traffic going back.
Test method
Always activate Autopilot, and keep on as much as possible. If not possible, use TACC (but this never occurred). Switch lanes responsibly, use several seconds or wait for the next chance. Abort lane change over speeding up.
Base line. Dirty car, collision detection set to early. 2x sunny days:
Phantom braking happens a lot. Nearly every trailer I pass causes braking, all overhead passes causes emergency braking, and switching lanes sometimes cause a minor brake situation. It's quite untolerable and I felt awful driving like this not only on round-trip, but two. It was not something I would be OK living with.
Second test. Clean car, collision detection set to early, 1 sunny day, 1 cloudy day:
It seems like phantom braking happens just as often. I had washed the car the day before, so I pulled over mid-trip to go over the radars and the cameras with a clean microfiber cloth. It did not make a difference, and I felt like a real danger in traffic just waiting for the car to suddenly brake again. I washed the car again (machine wash, shame on me) the next week and tried again on a cloudy day, but it was still as bad. Direct sunlight vs cloudy didn't seem to have any difference in the phantom braking effect.
Third test. Clean car, collision detection set to late, sunny day:
I didn't experience any phantom braking, but the follow distance seemed WAY off in this setting. I came up really close to cars in front of me before the car decided to start braking. It felt extremely aggressive, and I was worried people would get mad at me. I was really not comfortable on my way home again in heavy traffic. I braked manually many times because I didn't like how close I got before it did the job itself. Setting follow distance to 7 helped, but this left a gap that even trailers utilized so it wasn't really my cup of tea.
Fourth test. Dirty car, collision detection set to medium. Rainy day:
No phantom braking experienced for the whole trip. Follow distance when cars braked in front of me was uncomfortable, but I adjusted the follow distance up from 3 (what I used with early collision detection) to 5. This was quite an enjoyable ride, and let me drive all the way without any incidents.
Fifth test. Dirty car, collision detection set to late. Cloudy day:
Pretty much the same as the first time I set to late. I switched back to medium on my way home because I didn't like this setting, and it was stressful in stop-and-go traffic.
Sixth test. REALLY dirty car, collision detection set to medium. Rainy day:
Same as the previous time I set it to medium, except I had a valid emergency brake due to a motorcycle acting like a lunatic. It came pretty close but nobody was hit. 5 seems like the sweet point on follow distance for me. Set to 1 it works pretty well in stop-and-go queues.
Seventh test. Dirty car, collision detection set to medium. Really bright, sunny day:
No difference from other tests on medium. This is my go-to setting right now.
Wrap up
On my car it seems that the only setting which really matters is the collision detection settings. Dirty or clean made no noticable difference, and sharp, direct sunlight and rain had no effect on AP performance.
I would like to note that this is my second TM3, and the first one also had awful phantom braking. I did not, however, perform these tests with that car.
Now I can finally install 2020.16
Car: October build Model 3 Performance. Autopilot, no FSD. No vinyl, no PPF, no wax, no coating. 2020.12.6
Road: 50km of high speed (100-110km/h), 15km of medium speed (60-80km/h) two lane highway each way and physical separator between driving directions (by the book). In total 120km+ round trip every time on AP. Plenty of ramps, a lot of trailers, plenty of other cars.
Timing: Always middle of the day, no traffic going out, heavy traffic going back.
Test method
Always activate Autopilot, and keep on as much as possible. If not possible, use TACC (but this never occurred). Switch lanes responsibly, use several seconds or wait for the next chance. Abort lane change over speeding up.
Base line. Dirty car, collision detection set to early. 2x sunny days:
Phantom braking happens a lot. Nearly every trailer I pass causes braking, all overhead passes causes emergency braking, and switching lanes sometimes cause a minor brake situation. It's quite untolerable and I felt awful driving like this not only on round-trip, but two. It was not something I would be OK living with.
Second test. Clean car, collision detection set to early, 1 sunny day, 1 cloudy day:
It seems like phantom braking happens just as often. I had washed the car the day before, so I pulled over mid-trip to go over the radars and the cameras with a clean microfiber cloth. It did not make a difference, and I felt like a real danger in traffic just waiting for the car to suddenly brake again. I washed the car again (machine wash, shame on me) the next week and tried again on a cloudy day, but it was still as bad. Direct sunlight vs cloudy didn't seem to have any difference in the phantom braking effect.
Third test. Clean car, collision detection set to late, sunny day:
I didn't experience any phantom braking, but the follow distance seemed WAY off in this setting. I came up really close to cars in front of me before the car decided to start braking. It felt extremely aggressive, and I was worried people would get mad at me. I was really not comfortable on my way home again in heavy traffic. I braked manually many times because I didn't like how close I got before it did the job itself. Setting follow distance to 7 helped, but this left a gap that even trailers utilized so it wasn't really my cup of tea.
Fourth test. Dirty car, collision detection set to medium. Rainy day:
No phantom braking experienced for the whole trip. Follow distance when cars braked in front of me was uncomfortable, but I adjusted the follow distance up from 3 (what I used with early collision detection) to 5. This was quite an enjoyable ride, and let me drive all the way without any incidents.
Fifth test. Dirty car, collision detection set to late. Cloudy day:
Pretty much the same as the first time I set to late. I switched back to medium on my way home because I didn't like this setting, and it was stressful in stop-and-go traffic.
Sixth test. REALLY dirty car, collision detection set to medium. Rainy day:
Same as the previous time I set it to medium, except I had a valid emergency brake due to a motorcycle acting like a lunatic. It came pretty close but nobody was hit. 5 seems like the sweet point on follow distance for me. Set to 1 it works pretty well in stop-and-go queues.
Seventh test. Dirty car, collision detection set to medium. Really bright, sunny day:
No difference from other tests on medium. This is my go-to setting right now.
Wrap up
On my car it seems that the only setting which really matters is the collision detection settings. Dirty or clean made no noticable difference, and sharp, direct sunlight and rain had no effect on AP performance.
I would like to note that this is my second TM3, and the first one also had awful phantom braking. I did not, however, perform these tests with that car.
Now I can finally install 2020.16