And what would you expect the service center to do with the software? They might uninstall and reinstall it but this is clearly not a hardware problem but a situational awareness and evaluation problem. I think we have to wait until V11 and hope that this is forefront on their 'radar' (sorry) and that it gets fixed. Sadly, Tesla has a lot on their plate with phantom braking, with the removal of sensors and not getting Tesla Vision to work for close objects, with summon and reverse summon, and with a host of issues with FSD. I feel a bit sorry for them, and for all of us having to wait around for the software to get good enough to trust with our lives and those important (and less important) to us.
On my two recent trips where I was using NoAP I am quite certain that the camera was working because I kept my hands on the wheel and responded to the tug requests when made so the only thing I can guess why it suddenly disengaged is that it didn't like what I was doing with my eyes. On the first trip I was actually looking around because it was doing such a good job at driving, and on the second trip I suspect it thought I was drowsy and that might well have been true but I was still alert. Either way the camera would have been necessary to figure it out.
I do not think you have a software problem. I have never, ever had any sort of hard braking and I use the FSD all the time. I think you have a hardware problem that is creating a false positive. I am running this same software build on the same vehicle and the problem just does not exist.
In an optimal world, they'd find out if there was something wrong with their computer or vision that was making it stutter.
I am using FSD Beta 10.69.25.2. I haven’t done anything to the vehicle since buying it except to wash it regularly. I admit that I do not watch the screen waiting for the various indicators but I do respond to the blue screen. If this is a hardware problem then I will ask Tesla to consider it Thursday when they stop by.
I couldn't agree more, and since the path planner is [currently] based on rules they should be able to tell which rule [set] fired even if they cannot get the exact data from the scene analysis NN. They will not respond to individual incidents but I can assure you that I reported all of it to [email protected] as soon as we got back.
Back with my trip report. It is a bit long but will share my experience from today's drive.
Thank you for the great information. I should make something clear about my posts and that is that my car was most likely phantom decelerating and using regen but that is how I brake 99% of the time so I called it phantom braking. In my case it was dramatic enough that items flew and I disengaged , so mine might have done what yours did. Clearly you have a stronger stomach for this than I do. Also, there might have been cases where it thought it saw emergency lights as there were often flashing lights for snow conditions but sometimes there were none.
I really try to let the vehicle make its own corrections and I only intervene when I think its actions will result in an unsafe condition. (Well, except for that slow passing acceleration thing... that totally irks me.)
Agreed. As an Ontologist, I was working on integrating deep knowledge models and ML to support “Explainable AI”, and then I retired, so I know that people are working on it but not whether Tesla is. From what I gather Tesla is not keen on deep knowledge models so we should probably not do more than hope for this kind of explanation.
