Oh how I wish the display was captured by TeslaCam, along with the camera feeds. But it isn't, so I will attempt to describe the events that accompanied this video.
As I was approaching this crossing, I saw the lights come on and the gates drop. Oh good, I said, this will be an amusing test of TACC (which was already engaged). We got a little closer (about where the video starts) and the display showed solid red stop lights and a red limit line, along with the "stopping at traffic control" message. So far so good.
As I just mentioned, this is the point where the video starts, so you can follow along. The car slowed, well in front of the limit line, and started to creep up on it. So far very normal. But at this point, I notice the limit line is flickering between red and green! Hey now, I say, probably out loud. This is interesting.
If you notice in the video, there are 4 total "stops" as we creep up to the crossing arm. By "stop" I do not mean we came to a full and complete stop, but I do mean that the car applied the brakes, came to a near stop, considered the situation, and continued. The first three "stops" were done by the car, and the subsequent resuming was also done by the car. The last stop, which is almost not noticeable due to being right at the end of the clip, was me stomping on the brake, saying "Nope, experiment is over."
Again, as we approached the arm and lights, the limit line was red. As we got close to it, it started to flicker between red and green. By the second and third "stops," the limit line was solid green. I also noticed at this point that the stop light images on the display had stopped looking red, and had gone all grey. The train itself was rendered as diagonally crossing vehicular traffic. Looking up at the lights, I could see why... we were so adjacent to them that the light lenses weren't even visible any more.
I wish I'd had the nerve to let it go, for investigative purposes, but that arm was getting awfully close (as was the train). I will say that I was not at all convinced that the car would not have continued to inch itself right into the side of that freight train.
I think this is a side effect of the new version's feature that allows it to follow traffic through signs/lights without operator confirmation. Once we were far enough forward that it could not "see" the lit part of the light anymore, it decided that "not red" meant "OK to go", and the crossing train looked enough like traffic ahead that the "follow traffic" feature was also giving approval.
In general, due to this and other experiences at crossings, I would say it is time for the AI team to step up the railroad identification game. These are not the places to be making mistakes.
As I was approaching this crossing, I saw the lights come on and the gates drop. Oh good, I said, this will be an amusing test of TACC (which was already engaged). We got a little closer (about where the video starts) and the display showed solid red stop lights and a red limit line, along with the "stopping at traffic control" message. So far so good.
As I just mentioned, this is the point where the video starts, so you can follow along. The car slowed, well in front of the limit line, and started to creep up on it. So far very normal. But at this point, I notice the limit line is flickering between red and green! Hey now, I say, probably out loud. This is interesting.
If you notice in the video, there are 4 total "stops" as we creep up to the crossing arm. By "stop" I do not mean we came to a full and complete stop, but I do mean that the car applied the brakes, came to a near stop, considered the situation, and continued. The first three "stops" were done by the car, and the subsequent resuming was also done by the car. The last stop, which is almost not noticeable due to being right at the end of the clip, was me stomping on the brake, saying "Nope, experiment is over."
Again, as we approached the arm and lights, the limit line was red. As we got close to it, it started to flicker between red and green. By the second and third "stops," the limit line was solid green. I also noticed at this point that the stop light images on the display had stopped looking red, and had gone all grey. The train itself was rendered as diagonally crossing vehicular traffic. Looking up at the lights, I could see why... we were so adjacent to them that the light lenses weren't even visible any more.
I wish I'd had the nerve to let it go, for investigative purposes, but that arm was getting awfully close (as was the train). I will say that I was not at all convinced that the car would not have continued to inch itself right into the side of that freight train.
I think this is a side effect of the new version's feature that allows it to follow traffic through signs/lights without operator confirmation. Once we were far enough forward that it could not "see" the lit part of the light anymore, it decided that "not red" meant "OK to go", and the crossing train looked enough like traffic ahead that the "follow traffic" feature was also giving approval.
In general, due to this and other experiences at crossings, I would say it is time for the AI team to step up the railroad identification game. These are not the places to be making mistakes.
