Well, not quite. Sphere tracking was used to correct drift, more so than the magnetometer. In principle, Move tracking is pretty much the same as PSVR tracking. Sensors for speed + camera for accurate position tracking/drift countermeasure.
Right, but the same characteristics that make the bulb easy to track also make it impossible to glean orientation information from it; it looks the same no matter which way it's pointing. Hence the magnetometer and trying to keep track of "down" as external reference points.
I'm still a bit confused, you haven't said why it's less affected than IR? Unless you're referring to TTP again? As long as you're using an appropriate camera sensor with sensible passband it will be far easier to localise a small subset of IR with known wavelength range than to track a visible light which can vary in colour with an ordinary camera sensor.
But Sony
also know their wavelength range, and better still, they get to choose it fairly arbitrarily, which allows them to choose a color that contrasts easily with the background. (Then they lower the exposure on the camera to make the lights stand out even more.) Oculus are dealing with a fixed wavelength, and believe it or not, those LEDs aren't going to be the only source in a real-world environment. That's actually why they call it Constellation; it's picking out a
known pattern of lights from amongst a fairly large assortment of random "stars."
Thinking about it, using visible light may be precisely what allows Sony to use such large, arbitrarily shaped markers. The white, rubber bulb on a Move wand has a red, green, and blue LED inside. Those shine brightly inside the bulb, causing it to glow whatever wavelength they choose, giving them a large, known target to track under a variety of conditions. I'm not sure if an IR LED can really be used to illuminate a bulb the same way, which sorta dictates the use of "pinpoint" sources. Hence, Oculus are trying to pick out the Big Dipper on a starry night, while Sony are trying to pick out the Goodyear Blimp, lit from within.
Rift doesn't have a room scale tracking solution.
Sure, but when they add the second camera, that'll put them much closer to Vive than they are now.
PSVR doesn't have any sort of physical IPD adjustment as far as we know (barring any last minute CV changes), which is probably as big an issue (bigger?) as the lower resolution panel and horsepower deficiency (in comparisons to PC VR). No one ever seems to mention it though. The BoM cost add to Rift and Vive is no doubt enormous, necessitating 2 screens to accommodate it. Certainly the quality difference must be substantial, or they wouldn't have done it.
Apparently surprisingly, it costs less to build a small screen than it does to build a large one. Yes, per square-inch. The IPD mechanism will add some cost, obviously, but even a Blu-ray drive would seem mind-bogglingly complex by comparison; this'd be like just the pop-out disc tray, minus the motor.
Thanks for posting this.
In that video you can also see what happens when the light of the move controller is occluded, it's WAY of...
It also shows that occluded tracking doesn't really get "way off" until after a decent amount of fairly violent shaking, and tracking is regained the moment the target comes back in to view. So as some of us have been saying, temporary occlusion is a relatively minor issue. Obviously, you'll want to avoid situations that promote extended periods of occlusion, but folks didn't seem to have any trouble with occluded aiming in The Deep, for example, and it's a lot easier to lose the DS4 than a wand. The headset has LEDs all around, so you'd need to occlude it fairly purposefully.
