Ynot, two things:
A: There IS NO "pit" at the roller on our treadmill. I'll happily demonstrate this in my next video by placing the level at the rear (with the cart in place) and show there is no light under the level at that point. The belt is tight, spans less than an inch of open space and the rear wheel of the cart weighs barely over one ounce. The pit only exists in your imagination. (if the cart was heavier -- say our "truck" design -- it could create such a pit)
B: There IS NO BUMP across the belt created by the boards. The boards do not extend more than a half inch under the edges of the belt. They only turn up the edges a bit to create a fence effect and keep the cart from falling off. I can also demonstrate this for the next video by placing the level on the belt and show no light under it. The bump only exists in your imagination.
To run a device like this on a treadmill *unattended and unattached* is quite a challenge. You have to manage the side to side issue (thus the fence), and you have to get the balance right so it won't fall off the back or climb off the front.
Here's the reason that in the *unattended* 1:48 long video, the cart operates where it does on the treadmill: the propellor is less efficient when operating close to the belt -- simply because of gradient issues. This means that the device can operate at a lower speed with the prop hanging off the back of the belt then it can in the middle. We used this behavior simply to keep the cart from climbing off the top of the treadmill -- the prop generates it's best thrust hanging off and advances the cart slowly ... once the prop reaches the belt it loses thrust by a hair and the momentum carries the cart up just a bit and then with the lessened thrust it begins to back up until the prop bites again off the end. These cycles eventually end up big enough that the cart climbs far enough up the belt and falls back faster than the prop can recover from before the rear wheel falls off the belt.
You see, it's really easy for people to critical of the testing, but until you try to please everyone you really have no clue what it's like.
You want strings hanging off the device -- most see strings and say "WTF -- those strings are part of the hoax". Most want see the device run hands off -- you say "what the heck did you put boards along the edges for".
Get one working of your own and then you can have your very own 500 post thread where people say "your tests are bogus -- here, perform my bogus test".
JB