Finally, here it is : The demo of my proof-of-concept for realizing tracked vehicles (tanks) using only Newton’s material callback system. Head over here to download it (~3,6 MBytes) and please refer to the readme.txt before playing around with it. Also note that it’s possible that not all gamepads will work, if that’s the case just use the keyboard controls.
I have uploaded a new video (wmv-format, ~60 MBytes, ~5:40 mins) of the tracked tank vehicle implemented in newton. This video shows a more advanced, and almost finished version that allows you to control each track separate (using two analog sticks on my gamepad), as well as rotating the turret, which is implemented with a derived custom joint (and controlled using two neck buttons on the pad). It also allows for realistic turning when the tank is moving, something that wasn’t working in the version that the first video showed. For this I “just” had to check if both tracks accelerate in the same direction and then if the delta is too high I bring the other track up to speed. Works fine so far and now the tank really moves like a real tank. Once again, if you want to know more about how I implemented this and if you want to participate in the discussion about this technique, head over to this thread on the newton forums.
Update : Julio Jerez, the man behind Newton Game dynamics, has it’s own channel on youtube showing off stuff made with Newton. He already uploaded the above mentioned video there, you can access it here. But I suggest you download the video, as youtube has a rather low resolution combined with a horrible qualit.
Back in 2006 I tried to simulate a tracked tank in Newton using a very uncommon method. Instead of moving the tank around with forces I tried to implement an “inverted” conveyor belt material for the tracks, to achieve a more realistic movement. Sadly it didn’t work out 100% perfect back then, but recently I revisited my demo and I now got something that almost perfectly moves like a tank.
A short summary : Instead of applying a force that “drags” the tank around the terrain (not very realistic), or using a lot of wheels to mimic a track (more realistic, but very slow) I’m using Newton’s contact process callback, where I go through all contacts that the track is generating against the terrain and then accelerate along the contact’s tangent. Together with my new tank model (you may remember it, I made it for “Phase 2” of Projekt “W”) that has a much better track (more flat, which means more even contact generation), a joint to make physics for the tracks more dynamically follow the terrain (see end of the video for more on this) and the new beta of Newton 2.00 I now have an almost perfect way of simulating real tank track behaviour. Not only is it realistic, but contrary to the “multiple wheels for a track”-way it’s very fast and you could do dozens of tanks at once without a severe speed drop.
If you want to see it in action, check out this video I created (~3 min, 35 MBytes, WMV format). I also have a demo ready, but the man behind Newton Game Dynamics wants to optimize contacts generation first before I release the demo to the public. And if you’re interested in implementing something like this yourself, head over to this thread I created at the Newton forums where I also tell how I implemented this.