Dear all,
I would like to investigate how I might simulate a tracked vehicle/robot with Bullet. To clarify, by tracked vehicle I mean:
- Double tracks, like a tank or bulldozer type vehicle
-- Qinetiq Talon
-- Mesa Robotics Matilda
-- Jacobs Rugbot
- Single tracks, like some snow mobiles
- Multiple, movable tracks, like some important robot designs:
-- iRobot Packbot
-- IRS Kenaf (Japan)
- (maybe) Variable tracks:
-- Inuktun VGTV (changing geometry)
In general, I would like these track models to be defined by a (variable) non-intersecting 2D polygon. This way, more complex shapes are possible as well. Shapes (lines and arcs, piecewise) around these polygons, taking into account wheel radii, are easily found.
Another important feature would be shape profiles of the track elements. Some might include additional non-uniformities (such as the ridges seen on the Matilda robot above) to increase friction on uneven terrain.
Note that none of the robots above have suspensions, so I'm going to neglect it for now, including pseudo-suspension from deforming the tracks. If anyone has an idea how to include suspension in the simulation, I'd be very happy to hear it.
After some googling, I found the ADAMS Tracked Vehicle toolkit, which is a commercial simulation package used to design tanks, etc. Some nice videos of what is my long-term goal (as long as it is real-time) can be seen here:
http://www.youtube.com/watch?v=-c-o1jC7AKg
http://www.youtube.com/watch?v=ac3rnUu-TgA
The only mention of a tracked vehicle simulation on this forum is this:
http://www.bulletphysics.org/Bullet/php ... f=12&t=652
However, modeling the tracks as a series of wheels is not going to cut it for me.
My first idea was to simulate these tracks independently by sampling the track with many cylinders or capsules and making them move around the track shape with some user-defined speed. Various track elements can be modeled similarly.
However, does this work when done relative to a moving rigid body? This thread
http://www.bulletphysics.org/Bullet/php ... f=9&t=3555
mentions a problem with simulation artifacts when doing the same with a conveyor belt. Will the friction orthogonal to the drive direction cause jitter and instabilities during steering? Also, what setup of joints, etc, can I use to "manually" force the cylinders along the path of the track?
This diagram shows what I mean:
http://www.gliffy.com/publish/1885182/
Do you think this is a good idea? I would like to achieve a good balance between fidelity and efficiency.
Thank you all very much for your replies!
Max Pfingsthorn
PhD Student
Robotics Group, Jacobs University Bremen
http://robotics.jacobs-university.de