Re: simulated arm/gripper
Andrei Androsoff <[email protected]> Mon, 25 Jul 2005 13:37:18 -0400
| Newsgroups | gmane.comp.breve |
|---|---|
| Message-ID | <[email protected]> |
Tim, all, I've been trying to build an articulated arm too and have been running into a lot of problems with the physics engine. It seems that if I add too many links to my MultiBody, the arm tends to fly apart or fly away (!) when the simulation starts (often preceded by a painful-looking squirming and wiggling). If I tweak the physical dimensions of the Joints or Shapes it often removes the problem, until I move a particular joint too far, and then the machine flies apart. Sometimes the coordinate system of child objects seems to get messed up as well - moving a parent object will cause the child to rotate (e.g. flexing your elbow will cause your wrist to twist). Again if you make the joint a little bigger or longer, keeping the normals the same, the problem will go away or get worse). I don't know of any more genteel way to put it other than the physics engine is "freaking out". I have also built walking robots and experienced similar problems. The behaviour is sometimes determinate (same thing happens every time you start the sim) but sometimes it actually varies between runs (no random numbers in the code of course). Tim, since you say your arm works, did you run into anything like this? Any pointers? Does anybody else experience these things? By the way, I'm not enabling self-collisions and I'm not using set- relative-rotation for any of the joints. All my joints are instances of a subclass of RevoluteJoint. I know this screams for a code example but I want to clean it up a bit and see if there is any obvious fix in the meantime. Andrei _______________________________________________ breve mailing list [email protected] http://lists.spiderland.org/mailman/listinfo/breve