Question about SDF lib plugin and making SDF in general #2182
-
IntroHi! I am a graduate student at SDU, I use MuJoCo for my research on Contact manipulation. I found few ways to do it.
First one may be the best option to have good performance but required too many work. Third one works good without having bunch of objs but i couldn't really find the documents about SDF lib plugin. My setupMuJoCo 3.2.0 My questionIs third method just same plugin as first one but default? What are the differences between the methods mentioned above? Minimal model and/or code that explain my questionNo response Confirmations
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
Hi, the collision algorithm for methods 1 and 3 will be the same. The difference is that 1 requires writing a new plugin, while 3 means using the already existing sdflib plugin (like in this example https://github.com/google-deepmind/mujoco/blob/main/model/plugin/sdf/cow.xml) - the underlying representation of the SDF is a voxel grid computed from the mesh using this library https://github.com/UPC-ViRVIG/SdfLib Using 3 is likely not as fast as using 1 but it's obviously more general because it's really hard to come up with an analytic SDF for complicated objects, so computing it automatically using a voxel grid is a big advantage. |
Beta Was this translation helpful? Give feedback.
Hi, the collision algorithm for methods 1 and 3 will be the same. The difference is that 1 requires writing a new plugin, while 3 means using the already existing sdflib plugin (like in this example https://github.com/google-deepmind/mujoco/blob/main/model/plugin/sdf/cow.xml) - the underlying representation of the SDF is a voxel grid computed from the mesh using this library https://github.com/UPC-ViRVIG/SdfLib
Using 3 is likely not as fast as using 1 but it's obviously more general because it's really hard to come up with an analytic SDF for complicated objects, so computing it automatically using a voxel grid is a big advantage.