Raúl
Moderator
 Moderator
| Posts: 590 |  | Karma: 10
|
Re:Multi-Robot SLAM - 2010/03/02 17:36
Hi,
What I've noticed when doing processing other than artificial vision is that most of the CPU load was due to the simulator (the Microsoft Visual Simulation Environment).
While I don't see now an obvious way to optimize the code in your case, these are the things that come to my mind:
- Using an Ageia PhisX physics hardware accelerator (that should relieve the PC CPU from the physics simulation load).
- Move some services to another DSS node (running in a different computer). Here, there is always a tradeoff between CPU load in the DSS nodes and communication bandwidth required across the nodes.
None of these solutions are any good (in terms of code optimization) as they just consist on adding more hardware.
Another optimization you could try is to decrease the required load by the SLAM algorithm by:
- Considering a lower map resolution. - Considering a lower LRF angular resolution.
These would be equivalent to lower the resolution of images in vision processing...
If you prefer Spanish, please use the Spanish forum for RDS:
-> Foro de Microsoft Robotics Developer Studio
(I try to keep different languages separated in different forum threads so people don't get confused).
Hope this helps..
Raúl Arrabales Moreno. conscious-robots.com/raul |