There are several difficulties of speeding up a simulation (not to say that it is impossible, just can have strange side affects).

The first is the raw computing power, there is a limit to how fast the simulation can run as it still takes the same amount of CPU calculation to update each step.

Another issue is time synchronisation, if your client (i.e. wavefront) uses wallclock time for anything it will think less time has passed than has for the simulation, which will cause issues in the timing of its responses. (i.e. it is as if the robot is driving 100 times faster than you told it to)

Final one I can think of at the moment is whether the client is able to process all of the incoming data fast enough, if not it will either be discarding a lot of data or using old data depending on its design. I suspect this is the case you are seeing, although possibly combined with the other two.


2009/8/5 Ruben Gerlach <>

in order to increase the learning speeds of my reinforcement learning algorithms I want the simulation to be as fast as possible. I changed the interval_real variable in the stage world file from 100 to 1. The simulation now runs faster but I also experience the agents behave worse. Using the wavefront planners the agents move often in circles instead of finding the straight way. What is the reason for this behavior and is there some other way to increase the simulation speed? Maybe I would have to compile everything into one program instead of using network communication?

Neu: WEB.DE Doppel-FLAT mit Internet-Flatrate + Telefon-Flatrate
für nur 19,99 Euro/mtl.!*

Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.
Playerstage-users mailing list