java-player-users

[Java-player-users] JavaClient patches: Performance, Gripper

[Sebastian R. <seb...@go...>]

Hello Jorge,

I made some changes:

[0]
Does a huge performance improvement when you have a Player server containing a lot of devices.
E.g. Having a Stage simulation with 100 robots (each has at least one position and maybe a ranger device).
 
Formerly the PlayerClient just loops through all device for logging output only. Which takes on my config around 20 seconds (!) for connecting 1 PlayerClient (and minutes for 100 clients).
With the patch the standard behaviour is not changed, i.e. when logging level == INFO.
But when you set it to any higher value, e.g. SEVERE, it will benefit from the performance advantage. It will than take only about a second for one client.

[1]
Adds the gripper commands to more convenient methods, e.g. open, close ...

Kind regards,
Sebastian


[0]

[Java-player-users] Fwd: JavaClient patches: Performance, Gripper [Update]

[Sebastian R. <seb...@go...>]

I just found a bug when the logger is 'null'.

The patch is attached.

Sebastian

Anfang der weitergeleiteten E-Mail:

> Von: Sebastian Rockel <seb...@gm...>
> 
> Hello Jorge,
> 
> I made some changes:
> 
> [0]
> Does a huge performance improvement when you have a Player server containing a lot of devices.
> E.g. Having a Stage simulation with 100 robots (each has at least one position and maybe a ranger device).
> 
> Formerly the PlayerClient just loops through all device for logging output only. Which takes on my config around 20 seconds (!) for connecting 1 PlayerClient (and minutes for 100 clients).
> With the patch the standard behaviour is not changed, i.e. when logging level == INFO.
> But when you set it to any higher value, e.g. SEVERE, it will benefit from the performance advantage. It will than take only about a second for one client.
> 
> [1]
> Adds the gripper commands to more convenient methods, e.g. open, close ...
> 
> Kind regards,
> Sebastian
> 
> 
> [0]

Re: [Java-player-users] JavaClient patches: Performance, Gripper [Update]

[Sebastian R. <seb...@go...>]

Hi Jorge,

Am 09.03.2011 um 20:05 schrieb Jorge Santos Simón:

> Great! Thank you a lot.
> 
> [0] commited

Cool and sorry for the patch right after the patch;)

> 
> [1] sorry for the duplicated effort: I have recently commited a similar patch

Not too bad, it wasn't that much.

> By the way, concerning your issues with PlannerInterface, I have
> failed to give you any insights because....I'm not able even to run
> properly player with wavefront!!!

I found that for the position hypothesis using the localize instead of the planner IF is faster.
So I use the planner IF only for giving it goals.

>  It seems to work, however, with the
> real robot, but on Stage it NEVER finds a path, I guest because the
> origin of the map are not well configured according to the robot
> position. As soon as I make it work, I'll try with javaclient.

I have the JavaClient working with the Wavefront planner, both in real world and Stage simulation.
There is no problem currently.

If you describe your configuration in more detail I can help maybe.

Sebastian

> 
> Best regards,
> Jorge