Menu
▾
▴
Re: [simspark-devel] [Sserver-three-d] Feature discussions
|
From: Yuan Xu <xu...@in...> - 2012-09-13 09:16:44
|
Hi all,
I have put the implementation of realistic motor in the svn branch
'realistic_motor ',
you can check out and test the code by:
svn co https://simspark.svn.sourceforge.net/svnroot/simspark/branches/realistic_motor
simspark_realistic_motor
or simply review the code in
http://simspark.svn.sourceforge.net/viewvc/simspark/branches/realistic_motor/
The main changes are:
* stiffness of motor is controllable by StiffnessEffector (the name of
effectors are name of joint effector + 's', e.g. 'laj1s' for
controlling stiffness of 'laj1')
* the power consumption is modeled, the battery status is given by
BatteryPerceptor, the robot is not able to move when the battery is
empty
* the internal temperature of motor is modeled, and the overheating
protection is enabled when the motor is too hot.
* the temperature of motor is reported by AngularMotorPerceptor
The new effector and perceptors are documented in user manual as well.
All parameters are determined by experiments on real NAO, and we can
change them for competition.
Any comments and feedback is welcome!
Best Regards,
Xu, Yuan
On Thu, Sep 6, 2012 at 2:13 PM, Hans-Dieter Burkhard
<hd...@in...> wrote:
> Hello all,
>
>
> Am 05.09.12 17:37, schrieb Stefan Glaser:
>
> Hello Dieter,
>
> Am 05.09.2012 09:59, schrieb Hans-Dieter Burkhard:
>
> Dear Stefan, dear all
>
> thank you Stefan for your explanations and plans - I am looking forward to
> hear more
> (by the way, in the community I am simply "Dieter").
>
> you are very welcome.
> For turning off the soccer rules, simply comment out the line:
>
> gameControlServer.initControlAspect('SoccerRuleAspect')
>
> in the "naosoccersim.rb" file in "/usr/local/share/rcssserver3d" folder.
>
>
> Thank you, I did and it works in principle well
> (but it still does not allow to beam players at arbitrary position at
> beginning).
>
>
>
> As a repeated statement, I think that the simulator has a great potential
> for both, teaching and research.
> If we can manage to lower down the barriers for entrance, it could be of use
> for many purposes, not only
> RoboCup - but maybe to make people interested in RoboCup. We had just a
> course in Macedonia on
> "Robotics and Mathematics" (funded by DAAD), where we used the simulator
> together with an agent
> framework. It worked really well and we are going to extend the framework
> etc.
>
> For such work (and also for intended usage at high schools), a simple
> handling is recommendable.
> The usage for Machine Learning etc. leads to similar recommendations - if we
> dont think only about RoboCup
> people.
>
> How this is achieved best can be discussed (for the moment I would like some
> simple "switches", but it is also
> possible to distribute the programs with predefined configurations).
>
> That's great! I think the community is happy to hear that the server is
> further applied beyond the RoboCup competitions. We also discussed at the
> previous competitions that we want to simplify development with the server
> and make it easier to use for researchers. In order to do so, I decided to
> address the documentation gap first, since I know that the server already
> provides most/all of the functionality needed. After that I will start a
> discussion about improvements.
>
>
> Actually, the installation of the server is not so easy. So we used a
> windows-version compiled by and thanks to Yuan.
> For usage outside RoboCup we might need a "stable version" - while for
> RoboCup we want to extend it step by step.
> Dont know how to overcome this conflict. We have compiled a tutorial
> material about the server. We could provide
> it to the community (but it is not up to date because it was related to our
> server version from Spring, but needed
> changes should be not too much).
>
> At the other side, we will provide some easy understandable agent, which can
> be modified and extended without
> much efforts. Actually, the first version (written in JAVA with credits to
> MAGMA) is already under test. We will report
> about further progress and hope to present it at RoboCup in Eindhoven.
>
> Best regards
>
> Dieter
>
>
>
> According to low level control and close relations to reality, I am also
> looking forward to introduce stiffness,
> torque, power consumption etc. if possible - may be not for competition but
> for some "research modes".
> As I understood, something is already done in this direction? Can it be made
> available?
>
> I was also advertising the torque control topic over the previous
> competitions with moderate success. By looking into the server files I
> somewhere found that one can turn on torque feedback to joints. According to
> Sander, there were some problems with a crashing server when using this
> feedback, which should be fixed by now. So perceiving the torque may be
> possible, but as usual it is not documented how. Furthermore I also had a
> discussion with Hedayat and he meant that ode suggests controlling motors by
> speeds. But he also suggested to bring up a discussion on the mailing list
> about this, because he also was not completely sure about it.
> Since I'm not a developer of the server, I would suggest to write to the
> simspark mailing list for more information about this topic.
>
> Best regards,
> Stefan
>
> Best regards
>
> Dieter
>
>
>
>
>
>
>
>
>
> Am 04.09.12 13:21, schrieb Stefan Glaser:
>
> Hello Mr. Burkhard,
>
> nice to hear, that you took advantage in using our code. I'm always eager to
> get some feedback to our framework.
>
> As Klaus and Yuan pointed out, you can use the trainer/monitor protocol to
> control the players, the ball and the playmodes of the server. Since it is
> the monitor protocol, you should be able to get any position and orientation
> to any object on the server by parsing incoming messages to the trainer. If
> you just need the own position, "setSenseMyPos = true" as Yuan described
> could also be an option. Furthermore there is a possibility to turn off
> specific soccer rules. Unfortunately I can't remember where, yet.
>
> In my opinion an "exercise mode" doesn't make much sense. I don't like
> mixing up the agent code with the learning code. The
> trainer/monitor-protocol in combination with a special server configuration
> provides a very generic and independent interface for learning. Different
> learning scenarios need different setups. E.g. for walk-learning I may want
> to have a bigger field, to connect multiple agents in parallel, while for
> learning how to kick, I need the ball.
> I'm planing to put a list of all possible configuration parameters of the
> server and their meaning, as well as how to use the server in learning
> scenarios into the simspark-wiki. The time slot for this starts end of this
> month. In combination I may also try to improve configuration possibilities
> of the server with respect to learning. You will find news about this in the
> mailing list.
>
> Adding stiffness control would be very nice to have.
> I'm still looking forward to just have a torque control. Setting intended
> forces instead of target angular speeds. But I'm not sure if there is still
> development towards force-control.
>
> Wish you all the best,
> Stefan
>
>
> Am 04.09.2012 10:31, schrieb Yuan Xu:
>
> Dear Dieter and all,
>
> On Tue, Sep 4, 2012 at 10:08 AM, Hans-Dieter Burkhard <
> hd...@in...> wrote:
>
> Dear Klaus,
>
> thanks for the answers and the code - we will try it out! Hopefully you
> will
> hear still more from me :-)
>
>
> Thanks for your code, but I think 'excise mode' in server will help all
> other teams ( which do not use java), and it would be easy to use and
> functional out box.
>
>
> GPS even during work would make evaluation (e.g. fitness) easier. Maybe
> someone knows
> more about status?
>
> The vision perceptor can provide the global position, (setSenseMyPos in
> rsg/agent/nao/naoneckhead.rsg) I think it works similar as GPS, but it
> doesn't provide the direction of robot, which is also useful for evaluation.
>
>
> I cannot estimate the efforts for stiffness. It would bring simulation
> closer to reality at the lower
> levels (body control). Another approach in this direction would be
> measurement of energy consumption
> by the individual joints (how much force is needed). But may be that is
> even more complicated to simulate.
>
> We can simulate the stiffness by setting the maximum force of joint (this
> is easy in ODE), because the maximum currency of motor is determined by
> stiffness in real NAO, and for DC motor, output torque t = K * I, where I
> is currency and K is torque constant of the motor, it can be found in the
> specification from manufacture.
>
>
>
> Battery simulation is nice as well for the higher control aspects (as we
> have it in 2D).
> Actually, for real robots the energy consumption depends not only on the
> amount of movements in general,
> but on clever designed energy saving motions (which may lead back to
> measurement of energy consumption at the joints).
>
>
> Actually I have already implemented stiffness and energy consumption of
> motors for my thesis. If you are interested in, I can make a branch for
> testing.
>
>
>
> Thank you once again and greetings from Croatia,
>
>
> Wish you nice days!
>
>
> Dieter
>
>
>
>
>
>
> Am 03.09.12 15:28, schrieb Klaus Dorer:
>
> Dear Hans-Dieter,
>
> good to hear from you again!
>
> Some of the features you are asking for are already achievable with the
> trainer(monitor) protocol. We have implemented it this season and use it
> for machine learning purposes. I have attached the Java class doing the
> stuff. I think it should work with the ServerConnection class we published
> in our last source code release. A little usage example is attached below.
>
> With that you have access to what a referee can do through the monitor.
> I.e.
> - placing a player at any position at any time
> - placing the ball to any position at any time
> - dropping the ball to start a game (make sure to first drop ball and
> then beam a player close to it)
> - change playmode
>
> With respect to the "GPS" I weakly remember that it can be configured in
> the server to be added to player information in training mode. May be
> Stefan knows where exactly? At least after trainer beaming you know exactly
> where a player is.
>
> With respect to stiffness I am a bit uncertain. It sounds to me like
> being quite computational expensive although I really have no good insight
> into server simulation. When it comes to features that bring simulation
> closer to reality I would prefer things like energy simulation that also
> add value with respect to strategy.
>
> Greetings
> Klaus
>
> protected boolean tryToConnect()
> {
> tce = new TrainerCommandExecutor(**IServerConnection.SERVER_IP);
> if (!tce.isConnected()) {
> return false;
> }
> return super.tryToConnect();
> }
>
> tce.beamBall(0, 0);
> tce.dropBall();
> tce.movePlayer(Team.LEFT, 2, x, y, 0.65f);
>
>
>
>
>
> Am 03.09.2012 14:07, schrieb Hans-Dieter Burkhard:
>
> Dear Klaus and Yuan,
>
> we at Humboldt University are interested to use the server for teaching
> at different levels
> of complexity, including Machine Learning. I have discussed it already
> with Yuan. Our
> first approach based on our Nao-Code ("Simple Soccer Agent") appeared to
> be of limited use
> because to many details were hidden for unexperienced users.
> Now we are preparing a new agent in JAVA (were some credits are given to
> Magma Offenburg),
> and first tests are very promising.
>
> It would be comfortable to have some “exercise mode” in the server with
> features like that:
> - players can start immediately without any restrictions, i.e. no
> kick-off, no time limit, no kick-in etc.
> - players can beam at any time
> - a coach can place the ball to any position
> - a “GPS” informs about exact locations of players and ball (may be
> only the coach has access to the GPS)
> - (what more?)
> I don’t know, how far such features can be already realized by
> configurations. Anyway it would be nice
> to have it “at once” as a special mode.
>
> What do you think?
>
>
> Another topic:
> To come closer to real robots, it would be nice to control stiffness –
> what about that?
>
> Best regards
>
> Dieter
>
>
>
>
>
> Am 03.09.12 11:40, schrieb Yuan Xu:
>
> Hi Klaus,
>
> I have one more suggestion, hope it is not too late.
>
> I would like to include the battery and power consuming (I have
> already implemented),
> since limited resource exists in all real world application,
> and furthermore it will (hopefully) restricts the ability of one
> robot, e.g. dribbling ball all the time.
>
> On Fri, Aug 31, 2012 at 4:08 PM, Klaus Dorer
> <kla...@hs...
> <mailto:klaus.dorer@hs-**offenburg.de<kla...@hs...>>>
> wrote:
>
> Hello,
>
> thanks to those who have contributed to the discussion of new
> features
> for 2013 3D simulation league. If you want to add anything you
> should do
> so within the very next days.
>
> The technical committee is then starting internal discussion on
> which of
> the suggested features to use. We try to have decisions and
> specifications ready in October.
>
> Greetings
> Klaus
>
> ------------------------------**------------------------------**
> ------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond.
> Discussions
> will include endpoint security, mobile security and the latest in
> malware
> threats. http://www.accelacomm.com/jaw/**
> sfrnl04242012/114/50122263/<http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/>
> ______________________________**_________________
> Sserver-three-d mailing list
> Sserver-three-d@lists.**sourceforge.net<Sse...@li...>
> <mailto:Sserver-three-d@lists.**sourceforge.net<Sse...@li...>
> https://lists.sourceforge.net/**lists/listinfo/sserver-three-d<https://lists.sourceforge.net/lists/listinfo/sserver-three-d>
>
>
>
>
> --
> Best Regards,
>
> Xu, Yuan
>
>
> ------------------------------**------------------------------**------------------
>
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond.
> Discussions
> will include endpoint security, mobile security and the latest in
> malware
> threats.http://www.accelacomm.**com/jaw/sfrnl04242012/114/**50122263/<http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/>
>
>
> ______________________________**_________________
> Sserver-three-d mailing list
> Sserver-three-d@lists.**sourceforge.net<Sse...@li...>
> https://lists.sourceforge.net/**lists/listinfo/sserver-three-d<https://lists.sourceforge.net/lists/listinfo/sserver-three-d>
>
>
>
>
>
>
>
|
