Dear Stefano, as you have invited comments about the issue, let me put in my ten cents as the responsible professor in Osnabrück. Given that you have discussed intensely the concrete rules for the contest in Rome, I see the point of defending the results (although you have labeled them "tentative"). Our point is not to cause you to go back behind any decisions made collaboratively -- as you wrote, "too bad" for us. But let me forward the point for the next discussion about the USARSIM rules for the next event to come. We certainly agree that the whole point of this league (like in RoboCup in general) is promoting science. If you look at our Web sites, you will find Gigabytes of PDFs, videos, and downloads. In particular, we have made USARSIM models public, as Andreas has written. So, definitely no disagreement on that. In the case of the USARSIM league, I used to think that one of its goals were to allow for direct correspondence between the simulated robots and real robots in the Rescue Real Robot league. In fact, we feel that this is a perfect aid for science: provide a truly realistic simulation environment for beta-testing your real robots, and experiment with new control software or simulations of sensors that you don't have yet physically. This is the goal towards which we have worked, given that we keep participating in the Real Robot league (Kurt3D in 2004; Deutschland1 in 2005 and, hopefully, 2006). Consequently, we have worked hard to allow our real robot control software to be usable _as is(!)_ in the USARSIM environment, i.e., we have modeled our real robot, our real sensors (including the 3D laser scanner), etc. All this USARSIM code is public, even now!! Does the Real Robot league force teams to open all of their robot control code? No! Neither do the other RoboCup real robot leagues! And why? -- in addition to heaps of uninteresting technical stuff, each and every Real Robot team (at least the succesful ones ;-) have their set of crown jewels, which is at the heart of their scientific contribution and, if all else works well, the reason for their success in the competition. (Our crown jewels are undoubtedly in handling 3D laser data.) If you force these people (the ones with good real robots) to lay the code completely open, they will go away instantaneously. We all DO publish the stuff -- after all, we are scientists with a keen interest in publication lists and citation indices. The development of the soccer midsize league proves that there is a steep learning curve in the community by just promoting scientific publication of the techniques, not free code access. In consequence, if you enforce publication of the whole control code of USARSIM robot simulations, you apply an effective means that no team working with identical robots in the Real Robot league and in USARSIM will participate in the simulation contest. If that's what comes out of the attempt of pushing scientific progress, then maybe it was not an appropriate means. By the way, we will definitely NOT follow the idea of exchanging our real robot control software only for participating in the USARSIM league. There is enough work to do before participating in such a tournament that we want to save the additional effort of making our control software worse than it effectively is... So, "Roma locuta, causa finita." as another community tends to say. Okay if a decision is executed, once taken (although I am sure our students won't like it). But maybe Rome might rethink the matter next time. Good luck with the 2006 competition, which we will watch with keen interest! Feel free to use our free Kurt3D robot model and 3D scanner sensor model! Best regards, Joachim Am 14.11.2005 um 16:51 schrieb Stefano Carpin: > Dear Andreas, > > thank you for your email. Here are my answers. Others (Mike, Alex, > Steve, ...) are welcome to add, of course. > Andreas Nuechter wrote: > >> Firstly, we cannot publish our source code. We started USARSIM with >> our original Kurt3D software, where we've integrated the simulator >> into. (I assume that some participants choose the other way round, >> i.e., creating special software for USARSIM). Kurt3D's software >> development has lasted for 6 years, parts of the software belongs >> to Fraunhofer AiS and are _not_ open source, thus can't be made >> available to the public. > > > Too bad. We discussed this aspect a lot during the Robocamp in Rome > and we came to a uniform consensus about this fact. We see the > league also as a way to foster research and disseminate results to > the widest possible audience. I guess the same spirit moved Mike > Lewis' group to make available as open source the entire USARSim > project (he would be the best person to comment, though). This > requirement we've introduced is still less demanding than what is > asked in the Rescue Simulation League where ALL teams are required > to disclose the source code. If you can't publish it, you can avoid > using it and still partecipate. > >> Secondly, we aren't able to publish all new robot models by >> December 31. In Osnabrueck practical work is mostly perfomed by >> students when there aren't lectures, i.e., in February and >> March. Thus it must be possible to update the robot >> model. Currently our work plans are: improving the Kurt3D model, >> creating the Kurt3D model with the RTS 3D scanner, adding new >> sensors, etc. pp. > > I understand that December 31st is a close deadline, but you have to > consider two aspects: 1- people at NIST have to validate all these > models (possibly correct them to adhere to the new USARSim > standards), and embed them into the official USARSim release(s) that > will be used in the competition. The later we fix the deadline the > more pressure we are putting on them (and they are already doing a > lot!). 2- the idea behind the simulated leagues is to offer to all > participants a set of agents with the same capabilities, so that a > common starting point is shared among participants. Of course we all > need time to experiment and evaluate different platforms and so > on. If platforms are published in late March, than there will be > little time to develop code based on them. Only those who developed > the platform will take real advantage. As a starting point I think > that the platforms we've used last year (with the addition of the > Tarantula model developed by Alex Kleiner at Freiburg) are a good > starting point and we think are the reasonable starting set for this > initial competition. > > Best, > S. > > -- > Dr. Stefano Carpin > Assistant Professor of Computer Science > School of Engineering and Science > International University Bremen -- Germany ------------------------------------------------------------------ Prof. Dr. Joachim Hertzberg University of Osnabrück Institute of Computer Science Fon: (+49) 541 969 2622 Albrechtstr. 28 (Room 31/507) Fax: (+49) 541 969 2799 D-49069 Osnabrück hertzberg@informatik.uni-osnabrueck.de Germany http://www.inf.uos.de/hertzberg/ ------------------------------------------------------------------