Dryw Filtiarn

Here is a short time-lapse of the set-up of the factory after transport: The photo's have been made when I set up the factory at the office I work at. You will not that at the end of the animation the sun starts to set causing the photo's to be a bit darker. All in all it took me about 5 hours to set up, but I must say that it didn't help that people kept me from continues building as they had all sorts of questions ;)

In regards to the release of the part being placed in the model, the build head triggers a sensor as soon as it hits something. That means that it will simply keep going until it senses resistance. At that point it will increase it's motor power for a short moment to really make sure the part is really attached and not just lays on top of previous parts placed. To answer a previously asked question I have not answered yet: Is the robot aware of failure to place a part. The answer is no. It's very hard to determine whether placement fails or not, unless you would have some really intelligent video monitoring of the building area. It would be possible to realize such a thing, but the amount of effort you would need to put into such a thing is insane. Obviously things do go wrong in placement and that will basically mean, that you will have to stand by the factory during building to correct possible failures to place a brick. In regards to the number of failures in actual building. After some fine-tuning (which could have been improved even further if I would have put some effort into that) at the LegoWorld event, I got to a point there where only about 1 out of 50 bricks failed to be picked up by the build-robot. And about 1 out 10 failed to be placed in the model. I expect to get the placement failure-rate to be reduced when I improve the programming in the upcoming weeks and I'm hoping to get it reduced to about 1 out of 25 bricks failing. At that point, considering the accuracy needed for a piece of machinery like this, I think I can be pretty happy with the reliability. A situation where not a single brick will fail, will probably not be possible to achieve when you consider that the margin between correct placement and failure to place a brick is about 0.1 to 0.2 millimetres. In theory that even is a level of accuracy which is technically impossible to achieve with Lego, when you consider error margins in the rotation-sensors of the motors, the slack you have in gear transmissions, etc.

Technically the model is completed both in brick-building and programming. But Indeed, I will be investing some more time to get things working in parallel. At the moment everything is build around a task-based instruction, where basically one of the NXT's is the master which sends a sort or remote control command to the other NXT's and waits for the response of that. The current implementation of the software therefore means that everything is sequential. I'm going to rewrite this so that the robots will work more independent of eachother so I can create a situation where one robot is placing a brick in the model, while the other one is retrieving the new brick from the warehouse. In regards to limitations to models, basicly there aren't really that many limitations. Due to the way I handle bricks you obviously need to use stud-on-top building. There is also a current (software) limitation that you can only place parts at straight angles (0, 90, 180 and 270 degrees of rotation). This can be changed through software though, where nearly every angle would be possible. The situation you describe of placing a plate next to a brick and then continue building on the plate isn't a problem. The current building head can handle a 1 stud height difference between elements. My factory is capable of the following situation for instance (which I think is an excellent example of the posibilities in regards to how you can build a model): ┌─┬───┐ │o│o o│ │ ├─┬─┤ │o│ │o│ ├─┴─┤ │ │o o│o│ └───┴─┘ Take above sitation, 4 1x2 bricks placed in a "circle" with a clear stud in the centre. My factory can place a 1x1 plate or 1x1 brick in that empty centre hole. One thing you obviously do need to take care about is that Lego elements in a model are supported on both ends by other elements. The factory will pick up an element at one end of the element and after initial placement will push the element again on the other hand. If elements aren't supported this way of building will obviously result in the model being destructed again. So: ┌╨─╨─╨─╨┐ ├─┬─────┘ │ │ └─┘ INVALID! ┌╨─╨─╨─╨┐ ├─┬───┬─┤ │ │ │ │ └─┘ └─┘ VALID! The above situation where the 1x4 plate on top is only supported on one end by a 1x1 brick can't be done. Placing might go right, but once the element gets pushed on the other end of the element things will break down again. Even when I would not do this extra push it will certainly go wrong at a later stage when you would try to put other bricks on top of that plate. The bottom example is possible, as the 1x4 plate is supported on both ends by a 1x1 brick. It won't break when placing the 1x4 plate itself and you can build on top of it without any issues. The first example with the clear overhanging of a part is possible though with slopes! as long as the stud of the slope would be above the 1x1 brick. The sloped side of brick will never be pushed in any way, so that's a valid (and safe) building method. Finally in regards to the speed, there isn't much to speed up in regards to the robots. If I would use higher motor speeds it will also result in less accuracy. The factory can only function with the correct balance between speed and accuracy. Overall the building of a model can be done faster though once I get things reprogrammed so the robots will work in parallel to each other rather then sequential. Currently it takes about 1,5 to 2 minutes to build a brick into the model (so a 60 brick car will take around 2 hours). For further information on the project (and future versions of the factory... yes I'm already planning on Factory 3.0 which will be even more advanced possibly using SNOT and tiles) I will create a website with information about the factory, it's development and the techniques used (though some things I'll keep a secret as I consider it my trademark and I want people to think about it for themselves). Once the website is up and running I'll announce it here (and elsewhere ;)) More video's will come later also showing bigger models being assembled, but that will take a while. I'm going to improve the programming first so that the whole assembly will become faster overall and then I will try to get Mahjqa to come to my place to do some recording again. Don't expect these video's to come very soon though, it might be halfway or late next month.

It has taken a while, but here is a first video that shows all of the factory's features in a single video: De video recordings have been made (and edited) by Mahjqa (Peer Kreuger) on the last day of the LegoWorld 2010 event (in Zwolle, The Netherlands). Unfortunately by then we were unable to make recordings showing some advanced models being build as during the event a lot of dust and dirt collected in the factory's mechanics, causing it to loose it's accuracy and making it very unreliable. Anyhow, this video will now show you a bit more of the factory and it's functionality then the "work in progress" video's shown so far. At a later stage I will make new video recordings in a "studio" setting, and with a cleaned up and reprogrammed factory. The video's that will be recorded then will certainly show a lot more of what the factory is truly capable of!

Well obviously it would be possible to create a little program for the Mindstorms RCX that will make it possible to create a toggle function for a touch sensor. Unfortunately I don't have the software for the RCX at hand so I can't create an example for that, but you would be able to do it with the following psuedo-code: start loop set output on wait for sensor pressed set output off wait for sensor pressed end loop With above you would have a situation where the siren starts off making sound, then when you press the touch sensor it switches of and when you press the touch sensor again it would turn back on and the cycle is restarted again, where you can press it again to turn it off again. If you would use NQC (using Brixcc) to program it you could use the following code: task main() { int sirenOn = 1; SetSensor(SENSOR_1, SENSOR_TOUCH); OnFwd(OUT_A); while(true) { if (SENSOR_1 == 1 && sirenOn == 1) { sirenOn = 0; Off(OUT_A); } else { sirenOn = 1; OnFwd(OUT_A); } } } I hope that I understood your question correctly.

The rail ferry was build by a member of Freelug (France), I'm not entirely sure the name of the build was though. It was hard for me to communicate with them as I don't speak French and they didn't speak English :)

I was made aware of the project of Will Gorman and even have spoken to him in person at LegoWorld in Zwolle. He was there as well with his MakerBot. It's a really interesting machine which works excellent. Although the concept between them is somewhat different, there are quite some similarities between his project and mine. Some of the techniques we use to pick up bricks and place bricks are pretty much the same. TheNXTStep indeed has a picture currently on the blog. During LegoWorld we have made about 1.5 hours worth of video, which needs to be edited, merged together and put online. I expect this to happen upcoming weekend after I have revived a bit after 9 days of LegoWorld. Once these first video(s) are online, I'll be adding them to this topic and the videos of the factory will also appear on various common Lego blogs and sites. In about 2 weeks from now I hope to have some updates to the functionality (especially the programming of the factory) which will further improve it's efficiency and after that I will create some high quality video's showing every aspect of the factory without all the "noise" in the video with the crowds at LegoWorld walking by and such.

A first video of the factory functioning will be coming online before the end of the week.

I took me a couple hours of work but, but everything is disassembled now... All packed and ready to go to LegoWorld in 9 hours from now. For those of you who go there, make sure to come by and have a chat :)

A short update of last day. I am nearly at the point where I have gotten everything to work together. I've been doing some short trial runs, but I had to abort most of them due to the computer disconnecting from the NXT, causing things to go wrong :(. That issue is now solved as it was caused by two processes in the software trying to communicate with the NXT at the same time occasionally, crashing the blue-tooth chip on the NXT (which resulted in that only a restart of the NXT would get things working again). Beside the programming business, I've also been working on some of the detailing of the factory, so that there's more to see then just the technic construction and it's functionality, just a bit of fun to please the eye, so to say. You can click the images for a larger version of the image. The forklift was build after the instructions of Brickshelf user "smhltec"

I really have no clue at this point how long such a model would take I'm currently in the final stages of getting the instructions send from the computer to the factory. As soon as I get that solved I can do a trial to see if everything works as it should and how long things will take. If you ask me for a really, really rough guesstimate I expect it to take around 40 minutes (for 59 bricks in the house model), but then again, it might be more and it might be less. I will know no sooner then I actually get it to really build the first model.

A short video showing the software at work. It's not yet completed but I'm close to getting there. I expect to have everything up and running either by tomorrow evening or thursday. Progress is going smooth so far.

Things are going smoothly currently, all the mechanics have been verified and changed where needed and now work like a charm. Basically this means that it only comes down to writing software, software and even more software. The robots have now been programmed for the biggest part, although there's still some work left to do on that. I'd say each individual robot is now currently at about 95% completion in regards to the software for the NXT's. For the past few days I've been focussing a bit on the software that will run on a laptop, which will send instructions to the factory. I'm now capable of loading LDD (Lego Digital Designer) models and process them. Most of the functionality of the Windows software now also has been implemented and the main thing left there, is doing the final step of translating the brick locations in the model into factory commands, so the model will be build. Here are some screenshots of the software I've been writing to run on the laptop: This is the screen that I will be looking at, it gives me all the controls I need, to be capable of controlling the factory. I will also have a secondary screen attached to the laptop that will be directed at the visitors so that they also get some information on what the factory is doing. The "visitor" screen, is yet to be extended with more information. Note: you currently see some blank spots in the warehouse, these spots are all to be filled in, but I haven't exactly determined which parts to use there

A short update on the project. Unfortunately progress hasn't been as smooth as I hoped, but the warehouse robot is getting along nicely anyway. During programming I ran into some issues with the mechanical construction of the robot, requiring me to do some minor rebuilding/adjustment to the robot's construction. The "rebuilding" is nearly complete and I only need to change one final part tomorrow. After that I can continue on the programming of the robot and hopefully complete it within a day or two. Once done, that means that all robots are ready and programmed, and then the final thing to do is build some Windows software to do the controlling of the factory. Plan is still to be able to build a small model in LDD and then import it and have it build automagicly, I'll see how easy it will be, otherwise I'll do some fixed models for now and do some proper software programming for the controlling software after the LegoWorld Zwolle event. Unfortunately time is starting to run out, with only 14 days left, still hoping I will manage, but if there aren't any setbacks anymore, I think I'll make it. New video's will come later, but due to the amount of time left, might come only after the LegoWorld Zwolle event, when I'm back at home.