 So, I'm sorry, I'm not the inspector here, but I'm going to work for the task force. Yes, I don't think there's one. But you have done them, I really can take them. This is a mile up top. I did it already before. Maybe I just want to know if this one's working or not. You have a big appetite now. I would like to try to contact Sebastian. He just insisted that I should... Did you shoot the video? There's nothing to do, so I... You can try to call him on IRC. Yeah, but I'm not connected actually. I just try to... It's always deconnecting. Maybe I should try to... Oh, we're already late? I don't know, is it 15? Yeah, so, okay everyone. We're running a bit late because the last talk was a bit over time. Grab a seat and welcome our next speaker. Is it on? Because they can't... Hello? Is that working? Cool. Are you okay like that? Yeah, I'll be fine. Okay. Hi. I'm Nathan Hughes. I'm an undergraduate student at Aberystwyth University. And I'm currently on an industrial placement at the MPPC, which is the National Phenomics Centre in Wales. This is a project I've been working on. It's an open gravimetric phenotyping system, which essentially is an automatic system that waters plants and makes sure they don't die. So yeah, like I said, it's an automated system. It records biomass growth and water usage in plants. It records all its data automatically into a MySQL database, and it has the capacity to augment with loads of other cool sensors and that sort of thing, like temperature sensors, solidity sensors, that sort of thing. So one of the cool things I'm working on at the minute with it is automatically putting your data into a usable form so that the end user can see what's going on. This graph over on the right is water usage for a plant for the last 10 or so days. And it's really handy because you can see on day four I turned off the watering tap and just let me know because I could see, oh, there's something missing. Yeah, this is the web interface for everything that lets you change details on plants that are on your system. And yeah, it's all written in Python using WebPy. So the system as a whole is controlled with a whole network of Raspberry Pi's. Other pictures have come through great, but so this one here on the left, you have a master Pi which network boots a load of Pi nodes and SSH's into them and gives hardware instructions of what to do. All of the control software is all written in Python but the low level hardware control is all done in pure C. And I think the big question is why would you want a system like this? If you wanted to keep plants alive is the big thing. If you wanted to do any sort of research on plants and create data which is very accurate and you don't want to stand there 24-7 recording things by hand. So yeah, people also the other thing you want to automate a system because people make mistakes. So there's sort of computers but they make slightly less mistakes, I hope. Yeah. And also you don't want to be there at three in the morning writing down plant readings. So somebody might know you can get automated systems, commercial systems that already do this and do it supposedly well. So why would you not just buy one of these systems? Like why am I working on this project why don't I just use a commercial system? And the reason is there's loads and loads of problems with them. I'm not going to name names or point fingers but they're very, very, very proprietary and that makes me very sad. So yeah, all of these commercial systems are generally, all the hardware is closed and all the software is closed so we don't know exactly what is going on with the data that's going through the system. Yeah, and they're not adaptable. Our system is very adaptable. You can add on any sensors you want. You can get more data out of it and especially if you're working on a sort of scientific environment what you want to study today might not be what you want to study next week and you don't want to have to spend a few hundred thousand euros to change that. And also whenever parts break they're all proprietary parts that are extremely expensive whereas what we've done is very generic parts and easy to replace and modify. The other reason these proprietary systems terrify me is some of them will actually take data that you generate and they will send them off to their remote server to store them and then you don't already know what happens. So that creates all sorts of problems of, you know, from a simple as if your internet connection cuts out you can't access your data that was created in your greenhouse because I don't know why they do this. It's a bit silly really. And also the reason I don't like my not knowing what's happening to the data we're generating is because of this graph and anyone who knows me will know I love to show this graph off. This is a load cell, so a scale with constant weights put on it, so two kilogram, one kilogram weights and you can see over time there's a load of fluctuations because hardware isn't very good at reading data for a long time really. There's lots of noise. And the problem is on these commercial proprietary systems you get straight lines back you know what sort of averaging algorithm they're using what sort of sanitization of your data they're doing which is kind of scary because if you're using this for scientific research you don't actually know what's happened to your data in this process. So that's why we've made this system and we've made it really open so that you have full control over your data and anything you're doing with it and there's complete transparency on any sort of averaging we're doing any sort of sanitization we're doing of data and yeah, so we've made everything open from specifications we've used for hardware to the software we've written and I zoomed right through this talk super fast I'm really sorry, so I'd like to say thanks for listening to me thanks to Aber University for having me and thanks to a lot of people at the MPPC for helping me produce all of this work and be the SRC for funding it and as I said all of the source codes all on GitHub at the minute feel free to have a look and yeah, if there's any problems let me know because I'd quite like to find some bugs in it Any questions? I'm coming with the mic Hello, are you using any sort of well-defined data formats for exchanging for sending the data around the system so for example is there a common format that would already exist for communicating data about plants and moisture and things like that that you're using or is everything sort of new and different? Yes and no, I know there's quite a lot of I think it's isotab formats that are trying to keep data organized in and a few things like that I'm not 100% on that but I know they're trying to but it's very unorganized at the minute Hi, you said that the proprietary solutions are very expensive I'm just curious, the setup that you've got going at home how much did you totally spend on it? How much have we spent on it? Well, there's been a couple of iterations of it but I would say you're talking a tenth of what a commercial system would be so whereas a commercial system you might have spent upwards of a few hundred thousand euros our system is definitely well under a hundred thousand probably quite a lot under but I couldn't give you an accurate number unfortunately Which sensors are at the moment deployed in your greenhouse? What sensors? In the greenhouse there's a lot of sensors so we'll have air, humidity, air temperature those sorts of sensors in GravaMetrics itself in its pure form all we have is we have load cells and it's pretty much the way it was focused on for sensing at the minute but there's a lot of capacity to add an augment more on depending on what an experiment will want to find out really Hello, so my question is related to the previous one did you also design the sensors like load cells or did you buy them off the shelf? No, at the minute they're off the shelf ones but we've been looking into building our own to try and get more accurate readings and to play a bit more with what's actually coming out of them because there's so much variation in because load cells aren't meant to be left on 24-7 there's a lot of tweaking you'd want to be able to do on the very low end of the hardware so that's on our GitHub at the minute well some of our plans and some of the research we've been doing into the hardware side of how load cells actually work Me again, so do you have an API for this data or can it only be accessed through a web interface and did you think about using something like Grafana for example so you don't actually have to write the plotting on websites so you just bothered about data collection I think your mic just stopped working so I don't know what happened It's off or? Yeah, it's not working anymore So maybe his mic is not working anymore so if you have more questions then just meet him outside Thanks a lot for your talk Cool, thank you very much Is it your water? No, this is Okay, I don't know Cool No, members take on I've got it