 Welcome everybody, this talk is going to be about Adam. Adam is Dutch for breed and this project has something to do with air quality and breeding. My name is Dach Wies, but we only have 23 minutes, so I'm not going to talk about myself today. I'm not going to talk about tamlop, but tamlop is a faplop in Ghent where I live, so that's the only thing you need to know. It looks like this. We have lots of stuff to play with. We have a nice logo, and the logo is special in the sense that you know what this means. It's like electronic schemes, but it also includes a bird. Like in the coal mines, they used to have birds with them to monitor air quality. If a bird died, then there was something wrong with the air quality, and that's what we actually want to present as well. The Adam project we would like to do is collect raw data, in the first instance, raw data of air quality information. We are focusing on fine dust, because fine dust is something where we don't have a lot of information. In the sense that we do have some information, but usually it's on specific places where they have the expensive equipment to monitor. In Ghent, for instance, there are only two places in the whole city of Ghent where they are measuring quality of the air. The location of these measurement devices is very strange where they are put in a park and stuff like that, which doesn't represent the real quality of what you would like to know. We think that air quality only makes sense. It's like in nuclear reactors you have this personal device that actually measures what you're exposed to. So what we would like to do is have something similar in the sense that we can make something that tells you where you are, what the air quality is. Obviously it will not be exactly because it's very dependent on lots of things, but if we can collect a lot of information and we're going to do it by having a device that you can mount on your bike, if we can collect a lot of information during the day and we can use that information together, we can enhance that information because we get a lot of that information. We can maybe collect information that while the components we use may not give you exact information, because we get so much information we can get better results from it. And we want to get raw information in the first place because then maybe data scientists can use that information to put it on a map and do things with it that we didn't envision. So the most important thing today is the device itself, so I'm not going to talk about the health obviously. Find us is only one of the things you can measure about air quality, but it's one of the things that, like I said, there's not that much information of it and the quality of that information is debatable. Also for find us, the World Health Organization and the European Union, they have also debatable levels that are acceptable. For instance, in Europe, the first 35 days that are measured that exceed a certain level are not taken into account. So that's a really strange way of measuring things because it makes no sense. So you can have very high levels of find us, for instance, and those are not counted for the 35 days. Only the 36 days taken into account. Very strange, but it's because obviously find us is a problem. Okay, so there are different kinds of find us. You have PM10, which is 10 micrometrics, you have PM2.5 and PM1. And all of these have a different kind of, how do you say that? It's a sense of how it's dangerous. The PM10 is, you can breathe in, but it doesn't go as deep as for instance PM2.5. Those can get into your bloodstream, even if it's getting worse, ultra-fine particles, they go directly into your bloodstream, even not directly to the lungs, but even by other means. Find us may not be caused by human activity alone. You also have find us, obviously, that the nature produces. So find us is not necessarily always related to us, to cars, to central heating systems and stuff like that. But it's one of the indicators that our quality is not that good. So yeah, Ghent, the nice city of Ghent. We have lots of students, lots of bikes. A lot of people use bikes. I think in Belgium Ghent is probably the city where people use the bike the most. But we also unfortunately have a highway going almost directly into the city. And I'm living up there at the end of the highway, so I have a personal reason for doing this project. So not that important. Probably you all have these concerns where you're living, especially in cities or near highways. Like I said, yearly averages or daily averages don't mean a thing. It's the exposure you get for yourself that makes a lot of sense. The things that we would like to get is that we have a good indication when, for instance, it's good to open the windows to clean the air in your house. Sometimes you know from the weather announcements that the air quality is bad, but sometimes it's related to something closer to you. So it would be nice to be able on a map to tell when it's best to open your windows. People do it usually later in the evening, but in the winter that's worse because the central heating systems, everybody is burning gas or whatever, or wood. And then the air quality is even worse than during the day when there is just traffic. So it really depends on, you have to measure in-house and out-of-the-house to be able to know if it's better or not. The device, this is one of the first designs where all the components were put together. It's changed a bit. We're going to see what we had to change, but this is how it would be put together in the first version. Our idea is to have an open device, open hardware, open firmware, and also how the device is put together. That also will be open. We already have played with that as well because it's a track next to the technical part. We're using as much as possible open source libraries. I don't have to tell you that here because everybody is doing it here at this place. We tried to make it as cheap as possible and if we have time I will show you the bill of material. I hope we can make a device like this for 40 euros. Real devices that measure it in a scientific way, they easily cost 1,000 euros or more. So this would make a difference for us. This is also an earlier design. I didn't make a picture of the latest one, but this is basically how it works. We use the Sparkfilm ESP8266. The sensor at the top is the Find Us sensor. It can measure two and a half micrometrics and one micrometrics. Particular method that is in between those in that range. I will show you how that works later on. There is a GPS in there because we obviously have to locate where we are. We have a buzzer, we have a light to give some indication to the user itself because we think that we get more users if the users have feedback themselves as well, if they have a personal interest. And then we have a humidity sensor. Temperature is also in all of these accelerometers to see when we are moving. That's important for your device as well and barometer is the air pressure sensor as well. And while we're collecting that information as well, it's because that might influence the information you get from the air quality. Well, the Find Us sensor. We don't know yet what the relation is. We do a bit on scientific papers that are written about the sensor we're using, but we will be able to verify ourselves based on well data as well and obviously based on calibrated material. So, like I said, we have a microcontroller which is a Sparkfen ESP8266. The reason why we took that one is because it has Wi-Fi. That's the most important thing in this case. Because in the earlier designs we were thinking about using Bluetooth, but that's a lot more of a hassle. Like I said, the Find Us meter, we did try a lot of sensors. I think we bought five or six sensors. Also, there is a better one in the PPD 42NS, which is more expensive obviously, but we didn't see that it was much better in the results that we wanted to have. So, what is important to say in the latest design, we have started to use a second microcontroller to be able to convert from the GPS to I2C. I'm going to come back to this while we're doing this like this. I'm not front of the ID, but that's the only affordable way to do it at the moment. The Find Us sensor is very important because it's the basis of everything. One of the things I have to mention as well, we're now doing Find Us in future designs. This is all working and we have spare time. We're going to probably add other sensors as well because the design allows it. It's a modular design. You can add more sensors to the I2C bus, so that we can do as well. But we first want to focus on this and get something working. What's important about the sensor? The sensor works by sending out, emitting infrared light. We have a lens and a sensor where the light is being reflected to. It gets pulses based on what the sensor receives. The sensor itself already has some fine tuning. You see that here. It has two potential meters to be able to tweak it, but it's normally tweaked at the factory, so we don't have to use that. The air goes in here. There is a resistor to heat up the air, so you have an airflow for standalone devices. In our case, we're probably not going to use the resistor, and we're going to use some effect to have an intake of air automatically so that we don't need the resistor, which will save the battery. What do I have to tell about this as well? Well, the PPD-60NS does have two sensors, if I'm correct. So it gets better results. This does not measure black matter. That's the only problem with this sensor. The other problem with this sensor, because it doesn't reflect light, obviously, the second problem with this sensor is the fact that... I forgot what I wanted to say. Normally, in scientific and in the World Health Organization standards of acceptable levels, they measure fine dust by measuring the weight. Obviously, this doesn't measure weight. It only measures the size of particles. That's one of the problems if we do it this way. It's less scientific, but obviously, if you want to measure weight, you have to collect the dust and you have to weigh it, and that's impossible to do in real time, and that's the reason why these real devices are so expensive. There are better devices than this one using lasers and better techniques and stuff like that, but those are also very expensive. This one is not calibrated, so the calibration is... One of the things we have to do ourselves by buying calibrated devices and maybe based on statistics and where you are, and if you pass a calibrated device, we can take that into account. If cyclists cross each other, we can take that into account if some of them have just passed a calibrated device. Things like this are possible, but we're not counting, in this case, the weight of the fine dust, but we're only counting particles and the sizes of particles. Then, what we have to tell as well, it's an Arduino-based framework. I don't like Arduino myself, the IDE. I don't like the framework. Maybe it's better because there was a new release. I didn't test that yet, but we're also using Platform I.O., and that's a lot easier to work with, and we do the CI testing automatically on GitHub with Platform I.O. Components can be swapped, so that's the nice thing about the I2C bus. We also plan to have a driver selection, so if you have different kinds of devices, you can just plug them in, and if we support them, they will be scanned automatically as good as possible with the IDs. We accept contributions, go through the brouhards, this is all not. We have a state machine, obviously, this is already implemented, and everything is actually implemented now, but the integration, we didn't do a lot of testing yet, because we want to get rid of the breadboard. We have two PCB designs yet, but none have been produced yet, unless we have everything working on the breadboard first, but we're very close to that now. Let me quickly do this, because this is nice, because you have an idea of what it does. When you start it, obviously it has to do some initialization, and it directly goes into sleep mode, because that it's not doing anything and everything is disabled, except the accelerometer, which is always on. As soon as we're moving, we go into some sort of GPS test mode where we try to get a fix for the GPS, because that's not certain, and once we get a GPS fix, then we start to collect, because it doesn't make sense to collect information if you don't know where we are. Obviously, if one of those two are gone, we go back, or we go back if we're not moving. If you shake it, we go into a config mode that opens a Wi-Fi access point where you can configure what are the Wi-Fi credentials and what are the SSIDs that you're going to use for connecting to Wi-Fi. We plan to have multiple, so that if you cycle to work, then at work things are being uploaded. If you go back home, they are being uploaded. We expect that if the device is not moving now, it's always moving a bit, but not moving a lot. We go into Wi-Fi test mode. If we have data collected, so only if we have data collected, we will see if we have Wi-Fi. If we have Wi-Fi, we test every 30 minutes or something like that. We have to see what's acceptable. Then we upload, and as soon as it's uploaded, we go back into sleep. So that's very simple. There are certain cases where we probably want to do things differently. If you go into a tunnel, maybe we want to only go into sleep a little bit longer. We'll wait some time before we do that. The current device, like I said, the hardware design was done before I even joined the project, but things changed a bit. The components were already evaluated before I joined, so that was needed obviously for making the right choices. I still have seven minutes. That's fine. We now are on our second prototype, which is still breadboard-based, which also has its own problems because you have these cables that are not always making the right connections or there is lots of information and things like that. I'm not that technical. I'm not into electronics as much as I probably should be at this point, but we're working on a team. I'm not doing this alone. I'd like to learn more. We have a second piece of B-design already created, which also includes the second microcontroller. We also have redesigned the casing because of the connections. You only learn those things by making something and see how this would look. We also made it a bit smaller. The idea is to use a standard tube size that is transparent, but not too transparent, so the lights are reflected towards so that we don't need to make... It's going to be mounted on a bicycle. It could be raining, weather conditions, so we have to make sure that it's properly sealed. If you take a tube which doesn't have any openings on the side, then you only have to care about the front and the back. That's going to be 3D-printed at first. We already have a design where everything fits into it. At the back, we have access points for cables and at this moment for testing stuff. In the front, there will only be something where air gets in. That's also a worry in the sense that how will this survive extreme weather conditions or long usage? We still have to see. We have ideas about that as well, but it's too many to go into. We had a lot of issues as usual by these kinds of things. We had too few eye-opens, and that was one of the reasons to go to a second microcontroller. You could solve it in other ways, but by having a second microcontroller that only does the GPS at the moment, it only does the GPS, but in the future it could do other things as well. We can offload the GPS handling on that second microcontroller. That's one of the advantages. But the less interesting part is the fact that we have to write two formulas. One for that microcontroller, obviously, and they have to work together. That was one of the other problems that we had. We also had problems with imperfect signals. One of the things is that, for instance, our sensor expects 5 volts, and at the moment we have a battery that can go up to 4 or 2 volts, I think. So we probably have to use a better battery in the future. We still don't know because we haven't looked at the data the sensor is collecting yet. We first want to finish it, and then we want to do real testing. We did obviously do some things with it, but we don't know what the effect is of only giving 3 to 3 volts or more, a bit more to the device because obviously if the light is too weak, the results will be influenced. So we have to see how that adds up in the global picture. What can I tell as well? We had some issues with the ESP, stability, but also the fact that it has to give time to the integrated wifi stuff that for I2C communications we had a lot of problems that we didn't understand. We found lots of people that were having the same problems and then we found what the issue was and we could fix it. We had to change the I2C library a bit to make it work in a stable way. Some other stuff doesn't really matter. I still have three and a half minutes, so let's see. There are lots of things we would like to do. We have lots of organizations that are very interested in the possibilities that we have with this device. Collecting road quality information by if you have bumps and a lot of bicycles have the same bumps at the same location. We could probably do something with it. No air quality sensors, like I said. Obviously, when you have a moving device we could also make a stationary device. It's a little bit different because we may not include the GPS but we still want to know where we are so we still have to see how we would do that but it would be nice to have this as well. Obviously, if it's outside of the house we do want to have that information. If it's inside of the house obviously we don't want that information. But we may still collect it and show it to the user because the raw data is not usable for normal people because you have to see that in the global picture of all the data that we measure, both the calibrated stuff and the non-calibrated stuff. But the stationary devices if they're spread out through the city it would be nice to calibrate those maybe ourselves from time to time so that those are also giving a better indication of the air quality and those can be used when cyclists pass by. We also want to correlate air quality with weather conditions so that in the future we can also predict based on the predictions of what the weather is going to be what also the air quality will be based on obviously if it's a normal day a weekend and whatever. Because air quality like I said it's important to do that close to where you are because it doesn't matter if there are only two collectors where the information is guarded it's of no use it depends what you are breeding in. The reason why I'm saying that is that yeah, I lost that track as well the reason why I was going into that as well but I forgot Bill Boecht's showing in real time actually there's also some interest for using advertisement space that is underused to give that sort of information as well which would be nice in a train station that we could also display that information in between. Because one of the things that's important is that you can get people that people are interested in these matters as well and if one minute left so the demo will be I can give you a demo later on if you're interested and we'll go somewhere else if you're interested to help we're looking for more people to join as well now that the device is almost finished we need obviously people that also want to test drive around help with integrating and stuff like that like myself no prior knowledge is required and you learn a lot okay this is the information if you're interested take a picture now the slides will be up somewhere the video will be up somewhere we quickly see if there is something a lot of other projects are going on as well we learn from that as well because it may seem that I'm a little bit superficial over all the stuff but it's based on other projects as well it's just that we don't have a lot of time yes, questions quickly I've looked up online the data detector from Sexton and it said that it should be powered by 5 volts correct so repeat the question yeah we know can you repeat the question correct indeed, we know so the question was that the datasheets say that 5 volts is needed there can be 10% difference and we know that there is an issue but it's possible that the information we collect is lower also relates to the output we get but we want to have 5 volts for this definitely that's why we're looking for different batteries maybe I'm sorry to interrupt you we can take other questions outside so that we don't shift the schedule too much obviously thanks a lot for the presentation