Nicolas Sornin - Inventor of LoRa talks about the future of LoRa

next on stage the father of Laura and we're very very honored that he is here but let me just briefly go to his bio so we have a little bit more of background because a long long time ago he started at IBM research doing a radio frequency design he worked at neurologic Cambridge silicon radio it was director research & development at C Clio where he took the patent of Laura and now he's at Sentech allowing us to use Laura in all its ways with no further ado lady in German please welcome Nicolas surname [Applause] [Music] okay how do I start my presentation now here we go okay this is the first time in my life have some loud music come in a room I really hope you get to share this experience one day it's great and it's frightening so Nicola forty-one married three kids I've been asked to talk about the future of Laura so eight years ago I was lucky enough to invent a lot with a school friend so I could talk about the past we have a lot of certainty if four years ago I had to make some prediction that I've been totally wrong so talking about the future of law is very difficult so I've changed my mind I think I'm going to talk about the I can use next okay okay what are in my mind a few challenges we need to solve together as a local community and some interesting topic to work on so basically I'm not giving solution I'm just exposing problems today so Laura is actually just an enabler I've yet to see a product that has just a lot a chip inside it's always the low-rise backbone basically just transport the information but this is not the main functionality of any device so actually there's probably a lot more innovation around lower than in Laura at the moment and actually most of the problems we solved our or our around Laura so that's what I'm gonna be talking about mostly let me start with geolock I think we all convinced that geolocation is is one of the killer app of IOT networks but the problem is it's exactly like IOT when we talk about geolock nobody talks about the same thing so and and that covers so many different aspects let me enumerate a little bit this is one of the problems we will solve is my tag intrinsic somewhere between Paris and Amsterdam actually to solve this you cell ID is probably just enough you know you need 10 kilometer accuracy nothing special what address was my tag deliver to that's a totally different problem that requires probably a Wi-Fi sniffer or some other means of localization and and probably some very clever algorithm this is geofencing he's my tag in a specific area again that depend totally on how big the area is as a country a city a house where's Mike Oh Mike ow sorry this is probably something for for GPS or maybe triangulation but it usually happens in very wide areas where where it's very hard to to do anything actually in which offices asset actually if this is a tower you have another and on there is you're gonna need a for example a barometer to know the altitude and this thing tend to drift and you need some very clever algorithm to recalibrate them from from time to time so basically under all I wanted to illustrate this under geo lock we've got so many things and each of them actually needs a specific set of technologies and and and thinking so what I think my vision as of today is that loja triangulation or lower lock whether it's actually time stamping or so differential time of arrival or signal strength as a good candidate to be the lowest for course localization technology enabling the rest so basically with this you can do geofencing you know if your finger is moving you know roughly where it is and you can use this information to on demand trigger something more precise and this more precise stuff is probably going to be a combination of Wi-Fi and GNSS so basically GPS Glomus betaf kind of stuff but there's a lot of challenges first and we've got to make this Wi-Fi and GPS receiver as low power as possible because for the moments they actually burning a lot more energies and Aloha communicate that they're limiting the battery life of the device second and I think this is the most interesting area we have very limited connectivity to the cloud and mostly we have a very limited downlink we can have a lot of up links but we have very limited downlink Slowhand networks are by nature as symmetrical therefore you cannot do what you do on a cell phone you cannot tell the device what to use so the device needs to be autonomous and this is what I call like we why I say we need artificial intelligence is we need some clever algorithm in the device that decide when you need to do a position fix with what precision when when did you decide that you need to transmit that information to the cloud or lay the cloud what technology should you use you know how you use and very probably all of this relies on some starts with some accelerated data to detect if you moving or not in which direction so I think this is a totally wide field this is wide open Avenue for innovation and we are just just just at the beginning of this so I expect in two three years to see devices that are really you know remotely scriptable and adapt behavior to to the environment another kind of localization which is very coarse is actually on which continent am i and actually that's a very difficult problem and very essential problem to solve because depending if you are in Europe or EU or China or Asia you cannot use the same stack you have to you cannot use the same frequency story so and that's an egg and chicken problem because to detect why you are you would like to exchange with the network but to exchange with the network you must first know where you are and you are not allowed to transmit until you know exactly where you are so you must have some way of detecting where you are before you even open your mouth therefore this is a this is a challenge we are actually trying to tackle with team in Grenoble we have a so basically we're trying to develop a small software application that will run on the device that scans what are called signals of opportunity so basically cellular carriers TV broadcasting like this because or chip can actually scan this and from from the list of frequencies that we detecting determining the probability of being in the u.s. in Europe etc so this is very much research still but I'm quite up full that by the end of this year we can have something to demonstrate on that so and that really would be an enabler of like transoceanic like constant tracking or thing like that things that have to move from one continent to another this is gonna be a little bit technical but it's not gonna last more than five minutes it's about the security problem so I hate the the the security fear factor of the moment but the problem is if we don't demonstrate a good level of security people won't trust us so what where this is really where thing we're going as an ecosystem is I think all of our system whether there's you know system and chip modules simplex chip whatever that will all come with already embedded at fabrication a secure key and and in a unique device ID and that this module or silicon will come bundled with some key management services in the cloud so let me explain how this works basically you start with module or a system and chip or chip in which at fabrication we put the unique serial number and its root key and this is also replicated in a secure server which we call the joint server and that can be operated by you know gemalto Amazon whoever microchip any company basically your you can operate one then basically this gets pushed to distribution it's comprised by a device manufacturer the device manufacturer manufactures device and push the device devices in the in the sales channels those two guys device manufacturer the device man should never see Sookie doesn't have to manage any security at all in the loja alliance we have defined standard interfaces between this joint server which is basically a secure key storage server and network server for example TTN and application servers as soon as the device turns on and start broadcasting its join request packets the the key derivation process happened both in the device and in the secure server so the device is deriving these two keys the network key and the application key joint that's the same and both the network server and the application server through the standard interface that we've defined in the Lorenz they have a standard API they can go to join River and they fetch the network server at the network key the application server the application key so the network server has never seen the application key therefore you have end-to-end encryption nobody has ever seen the root key and even actually the user of the device might not even be exposed to to the ricci the root key is embedded in the secure element in the device and never goes out and never accessible so and even if you actually physically break down the device you can't extract it so I think really this is a streamlined process we'll try to - you know collectively to put into place - really is the provisioning process and that you know the the manufacturing ecosystem doesn't have to care about security this is all handled some news we for less humans we've been working on a new low have one very compact low a one stack for the end of Isis that we've called Mini Mouse because it's small and it's pretty it will be coded it is actually coding in c++ c plus plus so to be extremely modular it's it's super small it's six kilobyte without the encryption it will be fully collaborative so we'll push that very soon probably next week on github and the community is actually welcome to contribute so the ID there is a Sam tech will contribute just the initial version and the class a EU version that's comes to 35 next week and then it will be open for you guys to contribute other regions if you want to contribute Class B etc etc so we have well documented all the interfaces we have a clean API it's very modular and designed to be collaborative it's nonetheless although it's open source is designed for industrial applications so from scratch we have integrated watchdog fail-safe it can backs up its state in the flash to the basal even if you reboot the device it restarts from white left so if your device crashes actually the network will never even notice it so basically very nice features the basically also will try to reuse a 1.32 class a version very soon littered and basically the way we we will try to ensure the quality of this that it remains good quality this is very new is we will run overnight and over-the-air certification tests on the codebase so every day every day at midnight we'll take a snapshot of the github code or compile it push it into a real device that device is hooked to a real gateway we've run the certification tests on this setup and every morning you have a dashboard that basically says what you've broken ok this is another super important topic for the ecosystem I upgrade over the year so clearly we're starting to feel the pressure on this so this effort is led by the firmware upgrade task force inside the aloha Alliance I'm part of this task force we targeting the download of binary files of big binary files or big few kilobytes file into devices that could be either a fulfillment image or just a patch we've worked on the concept of a patch and I very much believe in this because I don't think we can download the full film well but actually patching just a function is super useful so what will be released by the Alliance all the LCD will include basically code compression integrity and authentication using public private cryptography and it will come with an open source actually several open source boot loaders that can basically perform the decompression and installation of a patch so basically it'll be complete toolbox that you can use to get firmware upgrade functionalities on your devices of course this will be open source you can modify it any way you want and adapt it to your device how does it work so basically up to now we add the lower one stack we are on top an application layer we have defined the concept of packages application packages so this is how we were going to progress forward from now so basically we've defined inside the Alliance for standard packages one is clock synchronization it's very useful to just keep the clock of your device in sync with the network and that's actually to actually to to set up a rendezvous for the film I upgrade so you need all your fleet of device to turn on the receiver at the same time so this is what is the clock synchronization package is about we have multicast setup of the year it is how you create a multicast group from just nothing you over-the-air set up a multicast group for a thousand device that you want to upgrade all in one batch fragmented file delivery is a clever file fragmentation mechanism that requires extremely little exchanges between the network and the device so it uses a clever error correction code you push fragment to the device and even if the device loses 10% 20% of this fragment is still able to recon to reconstruct the full file so basically you don't have to acknowledge every single fragment you just push fragments and when the last device Rises the sensing ok I've got enough you can stop this is that's gonna work and the device management package which enables you to remotely install this new package verify the validity of this package install it install that ephemeral gray and also control the reset remotely reboot your device that can be useful actually for in many other circumstances and just from our great this is another very interesting and rnd project I'd like to share with you because Laura is so assymetrical we've got lots of uplink but few down links we also have a very often of requests from customers I'd like a no missing point curve so what I have a tracker for example there I'm driving from chrono ball to Lausanne and I like to ensure that at the end of the day there are no missing point on this road it's supposed to send packet every five minutes but I've got some packet error rights let's radio right and I don't want to acknowledge every single packet because because that's too expensive basically and I don't have the danly capacity to do that necessarily so we're coming with this streaming coding library that that we we working on the moment that basically puts a little bit of redundancy on every frame with again error correction code said that if you have gaps in your data the follow-up data's allows the server to reconstruct the missing points basically so from an application standpoint advice you just push data the library add a little bit of redundancy on each of that and the gaps get filled in by the server as the packets arrived so a few minutes hours later you haven't you have no gaps anymore and you see the full data of your sensor that would work exactly for a temperature sensor you want no missing point that would work for everything pretty much and that doesn't require any downlink at all it's it's purely uplink only it's one way only so I think that's a very clever way of of solving this this kind of lack of systematic acknowledge that you could have in a lot of lp1 networks and we'll try to deliver something again by the end of this year energy harvesting so that's a very big challenge we have to solve having batteries everywhere is very bad ID for many reasons the I think all the components are there to make energy harvesting happen but we are people are very reluctant to use it because of the uncertainty is I never know if the device is going to be operational not does it have enough energy so how do I work with a device that might be on or off that's really big challenge actually so I I really think that the hardware part of this can already be solved but what we are really looking for as an ecosystem is a software reference design you know software stack on the end device and on the cloud side that can basically adapt the behavior of the device depending on how much energy is left in there and still provide valuable service even when there is no energy so this is really a big challenge to be solved for a wide adoption of of energy harvesting I think there the challenge is mostly on the software side I'd like now to talk about new network deployment that we're seeing and the trend there we are clearly migrating from you know big macro cell over-the-top deployments to what I call ibrid networks so more and more we're seeing a combination of macro cells and lots of femtocells so the femtocell start coming in many many different flavors you know that this is the same tech reference design on bottom it's just like this very small USB key it's going into DSL modem cable modems it's going into din rail from fakirs is going to Wi-Fi routers so we start to have available hardware to put basically Aloha one gateways everywhere so we very often challenge on capacity of network and I'd like to share with you the result of a real-life experiment has been performed by machine cue it's a it's basically a Comcast spin-off in the u.s. in Philadelphia unbelievable results you'll see basically on a 400 meter by 400 meter area in Philadelphia they've deployed some indoor femtocells in there in there basically modem cables so the blue dots here are some femtocells and you know and then they have done some real loading real-life testing of the capacitor so they have deployed 200 sensors and then why steps of 12 power just increase the throughput increase increase so after it on the last 12 hours what you see there corresponds to 1 million packets to a 24-hour on point 16 kilometer square and it's only using 8 Pico cells all right at the same time they achieve 96% packet success rate so we basically losing 4 percent of back to everything including interference collision everything early we losing 4 percent of the packets whereas we loading them the network 1 million packets on basically a point 16 kilometers so this this basically corresponds to 125 megabyte today kill meter square and every packet is just 20 bytes and all devices in this experiment or indoor so actually even in very dense urban environment like what you've seen we've very far from from network capacity as soon as we start deploying tend to sell so typically in the TTN deployment model which is actually only femtocell based more or less densifying the network gives you nearly unlimited capacity so what are the advantages of this ibrid networks basically they have virtually unlimited uplink capacity they provide longer battery life for the device because in average data rate is x forms therefore my energy consumption of the device divided by for mathematical and they also increase a lot to downlink capacity because I have a lot more Gateway I can push my downlink through so I've got better downlink capacity so I can now enable new services like VMware upgrade over the year or real-time downlink control you know closing your valve asking you device position fixed on demand etc having cybron network is super cool but we see that we already see that actually lower one is going to be a constellation of small networks and we need to find a way to actually interconnect this so that it behave as just one big network so inside the low Alliance we have defined end over and passive roaming interfaces people are defining roaming AB that's very good but it's still very peer to be oriented and those interfaces are quite complex to actually operate and create and mount and maintain so I'd like to introduce today the concept I've been trying for a long time but push it in the wide open today this is really for reflection nothing has been yet of a packet broker infrastructure so what would that be it would be just a global publish/subscribe platform on which any network can through a very simple API can publish all the packets that off from device is not belonging that it works so if I receive a packet that I don't know what to do with I can just by default publish it on the broker and then somebody can come and subscribe this dev address and say basically I want to this sensor on the packet broker so the packet worker would basically perform the authentication basically you know is that a valid packet does it really belong to you and route that over to to the owner and it will also maintain kind of a balance okay so that we absolutely know monitor exchanged inside the broker but the broker would maintain the balance that I proposed to express in time on air it off in seconds of Laura over the air so you you credit you once again I'm charge one second because you've routed the packet for me okay so for for broker to be actually successful it must be monopolistic that basically mean if you have ten brokers they use less so you have to have one broker and the condition I think to have only one broker that everybody can subscribe to is that it must be neutral so it must treat equally all players and everybody should be able to join given a certain size it must be transparent so that basically mean the strategy and technical choice made there must be agreed by by all members so it must be some kind of association if you want I think and it must be free meaning you publish/subscribe operation on this worker are free it doesn't mean that actually the operation of this wing is free of course but it could be offset by just membership fee or something like this but that would just pay for the maintenance of the infrastructure so the more I talk about it with you know people and a peer-to-peer base and now to you guys the more I think it's a nice concept but but there are many challenges to be solved first capacity a very crude calculation very fast indicated that you can have up to alpha million published operation per second quite fast that's a lot to handle so it must be a totally distributed architecture how do we authenticate device and make sure that we route them only to the legitimate owner of that device because else we go into some privacy problems what happens when multiple networks and I will Apple all the time when multiple network forward the same packet you know that packet was received by orange and by TTN and by Wigan friends who gets credited do we share is that Leonard an auction you know so many things possible so we need to find a solution should we actually have a reliability rating on the network like on on uber you know yes it's a good very good and good network this is a bad network should there be a minimum size to become a player many many different things we have to we have to do but I really think this is a way forward to interconnect private and public network together and to make them appear as like one big network for the user and that is this is the condition for explosion of the ecosystem basically to make it transparent to the user finally even if you have this bracket broker some places will never get network coverage so there will always be isolated size you'll always have cows in the middle of Brazil or Argentina or actually animals in the French Alps that you want tract so my my pet project at the moment my hobby is satellites and I'd like just to introduce a talk that's going to come after by Tomas so I'm working with a moment on some things that could provide some form of connectivity to this isolated site you will describe all the constraints going with excetera so finally I think as I mentioned it's not so much about Louis is really all most of the innovation now is actually around law there are so many opportunities there for developers or people that have creative ideas to go around that because actually lava has just created something that never existed you know you you don't play with Laura's the same way that you play with a cellular connectivity or or Wi-Fi it's it's low power it's low cost but it come with this constraint and advantages so we have to basically adapt around us so I really encourage you to think out of the box and and basically I think the winners will be the one that actually really rigea quality of underlying qualities and can and can provide basically the the best service with with what's available so I wish you all a lot of success and I'm super pleased to be there today with you with all of you [Applause] so please stay with us because we've got ample time for some questions theories philosophies remarks anything you would like to share with Nikolas I'll just start up you say think out of the box right the take-up of Laura one was quite rapidly so there might be all sorts of use cases you never thought of so what is according to you the most funny use case you ever bounced up against actually one of the biggest use case and most the one we had never anticipated is a connected mousetraps oh yeah we'll have a presentation on that yeah and it actually happens to be a very serious use case with a very bizarre design constraints yeah so it's I wouldn't say it's fun but it's quite unusual I also like the phrase end to end solution but if you look at your own invention since it let's say the last four years but what are you most proud of yourself because that this was about the challenges but able to get it back to you again what were you most proud of beings are today actually it's really [Applause] which is a big thanks to you the community yeah I never you know never thought this would become this big really so big yeah yeah actually when we invented lower the IOT world didn't even exist no so we've gone a long way this made IOT fly so if there's a question here in the front the sort of centralized packet broker idea I'm talking about for data integration between different platforms do you think you could utilize multi network geo lock in that case given the the gateways that received would all be in differences very good question yes I think we could so basically when you publish a packet in that broker you could come with a lot of meta tags or a meta information and clean this information you could have signal strengths and timestamps and basically there's nothing preventing using packets coming from several gateways on different networks to to perform a global more precise Sherlock I think that's the goal actually thank you very much you're from the University of Edinburgh right okay any other questions yes zero for that so let me see if I can reach you yeah sorry [Laughter] so Laura makes use of chirp spread-spectrum modulation so it's these chirps you transmit you the inventor of Laura way did you get this idea of using these chirps because this is a unique thing of the Laura modulation and where you're from I'm mom JP Myers I'm the founder of DT and mapper I'm from South Africa okay thank you for being there it's a long travel so where did we find okay so to be very honest I didn't have the idea of the chirp okay my friend did and it was actually from from radars so radars also using chirps and we were looking for a waveform that has some specific mathematical property and actually you have only two existing the Dirac and the chirp and the church with the chip we had in our hands to play with was the only thing we could do so that's how we started with this but the the the inspiration was from the radar systems one final question from the audience please allow me to soon well the three news chips came out from the Essex family what will be their impact on the use of robots so you're talking about 1261 so basically this chip I don't have any else a revolutionary feature they just designed to be actually thinner smaller lower power and to to to unable to build cheaper lower one modules so basically is that really as the volume is ramping up there is a pressure from the ecosystem to make smaller devices lower power and cheaper yeah so essentially it's it's an ends were just so there are a few Eden feature in there I would like yeah but they're not disclosed yet oh okay soon to come but there are a few tricks in the silicon but not in the open yet okay yeah so yes improvement on all fronts and a little Easter egg in it and a little bit there exactly finally some folks here but most of him did at the lower one cause they they will receive a certificate will he be around to sign it will be kind of cool you know you are way too numerous you know okay so it was the first time I ever came with loud music in a room it's the first time I'll sign an autograph it's all due to you so finally a very big applause again [Applause]

Channel: The Things Network

Views: 22,915

Keywords: LoRa, LoRaWAN, iot, internet of things, ttn, the things network, nicolas sornin, semtech, tech, future, technologie, learn, understand, learn about lora, learn about lorawan, lorawan alliance, understand lora, understand lorawan, what is lora, what is lorawan, what is, The Things Conference, Conference, Picocell, geolocation

Id: jNnPTxWRNxs

Length: 33min 28sec (2008 seconds)

Published: Mon Feb 12 2018