Vous êtes sur la page 1sur 11

So, you want to build a connected sensor device? Here's what you need to know.

-- A collaborative resource -The purpose of this document is to provide an overview of infrastructure, options, and tradeoffs for the parts of the data ecosystem that deal with generating, storing, transmitting, and sharing data. In addition to providing an overview, the goal is to learn what the pain points are, so we can address them. This is a collaborative document drafted for the purpose of discussion and contribution at Sensored Meetup #10.

Figure 1. The data ecosystem, a simplistic representation. (Image credit: xkcd.com)

Outline
Outline

Speakers Audience Motivation Topics of interest Challenges Who else is working on this? Consortiums, groups, etc. Related meetups and community groups Other useful resources (blogs, websites, etc) FAQ Infrequently asked, but related questions Other notes, comments, feedback Chris Jefferies Talk Michael Kosters talk Tanteks Talk Duppys Talk Open Q & A Session with all the speakers Open Call for Announcements

Speakers
David Proctor (@duppy) Michael Koster (@michaeljkoster) Tantek elik (tantek.com @t) - https://etherpad.mozilla.org/sensored Chris Jefferies (Chris@tinajalabs.com, @tinajalabs) tinajalabs.com shared slides: Sensors, Data, and Application Logic Rachel Kalmar (@grapealope)

Audience
Anyone working on sensor devices or applications.

Motivation
People working on sensor devices come from many different backgrounds, and many of us are new to this field. What do we need to know? One critical issue is about data accessibility. If your hardware has an outgoing data stream (e.g. say, a fitness sensor devices), how do you decide how much data to make available to the user or developers? Ideally, you'd provide access to all data. But how do you remain competitive if you allow others to build competing software using data from your device?

Topics of interest
How do I give people access to data from my device? Downloads APIs Push formats Google Drive as an API endpoint, then share access https://googledrive.com/host/0B0Ubue0bd6FfRkZmSGZLNV9Jdnc/collecti ng.html Data formats Platforms, API aggregators, etc. Protocols 802.15 BLE ANT+ WiFi Working standards Privacy issues Available Libraries Prototyping Resources

Challenges
Lack of discoverable API definitions Discoverability of sensor attributes Discoverability of actuator attributes The formulas to make meaning from attributes (i.e. x[t0-tn] accelerometer movement = y calories burned). Unclear location of rule sets to manage interaction of sensors/actuators; at the sensor, in the gateway, on the web, all of the above The autonomy of actuator behavior in the face of lost connectivity. How redundant should the rules be? How do you monetize your system? How much is the data worth? How much should you charge? How do you make meaning from the data?

Who else is working on this? Consortiums, groups, etc.


Internet of Things Consortium - http://iofthings.org/ IEEE 802.15

http://www.comsoc.org/ Body Area Networks http://www.ieee802.org/15/ http://www.ieee802.org/15/pub/TG6.html 802.11 (WLAN) Bluetooth Specifically the Bluetooth Low Energy (BLE) portion of the BT4.0 spec (there are many profiles of a specification) http://en.wikipedia.org/wiki/Bluetooth#Bluetooth_vs._Wi-Fi_.28IEEE_802.11.29 ANT http://en.wikipedia.org/wiki/ANT_%28network%29

Related meetups and community groups


Hardware Startup Meetup (everything hardware) Open Source Internet of Things Meetup San Francisco Internet of Things Meetup Rock Health (many resources about health data)

Other useful resources (blogs, websites, etc)


Quantified Self website/meetup/FB group -- many interesting conversations here about, but not limited to, data Singly -- API aggregator http://quantimodo.com

FAQ
Whats the difference between open data and open standards? Does open necessarily mean free?

Infrequently asked, but related questions


Where should I store my data? http://quantimodo.com If I want to buy a NAS (network attached storage), which one should I get? Build your own with FreeNAS

Other notes, comments, feedback

Chris Jefferies Talk


How can the data be discovered? Is the data stored local to the device, out on the cloud, or elsewhere? * Idea on your dilemma on where to store logic: Why not model biological systems, such as central nervous systems that control local (e.g., retina), central (e.g., neuron) and relay nodes (e.g., thalamus)? Scale it your complexity, depending from single cell org to human brain. What APIs do you make available? What do you do with the data when youre not online? Do you own the data? Chris project is about wiring his house up in a detailed sensor network. The actuators come later. He routes data through many web services: Cosm, ThingSpeak, Open.Sen.se, Tinaja. The Autonomy of Actions: What happens when the communication breaks down? We want a short circuit between the sensors and the actuators, so that if the web connection breaks down, things still happen. The cloud is dandy, but flawless access to it cant be guaranteed. You dont want your house to fail to turn the heat on or off whenever the net connection gets flaky. Additional notes from Chris Jefferies: In 2007, attended an un-conference, the Data Sharing Summit, where we discussed, among other things, The Bill of Rights for Users of the Social Web There is a programming pattern (Gang of Four) that says program to an interface, not an implementation. This is why we want to build strong APIs in which the implementation can change, but the API definition should not. If the interface changes, it should be revisioned into a new method, leaving the original intact. This helps from breaking existing apps. Live chart from Cosm - typical feed from Cosm. Data from my sensor network; TMP36 sensor -> XBee transmit -> XBee receiver -> Raspberry Pi -> python -> Cosm

Michael Kosters talk


On: The Open Source Internet of Things. Whats the difference between M2M (machine to machine) communications and IoT (the Internet of Things). In M2M, you often have top-down control. A single company can regulate all the devices theyre wiring together to make sure that they can talk. In IoT, youre integrating between devices and protocols that dont have a body regulating them and making sure they play nice. So you need agreed-upon standards on how devices talk, in order to allow IoT communications without there being a company over it all. You also need a shared data model, so you know what values leaving one system and coming into another mean. CoAP (from the IETF) is an effort in the direction of allowing IoT communications. Semantic Sensor Network (from the W3C) is another standard. When evaluating communication technologies, beware restrictive TOUs (Terms of Use) that dont allow you to use your own data. Do you have to have a SIG membership or be a representative of a company to work with the standard? Just because something is Open Source it may not be truly open in how youre permitted to use it. Theres a spectrum of how open standards are. Its not just fully proprietary vs. fully open. In the middle there are industry alliances who agree to allow each other to use their patents, but dont welcome outsiders. And there are standards bodies that try to make open standards, but allow proprietary businesses to slip proprietary licenses into the schema as long as the companies providing them promise to allow them to be used (mostly) freely.

Tanteks Talk
etherpad.mozilla.org/sensored Hes a Mozilla Standards guy. He wants us to start thinking about: Longevity Standards/Openness Building Blocks

One of the problems of proprietary APIs is that the API owner can change it on a whim, causing software built on top of it to break. Does anyone have any electronic devices on their person in this room thats more than 2 years old? Devices may be disposable, but your data shouldnt be. Why should I put any data into a service when young companies doing interesting things like youre doing always just get acquired by bigger companies who promptly shut down half the stuff you were pouring your data into? When setting your data up to survive transitions through who knows what comes, the smaller and simpler your data format, the more likely it will survive. Simple text files are near immortal.

Duppys Talk
Why do we care about the data? What do we want to get out of the data? What we want is for things to be talking to each other and doing stuff without us needed to manage everything all the time. What's cool is getting people out of the middle and letting the devices talk to each other

Open Q & A Session with all the speakers


Q: How do you incentivize companies at every layer (sensors, etc.) to buy in to open data, and to making data available for the long term. A: Kostner: ~People at these companies are starting to realize the value of connecting these devices together at a high level and are asking for standards, despite the perceived cost. Chris Jefferies: If people start realizing they dont own the data, people will stop supporting them. ... If people start demanding to have their data the way they want it... Rachel: Were heading toward the Apple Internet of Things vs. the Google IoT, etc. We want there to be only one. What we really want to be able to do is to be able to take in multiple streams of data [to do something interesting]. Tantek: We need to create an Ecosystem of Open Data. Once youve done that, it lowers the cost of building into the open ecosystem rather than building 1,000 unique silos for each of 1,000 companies. Mozilla is making FirefoxOS for cheap/nearly free devices. He thinks small devices

will lead the way in promoting an open ecosystem. For expensive devices like smartphones, companies can budget to build a walled garden. For little devices made by little companies, sticking to standards lowers the cost of creating the device. Rachel: (regarding Tanteks comment that things that fit in your pocket will lead the push for shared standards.) I had to have a custom skirt made to carry my sensors. Guys have pockets to put their sensors in. Girls clothing doesnt. I had to have this skirt custom made to have pockets. Tantek: So we need a fashion disruption as well. (laughter) Duppy: Kickstarter and its like give people the chance to show their support for companies building products on open standards. Q: Is there a specific language people should be writing open data sharing standards in. A: ChrisJ and Duppy basically say that they dont want there to be a single scripting language so much as they want open sharing of scripts that people or automated systems can then compile. Q: How do services like IFTTT fit into this idea. A: Tantek: I think its a great system for prototyping, but at the end of the day, its another Single Point of Failure. Duppy: Its good... but: Polling for data is made of fail. It needs to be live data. Great way of showing people how you can do cool things, PubSubHubbub is a protocol for sending and subscribing to updates of data. Pu.S.H. is the acronym. https://code.google.com/p/pubsubhubbub/

Q: [A restatement of the question, how do we create an ecosystem of open data. How do we make it so that companies build into this ecosystem, and lay the foundations so that something amazing gets kickstarted and becomes the next killer app?] A: Duppy: My model is(Information Knowledge Wisdom) - and you cant do anything with any of those levels if you cant GET at the data. Q: Most consumers dont want to see the raw data, but early adopters do. We, as early adopters have a disproportionate voice, because the next wave of customers listen to us. Is there an organization that lets us speak with a loud voice? A: Rachel: I see the opposite happening, where the companies say theyre not developing for the

early adopters. Misfit is excited about being really open, and Im actually the one saying lets be careful how we do this. How do we do this in a way that serves everyones best interest? Tantek: I dont understand this fear youre talking about. Rachel: If we say hey, you can access all my data in real time, youre essentially commoditizing the hardware, and then you have a very different economic model. Audience: Maybe you should write good apps on top of your hardware then? Rachel: Yes, but we have limited resources as a start-up to put into making awesome hardware vs. making awesome apps on top of it. Tantek: If you have limited resources, dont you want to widen the search for the killer app? [i.e. let other people build on your stuff.]

Q: If Nest opened their data, we (a competitor to nest) would build a thermodynamic model of houses and make our next thermostat better than Nests and eat them. A: ~discussion of more stuff than I caught or kept up with~ Q: In the discussion of the ideal case for users, is the hope that with completely open data, people will build awesome things themselves? A: Kostner: We have a failure of business models, where people havent figured out how to make money on open systems. Rachel: We dont want another MakerBot, where people start off open and then, after getting a pile of users, become closed. End of Q&A.

Open Call for Announcements


Next months Sensored theme will be all about demos. (In prep for MakerFaire)

Tia anounces a hackathon via Technology Innovation somethingorother. Hertz car rental is sponsoring it, and wants to get people helping them get involved with car share stuand have a fleet of hundreds of thousands of cars to play with. IoT conference - theyre doing one of those, too. Adam Ellsworth: Announces Crowdfunding meetup. Discussion of videography, photography, manufacturing, and shipping. May or June starting. Talk to him to get on the list for the first meetup. HealthTech Forum: Like Health 2.0, but better. Having a CodeAThon this weekend. Hosted at UCSF. Check out their Meetup. If anyone wants to hack a system that has 100,000,000 sensors in it... were working on hacking the human body. At HackerDojo. The human body has a terrible bug where it quits after about 80 years. healthextension.co 7HackerDojo also having other data sharing meetings.

On the matter of openness of APIs http://news.cnet.com/8301-1023_3-57556103-93/peoplebrowsr-wins-temporary-injunction-again st-twitter/ "We relied on Twitter's promise of openness when we invested millions of dollars and thousands of hours of development time," PeopleBrowsr CEO Jodee Rich said in a blog post. "Long term supply is essential as this industry matures. We made this application to ensure full unrestricted access to the Firehose for our Enterprise and Government clients."

Vous aimerez peut-être aussi