The Internet of Things (IoT) and Analytics Class 1: The Different Things of the IoT March 7, 2016 Louis W. Giokas
This Week’s Agenda Monday The Different Things of the IoT Tuesday A Look at Communications and Devices Wednesday Cloud Storage and Formats in the IoT Thursday Examples of Big Data Analysis Friday Machine Learning & Analysis Techniques
Course Description The IoT generates a vast amount of data. This data can be used for many purposes, from product design, service and support, marketing, and control. There are three levels of devices: the things, communications infrastructure and storage. Tying it all together are analytic techniques. In this course, we will build from the bottom up and then look at how the analytics infrastructure can be used in applications.
Today’s Agenda Course Motivation and Outline Application Areas Consumer Industrial Medical Infrastructure Some device types Class input
Course Motivation and Outline The IoT is very diverse. This diversity extends from the types of things to their applications. The IoT also generates unprecedented amounts of raw data. The use of this data is similarly diverse, as are the consumers of the data. Analytics is how we make sense of the data an either present it to human consumers or communicate it amongst the things and applications that manage things.
Course Motivation and Outline In statistics (or analytics) the type of analysis done is generally dependent on the data available and the “research” questions to be answered. There are many ways to perform each analysis, in general. Requirements driven by the particular situation can determine which are applied.
Course Motivation and Outline Statistical considerations are also critical. One must be able to explain more than the quantities derived. Sources of variation, confidence intervals, applicability and limitations of the analysis are also critical. Probability theory is also key. Many machine learning algorithms will always find a “pattern”, but is this an artifact of the process or real finding?
Course Motivation and Outline We will follow a Systems Engineering approach to bring this all together! Requirements are key! These are driven by what we want to achieve, but… There are also considerations driven by the available devices, their measurement accuracy, update rates, etc. In the IoT world, communications and storage are also key. We can’t analyze what we can’t collect or store.
Course Motivation and Outline You may have heard about the “struggle” going on between “statistics” and “data science”. There is also a current in engineering that posits that we don’t need models, but can get the information we need from the data. Some of these discussions have appeared in IEEE journals. The truth is that statistics covers data science and, indeed, it is essential for real understanding. Data engineering (storage, access and communication) are also important. There is significant overlap here.
Course Motivation and Outline Plan of the course: Start with the things, looking at types and information they generate, as well as available methods of communication. Communications infrastructure is a major consideration for the IoT in many areas. We will look at this in terms of specific IoT implementations and equipment.
Course Motivation and Outline We need to learn how to store unprecedented amounts of data. This means utilizing technologies developed for the scale of the Internet. Once we have this data and can operate on there are two orthogonal sets of considerations. Purpose of the analysis Statistical techniques
Application Areas Consumer Consumer uses span a broad range of uses and devices Smart phones and smart watches/fitness bands Home control Thermostats and smart appliances Home security systems including detectors (CO, etc.) Entertainment
Application Areas Consumer Types of data generated Location Orientation Heartrate Events (e.g., threshold exceedance) Status For a thermostat, temperature and presence, settings
Application Areas Consumer Communication types Fitness band: blue tooth to a cell phone to the “cloud” Direct Internet connections Direct cellular connection Some are one way, some go both directions Collection only vs. collection and control
Application Areas Industrial The things span a number of applications Plant equipment Robots CNC machines Networked monitoring and control equipment “Dispersed” machinery Large engines (e.g., from Cummins for vehicles and power generation) Monitoring equipment Data Centers
Application Areas Industrial Types of data generated and consumed Engine status (periodic, e.g., 5 minutes) Machine actions (events) and status Program in operation Specific movement commands
Application Areas Industrial Communication types Industrial, in plant Industrial Ethernet/EtherCAT Profibus Modbus Wireless (WiFi for industrial control) Fieldbus Many vendor proprietary protocols
Application Areas Medical Medical devices In care facilities In homes In-body (embedded) Facility monitoring Supplies Devices (location and scheduling)
Application Areas Medical Types of data generated and consumed Patient data (heartrate, glucose levels) Dose control information Equipment location Medication and supply level and location Equipment health and status
Application Areas Medical Communication types Ethernet WiFi Cellular Bluetooth
Application Areas Infrastructure Areas of application Electrical infrastructure (utility and private) Water systems Road systems Building systems Gas infrastructure
Application Areas Infrastructure Types of data generated and consumed Meter data Customer premise Flow (for water utilities) Intermediate devices Utility infrastructure (substations, generation, transmission equipment) Structural status Roads, bridges SCADA
Application Areas Infrastructure Communication types Cellular For dispersed meters, engines, etc. Special data only Ethernet over powerline Satellite For very remote locations
Some Device Types Consumer devices Industrial devices ARM based microcontrollers from companies such as STMicro (STM32), Texas Instruments (MSP430, C2000), NXP/Freescale (Kinetis) Industrial devices Intel based devices Quark Atom ARM based devices Cortex R Cortex M
Class Input What types of analytics/control applications are you interested in? Type into the chat. I will try to address these in the last two classes Are you doing any of this now? Examples that can be worked into the analytic framework
Summary and Preview Today we covered the requirements aspects of the IoT that might affect analytics What are the device types? What data do they generate? How do they communicate? Tomorrow we will build on this with a look at how the devices communicate to create a “system”.