The Internet of Things Prof. Anurag Kumar Department of Electrical - PowerPoint PPT Presentation

The Internet of Things Prof. Anurag Kumar Department of Electrical Communication Engg. Indian Institute of Science, Bangalore anurag@ece.iisc.ernet.in

Cyber Physical Systems (CPS) • Engineered systems comprising dense embedded smart sensors (and even actuators) Motes: Smart Sensor – in some physical domain (e.g., the environment, Devices large buildings, farms, public utility systems (water/power), networked health care) • with distributed computing, signal processing, and inference, – thus providing unprecedented visualisation • and even control and actuation (fine grained), Agriculture • all interconnected over a communication network (with usually a substantial wireless component) Instrumented Cities

Application Domains Wireless • Smart and green buildings Blood – Structure, energy, environment Oxymeters • Networked healthcare – Mobile patient management, geriatric care • Smart Cities – Transportation, pollution, etc. • Agriculture • Smart power grids • Water monitoring and management • Forest and wildlife Forest Fires Endangered Wildlife

Monitoring at the Extreme Ends of Life • Core body temperature – Important health metric for neonates • Remote embedded monitoring of neonates in rural homes • Very old people living alone • Wirelessly connected sensors attached to common objects that an elderly person uses • Analytics algorithms analyse the data on the cloud • Change in use pattern could indicate a problem Courtesy mylively.com

Smart Buildings http://www.digikey.com • Sensors and actuators in appliances are wirelessly connected • Algorithms analyse usage patterns, grid condition, and power pricing – Optimise the use of energy

Devices Per Person Mote Cisco IBSG, April 2011 Culler and Estrin Devices: • In automobiles • Communication devices • Computer and network peripherals • Household appliances • Personal care devices and equipment

The Internet of Things (IoT) Link speeds Resource Challenged 100 Kbps 100 Mbps Resource Rich 100 Gbps Shelby and Bormann 2009 The platform for the emerging Some have begun to call it Cyber Physical Systems (CPS) “The Internet of Everything”

IoT Technology

Networks in 1988 • LAN and WAN packet networking were still evolving

Networks Now • Device to device (D2D) communication • Ad hoc wireless networks will finally come of age

A Smart Wireless Sensor Node • A smart sensor node is popularly called a mote (“speck of dust”) The Berkeley • Sensing: temperature, chemicals, Mote light, infrared, biosensors, strain, sound, vibration (often using The TelosB MEMS technology) Mote • Processing: e.g., 16-bit, 8 Mhz, 48KB flash, 10 KB RAM with a A Mote simple OS with a PIR Sensor • Digital radio: e.g., ISM band; a few Array 100 Kbps • Battery: e.g., 100mAh “button” The Vigi’Fall batteries to 2000mAh (2 AA Fall Detector batteries) Mote

Getting a Multi-Year Lifetime • Today's devices: active : 5 - 10mA; sleep : .001mA • 1000mAh battery; multiyear lifetime ⇒ 1% active A wake-up • Devices need to alternate through sleep- radio wake cycles (ESE, IISc) • Future devices: active : 0.1mA to 1mA; sleep : .001mA • Energy scavenging: – Nodes can draw power from their microstrain.com environment, – Using appropriate devices or mechanisms – e.g., from ambient light, vibrations, or mechanical use • Need software and algorithms that further conserve energy A self-powered wireless light switch (ESE, IISc)

IoT Communication Technology Evolution • 1980s and 90s: Wireline industrial automation networks (CAN bus, HART, Fieldbus) • 1995 onwards: research in wireless sensor networks – Predominantly academic research • 2003: IEEE 802.15.4 PHY/MAC standard (“ Zigbee ”) • Industry gets interested; adopts the IEEE 802.15.4 PHY – 2007: WirelessHART • Frequency hopping TDMA over the IEEE 802.15.4 PHY – 2009: ISA 100 • 2007: 6LoWPAN RFC4919 (2007) and RFC4944 • 2008: IETF ROLL (Routing Over Low power Lossy links) working group formed • 2009: Zigbee Alliance adopts 6LoWPAN and ROLL

The Network Protocol Stack • IEEE 802.15.4 PHY at 2.4 GHz • Mesh support Sensors • UDP over IPv6 with 6LoWPAN adaptation layer Sensors Wireline Local Sensors Area Network and the Internet Application UDP IPv6 IPv6 6LoWPAN 6LoWPAN IEEE 802.15.4 MAC IEEE 802.15.4 MAC Ethernet MAC IEEE 802.15.4 PHY Ethernet PHY IEEE 802.15.4 PHY

PHY and MAC • IEEE 802.15.4 PHY over 2.4GHz (ISM band) – The most popular PHY standard • 16 channels over 80MHz – Each channel of 2MHz, spaced 5MHz apart • 250Kbps bit rate – Achieved by a spread spectrum modulation – 2000 chips per second – 62.5 symbols per sec., 32 chips per symbol – 16 PN sequences code 4 bits/symbol • Thus yielding 62.5 x 4 = 250Kbps TI’s Single Chip • Medium Access Control (MAC): PHY/MAC for IEEE 802.15.4 – CSMA/CA – TDMA for time critical applications

Packet Error Rate vs. Received Signal Strength 103 Bytes over the air (13 B header, 90 B payload) • CC 2420: receiver sensitivity of -88 dBm achieved – Noise floor -110 dBm; processing and coding gain: 10 to 11 dB – Link lengths of 10 meters indoors and 30 meters outdoors • CC 2520: receiver sensitivity of -98 dBm – Also higher transmit power (5 dBm) – Link lengths of 100 meters outdoors are achievable

IEEE 802.15.4 CSMA-CA MAC Performance • Example: 70B packets – Packet error rate 1% • Can support a few packets per second per source • Non-negligible packet loss probability – There is no TCP • Application level resilience needed • Not for applications with tight delay/loss requirements

Layer 3: IPv6 and 6LowPAN • IPv6 can support a large number of uniquely addressable devices • 6LoWPAN: an adaptation layer Sensors – Fragmentation and reassembly – Header compression – Mesh routing Sensors Wireline Local Sensors Area Network and the Internet Application UDP IPv6 IPv6 6LoWPAN 6LoWPAN IEEE 802.15.4 MAC IEEE 802.15.4 MAC Ethernet MAC IEEE 802.15.4 PHY Ethernet PHY IEEE 802.15.4 PHY

Routing over Low Power and Lossy Links • Unreliable and time varying links – Short term variations • (coherence time) – Long term variations • (e.g., seasonal variations) • Need to find routes in this resource challenged setting • Why can’t MANET protocols be used? • Link state protocols (e.g., OLSR) – Very high in overheads; hence not energy conscious – Specially when link states change frequently • On-demand protocols (AODV, DSR) also not found suitable • IETF’s ROLL working group – Routing Over Low-power Lossy-links

RPL: Routing Protocol for Low power lossy links • Based on dynamic shortest path • Bellman-Ford type algorithm • Flexible notion of link cost – E.g., average number of attempts needed to send a packet over that link • Nodes are dynamically “ranked” in terms of their relative costs to the sink • This partial order yields a directed acyclic graph • On which RPL finds a routing tree • The cost of each link is constantly updated • The tree, thus, changes over time as link qualities change

Some Experience with RPL SmartConnect • Network was designed with two guaranteed paths – Delivery probability within a delay bound • Static path routing does not exploit other paths that appear over time – 70% delivery (was the target)) • RPL is very dynamic – Link metric: packet loss rate – Exploits all available paths 100 pkts per source in 25 minutes – Can have large convergence times 5 days of continuous data

Network Computing 𝑦 1 𝑦 3 𝑦 2 𝑦 𝑜 BS • Signal processing requires computing functions of the data – Say 𝑔(𝑦 1 , 𝑦 2 , 𝑦 3 , ⋯ , 𝑦 𝑜 ) ; e.g., average, max, etc. • Send all the values to the base station (BS) – Communication complexity 𝑃 𝑜 2 – A common approach in simple low duty cycle applications • In-network computation of, say, max – E.g., max( ( 𝑦 1 , 𝑦 2 , 𝑦 3 , ⋯ , 𝑦 𝑜 )) – Communication complexity 𝑃(𝑜) • Need simple distributed algorithms • Resource challenged nodes; lossy and intermittent links • Distributed clock synchronisation, function computation, optimisation, signal processing, tracking, etc.

Complete Application Architecture Sensor Mote Application Dashboard Sensor Analytics in the Mote Cloud Wireline Local Relay Sensor Mote Sensor Area Network Mote and the Sensor Internet Mote • Databases Relay Sensor – E.g., patient records Mote Relay • Analytics and inference Relay Relay Sensor Relay Mote – Is the patient Sensor Sensor Mote experiencing an episode Mote Sensor Mote Sensor • Actuation Mote Sensor Mote Sensor Mote Sensor – Send the doctor an alert Mote Sensor Mote Sensor – Activate some Mote embedded actuator

Challenges for IoT Applications • This talk has focused on communications technology for IoT • There are many other challenges • Building provably correct and reliable systems from a large number of resource challenged embedded sensing/computing devices • Designing such systems to meet real-time performance objectives (difficult with wireless interconnections) • Control over resource challenged networks • Security and anonymity • Development of common middleware that can be reused across application domains

Whither IoT?