Wireless Sensor Networks & IoT: Midterm

exposed terminal problem

(A)~~(B)~~(C)~~(D) *if C->D, B postpones tx->A due to carrier sense, even if collision wouldn't happen

hidden terminal problem

(A)~~(B)~~(C)~~(D) *node B talks to A&C *A&C cannot hear each other *C cannot detect when A->B *if C->D while A->B, collision at B (signal omnidirectional)!!

S-MAC vs. B-MAC

--B-MAC (w/o ACKs or RTS) has higher S than S-MAC --as #nodes inc., gap narrows --B-MAC consumes less power and has less default latency at higher S

established IoT radio technologies

--IEEE 802.15.4 ("ZigBee"): widely used, low-power mesh, lasts for years --BLE: short-range RF, phones/peripherals, beacon --NFC: small readers in phones, large in infrastructure

MANET vs WSN

--MANET=more powerful equipment, higher data rates, more resources --WSN=larger-scale, app-specific, id-centric

types of CPUs

--MCU: gen. purpose, flexible, programmable, low power --DSP: signal processing tasks --ASIC (app specific IC): not re-programmable, high performance --FPGA: field programmable, configured via HDL, array of logic blocks ** flexibility-performance tradeoff** ex: ARM, TI, Intel...

adaptive listening

--S-MAC feature to reduce multi-hop latency from periodic sleep --let overhearing neighbors know tx duration from RTS/CTS then wake up when tx is over --saves energy at heavy loads by reducing latency (inc. throughput)

message passing vs. 802.11 fragmentation

--S-MAC message passing: long message broken into smaller packets & sent continuously after RTS/CTS handshake (inc. sleep time, leads to fairness problems) --802.11 frag: other nodes keep listening; if ACK not received, give up on tx (greater fairness)

adaptive election (AE)

--TRAMA feature --decides which slot in access period node can use (hash function)

neighbor protocol (NP)

--TRAMA feature --gather 2-hop neighbor info using signaling packets --if no updates, signal packets = "keep-alive" packets --node times out if silence from neighbor

schedule exchange protocol (SEP)

--TRAMA feature --node creates schedule interval (SCHED) based on packet production rate --#slots announced to neighbors and winning slots assigned --leftover vacant slots released to other nodes

explicit contention notification (ECN)

--Z-MAC feature --node informs neighbors not to send during its time slot --neighbors rx ECN, set high contention level (HCL) flag --high contention = lost ACKs, repeated backoffs

Z-MAC vs. B-MAC

--Z-MAC has better avg. S and energy efficiency as packet rate increases --Z-MAC has low overhead

channel models

--analog: behavior of wireless channel captured via statistic models (Gaussian, Rice) --digital: model the resulting bit error behavior (Markov states)

IoT protocols

--apps: smart health, grid, transport... --session: MQTT, CoAP, AMQP --routing: 6LowPAN, RPL --datalink: ZigBee, WiFi, BLE, NFC --mgmt: IEEE 1905, 1451 --security: Oath 2.0 O.A.

universal asynchronous receiver-transmitter (UART)

--asynchronous serial --converts outgoing data from parallel to serial and incoming data from serial to parallel --simple, only one line needed --cons: comm. b/w only two nodes, very slow

S-MAC vs. 802.11

--at heavy load (low packet-arrival gap), idle listening is rare & sleep energy savings limited (worse S-MAC performance) --at light load, periodic sleeping is key --w/o sleeping, S-MAC consumes less energy than 802.11

traditional wireless network

--based on infrastructure --base stations connected to wired backbone --traffic relayed by base stations --limits: infrastructure-reliant (disaster areas?), expensive & time-insensitive to setup

CSMA/CA

--carrier sense: station sees if channel is idle and transmits if so --collision avoidance: contention window ensures stations transmit at different clock expiry times (window doubles if collision) --inter-frame spaces (IFS) define priority (high->low: SIFS, PIFS, DIFS)

classification of MAC protocols

--centralized: each device connected to base station that controls access --distributed: each device has logic that decides access --schedule-based: based on time slots & reservation --contention-based: based on CSMA (ex: IEEE 802.11)

CSMA vs TDMA

--channel utilization varies w/ # of contenders --CSMA (channel-based): low utilization @ high contention --TDMA (time-based): low utilization @ low contention

objectives of MAC

--collision avoidance (reduce re-tx) --energy efficiency (avoid idle listening) --scalability (as #devices inc.) --latency (packet delay) --fairness (who sends?) --throughput (data/time) --BW utilization (%time BW used to tx data)

3 domains of transceiver power consumption

--communication: rx/tx electronics & radiated power (energy wasted w/ small packet sizes b/c start-up time comparably high) --data processing (computation): clock cycles, supply V, switching freq. (local data processing >>>) --sensing: DAQ systems (sensors, signal conditioners, ADCs)

5 IoT business opportunities

--components (sensors, wireless radios) --smart objects --systems (buildings, cars, health) --network service providers (ISP) --app service providers (monitoring, analytics...)

B-MAC

--configurable MAC protocol (backoff/timeouts, duty cycle, optional ACKs) --goals: low power, CA, efficient channel use, RF-adaptable & scalable --minimalistic: small amt. of medium access functions (can call on higher protocols)

IEEE 802.11

--contention-based --random access (DCF): prob. that station tx (Markov chains); used to evaluate saturation throughput (S) --optimal backoff window dir. proportional to #stations --with RTS/CTS: S less sensitive to tx probability, ind. of window size, and improved w/ high packet size --cons: fixed data rate & one-hop network

multi-hop wireless network

--device broadcasts to intermediate device; direct comm. limited b/c of distance, obstacles --saves energy

sensor

--device that responds to a physical stimulus and generates processable outputs (measurands) --good sensor: objectively sensitive to measured property --ex: motion, imaging, pos., nanosensors

wireless ad-hoc network

--devices broadcast directly to each other instead of to central access point --challenges: lack of central organizing entity, limited range, battery-reliance, mobility of participants --devices must self-organize (requires MAC and routing protocols)

mobile ad-hoc network (MANET)

--devices move around --routes configured adaptively --complicated in large scale --needs long network/device lifetime & energy efficient protocols

wireless propagation issues

--distortion: rx waveform different from tx (reflection, scattering, doppler fading...) --attenuation: energy distributed to larger areas w/ inc. distance (path loss modeled by Friis eq.); can lead to freq-selective channel --noise: caused by temp-dependent fx in rx electronics --interference: co-channel (sender uses same spectrum) or adjacent-channel (filters don't fully suppress neighbor spectrum)

schedule-based MAC

--establishes tx schedules to avoid collisions at receiver --schedules are adaptive & dynamic to traffic patterns --nodes switch to low-power mode according to schedule

WSN application types

--event detection/classification --periodic measurement --function approx. (temp. map) --edge detection --tracking

signal interference and noise ratio (SINR)

--extracting corrupted waveform causes errors based on strength of rx signal to strength of corruption --leads to bit error rate (BER, % rx bits w/ errors) ***low SINR=high BER (bad signal!)***

why infrastructure-less networks?

--factory floor automation --military comm. --disaster recovery --car-car comm. --personal area networking (glasses, medical device)

on-demand beaconing

--funneling-MAC feature --sink periodically beacons --beacon receivers know they're in the funnel & coordinate schedules (# hops away from sink) --sink can inc./dec. beacon tx power to add/remove nodes from funnel region

physical layer

--generates/detects signals to tx/rx data over network medium --sets data tx rate and monitors data error rates --form of protection for link layer

why IoT?

--growing industry (50 billion devices by 2020) --high funding + business hype --many apps: consumer, gov't, enterprise

funneling-MAC

--hybrid MAC protocol --problem: packet loss rate is high close to the sink (denser node region = funnel) --hybrid TDMA/CSMA scheme inside funnel; pure CSMA outside funnel --mitigates funnel effect in choke points **proves that multiple MAC schemes can coexist**

Z-MAC

--hybrid MAC protocol: combines strength of CSMA & TDMA (high channel eff. & fairness) --owner of time slot always has priority over non-owners --non-owners can steal time-slot if owners have no data --switch b/w CSMA + TDMA depending on contention (low: CSMA, high use, low latency; high: TDMA, high use, fairness, low overhead) --uses DRAND to create scheduling scheme

specific WSN apps

--military (comm, control, surveillance) --environmental (tracking, bio mapping) --health (diagnostics, monitoring) --infrastructure (roads, traffic) --list goes on (MANY APPS!)

9 IoT research challenges

--naming & addressing (b/w many nodes) --power/energy efficiency --things to cloud --miniaturization --big data analytics (35 ZB!) --semantic tech --virtualization (aggregated sensors) --privacy & security --heterogeneity & scale

network allocation vector (NAV)

--provides virtual carrier sensing --tx sets NAV to duration it expects to use medium --other stations count down from NAV to 0 (if NAV>0, medium is busy) --when channel virtually available, PHY is checked --source can send packet w/o contention then send ACK to terminate NAV signal for other users

WSN deployment options

--random (dropped) --regular (well planned) --mobile sensor nodes

demodulation

--receiver looks at waveform & matches w/ data bit that caused tx to send waveform (impossible for analog signal) --problem: how to sync. carrier, bit, and frame (changed signal!)

WSN characteristics

--scalability/wide density range --limited resources/device --mostly static topology --service (in-network processing) --fault tolerance --long lifetime --programmability --maintainability

TRAMA

--schedule-based (TDMA) --time divided into periods: random access (signaling, collisions possible) vs. scheduled access (data exchange b/w nodes) --limitations: complex, overhead, queueing delay, high memory/CPU need

modulation

--sending data w/ radio waves --wave parameters (amp, freq, phase) manipulated for sender to encode data --sine wave modified thru analog modulation + digital keying --> has center freq. and requires BW to be tx

sensor node hardware

--sensing unit/sensor ADC --processor/memory --power unit --location finding system --mobilizer --transceiver --antenna ex: Arduino, Raspberry Pi, Telos, MICAz

wireless sensor and actor network (WSAN)

--sensor-driven automated interaction w/ environment --increased need for automation (smart houses...) --dense sensors, loose actors

roles in WSNs

--sources: measure data w/ sensors, report them --sinks: receive data from WSN (gateway) --actors/actuators: control sources based on data

serial peripheral interface bus (SPI)

--synchronous serial interface --multi-slave, single master --10 Mbps clock speed (faster than asynch. serial) --four unidirectional lines --cons: more wires required, no inter-slave talk

inter-integrated circuit (I2C)

--synchronous serial interface --packet switched --multi-master, multi-slave --two bidirectional lines --cons: slower than SPI, more complex

transceiver states

--transmit (tx) --receive (rx) --idle: ready to rx but not --sleep: switched off, requires energy to boot-up

emerging IoT comm. interfaces

--ultrasonic: small, low-power --visible light: LEDs/cameras --vibration: accelerometers + vibration motors

contention-based MAC

--utilizes CSMA/CA --provides network robustness & scalability --collision prob. inc. as node density inc. (lower channel utilization & battery life...)

wireless multimedia sensor network

--wirelessly interconnected devices retrieve video/audio streams, images, sensor data --can store, process real-time, correlate data

wireless channel quality

--worse than wired channels (throughput, bit error, energy...) --extremely diverse --various improvement schemes --main challenge: limited BW & energy efficiency

data path

1. acquisition 2. local processing 3. communication 4. stream 5. storage & cloud 6. analytics

6 components of IoT

1. micro-sensors (temp, pressure...) 2. tags (RFID, QR...) 3. energy efficient comm. (small batteries, BLE...) 4. micro-computing (Arduino, RPi...) 5. cloud computing 6. open/small OS (Linux)

analog to digital converter (ADC)

1. samples signal in time (above Nyquist freq.) 2. quantizes sample amplitude (resolution) --major bottleneck for hi-speed DSPs

clear channel assessment (CCA)

B-MAC design feature --effective collision avoid. --estimates channel noise floor by taking RSSI samples: reduces false "busy's", ensures channel is idle, maximizes BW

low power listening (LPL)

B-MAC design feature --goal: minimize "listen" cost --fixed node wakeup time, variable "check time" -> if energy detected, node stays up to rx incoming packet --node sleeps after packet rx OR after false positive (no packet) --preamble length = channel check interval --large preambles > frequent checking

why are frequencies allocated into tx bands?

infrastructure costs & physical wave properties (low freq/high wavelength suitable for long distance...)

first smart device network

vending machine at Carnegie Mellon (1982) --> used internet to report inventory & product temp. readings

what makes something "smart"?

wireless communication!

link layer

handles all the physical details of interfacing with the cable, including the network interface card and a device driver

Moore's Law

computing power (IC transistor count) roughly doubles every two years --more comps/person --smaller computers --lower prices

RF transceiver characteristics

crucial for network performance evaluation --front-end: interface (bit or packet level), freq. range, data rates, range --energy costs: send/rx data, change states, tx, efficiency --performance: modulation, noise, gain, rx sensitivity, BER, RSSI, voltage range

sensor interconnects

enable comm. b/w sensing & processing units --serial: one bit at a time (Ethernet, I2C, USB...) --parallel: multiple bits tx (CAMAC, ISA, PC card...) --our focus: on-chip (SPI, UART...)

medium access control (MAC)

enables devices in a network to share a wireless channel: --controls how shared medium is used --controls when to send/listen for a packet

Internet of Things (IoT)

globally interconnected continuum of embedded devices interacting with environment, people, and each other --collect/exchange data --receive control comm.

signal conditioning

manipulation of signal in a way that prepares it for the next stage of processing --filtering (noise reduction) --amplification/attenuation --all in analog domain

fieldbus

network type for real-time comm. over wireless & smart sensing/measuring/controlling actuators

Bell's Law

new computer class forms roughly each decade establishing a new industry

RTS/CTS

request to send/clear to send --sender sends RTS before it sends a packet (idle>DIFS) --receiver responds w/ CTS if ready to rx (idle>SIFS) --solves hidden/exposed terminal problems b/c if node does not receive CTS, it won't send or cause collision! --RTS collisions are ok b/c of small packet size

S-MAC

sleep MAC --contention-based protocol --nodes go into periodic sleep mode (radio off) & timer is set to awake later & listen for activity --requires occasional SYNC packet for nodes to maintain schedule & avoid clock drift --nodes set schedules (synchronizers vs followers) --border nodes sleep less: adopt one schedule only! --immediate neighbors sleep to save overhearing energy **causes latency but saves energy from idle listening** **inc. collision rate due to sleep schedules**

wireless sensor node

small, battery-powered embedded system with dedicated function

ISM band

unlicensed band of RF associated w/ industrial, scientific, and medical devices