Wireless Sensor Network 0.1 Wireless Sensor Networks • How is it different from traditional wireless network? – for specific application – embedded system with very limited resources (memory, battery, os) – typical deployment with thousands of nodes – data-centric, individual node’s performance not important 0.2 Applications for WSN • Environment and habitat monitoring • Precision Agriculture • Indoor climate control • Military surveillance • Intruder detection • Earthquake/volcano prediction • Patient vitals monitoring 0.3 WSN System Challenges • Very Large Scale – Dense instrumentation – Limited device capability • Sometimes partial measurements have to be correlated – Limited Access • Deployed in remote places • Leverage wireless communication to gather information – Dynamic condition • Environmental condition, event reporting (change in load) • Death of nodes – Change in topology – Routing protocol – Operating Systems • Should be scaled down to fit the embedded architecture 0.4 WSN OSs • Should be scaled down to fit the embedded architecture • Example OSs – VXWorks, Linux variants, WindowsCE, GeoWorks • Component based OS – TinyOS from Berekeley • Concurrency, fine-grained power management, light-weight event scheduler 0.5 MANET vs WSN MANET WSN Application General Specific Devices (energy) More powerful, relatively more resources Small embedded system (with constrained resources) large Individual nodes important, ID centric, high data rates Coordinated effort among nodes, datacentric, mostly low data rate Scale Communication 0.6 WSN MAC • Attributes – Collision avoidance – Energy efficiency – Scalability and adaptability  important – Channel utilization – Latency – Throughput  - not so important – fairness 0.7 Classification of MAC • Scheduled protocols – Nodes send data in predetermined times • TDMA, FDMA, CDMA – Contention based protocols • Nodes compete with probabilistic coordination – ALOHA, CSMA 0.8 Scheduled vs contention-based protocols Scheduled protocols Contentionbased protocols collision No yes Energy efficiency Good bad scalability bad good Multihop communication Difficult easy Time sync precise Loose or not needed 0.9 Energy Efficiency in MAC • Sources of Energy Wastes – Collision – Control packet overhead – Overhearing unnecessary traffic – Idle listening (a major source of energy waste, consumes 50-100% of the power for receiving) 0.10 S-MAC • Major Features of S-MAC – Collision avoidance – Periodic listen and sleep – Overhearing avoidance – Adaptive listening – Message passing 0.11 Collision Avoidance • Based on CSMA • Similar to 802.11 DCF – Physical and virtual carrier sense – Randomized backoff – RTS/CTS for hidden node problem – RTS/CTS/DATA/ACK sequence 0.12 Periodic listen and sleep • Since idle listening consumes lots of energy, S-MAC employs periodic listen and sleep listen sleep listen • Turn off radio while sleeping • Reduce duty cycle to ~10% • Increases latency but decreases energy consumption 0.13 Periodic listen and sleep • Neighboring nodes will have the same schedule • But two nodes who are multihops away, may end up with different schedules • Border nodes will follow two schedules – This enables broadcast packets to be sent only once across the two clusters 0.14 Overhearing avoidance • Problem – Nodes receive packets destined for others • Solution – Sleep when neighbors talk • Who should sleep? – All immediate neighbors of sender and receiver • How long to sleep? – The duration field in each packets should be the duration that the neighbors should sleep 0.15 Adaptive Listening • Reduces latency in multi-hop scenario • Wake up for a short period of time after transmission from neighbors (if it overheard the corresponding RTS/CTS) – This way, if the node is the next hop node, the node will be able to send the data immediately instead of waiting for the scheduled listen time – Reduces latency, but increases duty cycle => more energy consumption 0.16 Message Passing • Fragment large messages into small fragments • Have one RTS-CTS exchange for the entire message – Reserve the medium for the entire message • But ACK is sent by the receiver for every fragment – If an ACK is not received, only that fragment is retransmitted and the reservation period is extended for one more fragment • If the entire msg were sent at once, then retransmission would have been costlier • If only one fragments were sent per RTS-CTS, the control overhead would have been higher and the msg level latency would have been higher 0.17 Implementation • Platform – Mica Motes (UC Berkeley) • 8-bit CPU at 4MHz, • 128KB flash, 4KB RAM • 20Kbps radio at 433MHz – TinyOS: event-driven (modified stack) • Configurable S-MAC options – Low duty cycle with adaptive listen – Low duty cycle without adaptive listen – Fully active mode (no periodic sleeping) 0.18 Experiment Ten hop linear topology Source: S-MAC paper 0.19 Energy Consumption vs msg inter-arrival Source: S-MAC paper 0.20 Latency vs number of hops Source: S-MAC paper 0.21 References • W Ye, J Heidemann and D Estrin, “Medium Access Control with Coordinated Adaptive Sleeping for Wireless Sensor Networks”- IEEE Transactions on Networking, vol 12, No 3, June 2004 0.22 ...Wireless Sensor Networks • How is it different from traditional wireless network? – for specific application – embedded system with very... Estrin, “Medium Access Control with Coordinated Adaptive Sleeping for Wireless Sensor Networks”- IEEE Transactions on Networking, vol 12, No 3, June 2004 0.22