Energy Reduction in Wireless Sensor Networks
Transcription
Energy Reduction in Wireless Sensor Networks
Wireless Sensor Networks (WSN) ! What is a Wireless Sensor Network ? ! Dense network of small nodes communicating through wireless links Energy Reduction in Wireless Sensor Networks • for sensing, actuation, processing, storage, communication and control • ad hoc network Olivier Sentieys with contributions from: Thomas Anger, Jérôme Astier, Arnaud Carer, Olivier Berder, Duc Nguyen, Vinh Tran, Adeel Pasha, Steven Derrien, Hai-Nam Nguyen, Daniel Ménard, Vivek T.D., Mahtab Alam Relay Sensor and relay IRISA/INRIA, University of Rennes 1 ENSSAT Lannion [email protected] Base station and gateway Wireless Sensor Network ECOFAC’2010 Ex. Temperature sensing (1,8) !(7,7) (8,8) (7,7) ?(7,7) (5,6) (1,5) !(7,7) (4,2) !(7,7) (8,4) (11,3) ?(7,7) Wireless Sensor Networks (WSN) ! Set of nodes with coordinates (X,Y) ?(7,7) 2 ! Temperature sensing at (7,7) ! BS(0,0) ! ?(7,7) ! (1,5) ! ?(7,7) ! (5,6) ! ?(7,7) ! (8,8) ! !(7,7) ! (5,6) ! !(7,7) ! (4,2) ! !(7,7) ! (0,0) receives temperature ! Simplified deployment, fault tolerance • No maintenance and battery replacement ! Network characteristics • Low mean distance • Limited amount of data • Multi-hop routing ! Low cost ! Small size ! Long autonomy, low power • Towards autonomous sensor nodes (0,0) ECOFAC’2010 3 ECOFAC’2010 4 On-demand model ECOFAC’2010 Event-driven model [Simplot, 2009] 5 [Simplot, 2009] ECOFAC’2010 Emerging applications Tremendous space of applications ! Indoor environment ! Monitoring space: ocean water, pollution, ! Monitoring things: robots, human body,… • Industrial control, machine health monitoring, condition based maintenance of equipment in factory • Indoor localization • Home automation, intelligent lightning and heating • Health care 6 ! Outdoor environment • Monitor natural habitats, remote ecosystems, agriculture, forest fires • Structural Health Monitoring (e.g. bridges, buildings) • Automotive • Disaster sites • Security & military surveillance ECOFAC’2010 Fire detection Target monitoring 7 ECOFAC’2010 Monitoring in agriculture [Fischione, UCB, 2007] 8 ANR SurVeiller et Prévenir Captiv http://captiv.irisa.fr ! Capteurs de paramètres physiologiques • Suivi de l’activité physique • Prévention de l’obésité Cooperative Wireless V2I Communications ! http://svp.irisa.fr ECOFAC’2010 9 ITEA2 Geodes: fire-fighters ECOFAC’2010 10 Fire-fighter scenario ! Indoor network • Temperature monitoring, smoke detection, motion detection • Camera nodes WSN nodes Fire-Fighters FIRE-FIGHTER LOCALIZATION ! Mobile nodes (firefighter) PEOPLE DETECTED • Health monitoring, camera, etc. • Connected with indoor network ECOFAC’2010 PEOPLE DETECTED FIRE DETECTED 11 ECOFAC’2010 PEOPLE DETECTED PEOPLE DETECTED 12 Challenges in WSNs (1/2) Challenges in WSNs (2/2) ! Networking ! Energy reduction of HW/SW • Physical design • MAC protocols o Power saving, collision avoidance, ... • Routing protocols Cross-layer optimizations • Digital processing o Low power processors • Radio front-end, analog processing o Small radios with small bandwidth & transmission ranges o Data aggregation, dissemination, ... • Light and efficient software ! Network management o No complex operating systems o Mainly hardware-dependant software • Deployment, redeployment, topology control, maintenance • Localization and positioning • Coverage and connectivity problems ! Process or transmit ? • Limited processing power but communicating wirelessly is power-hungry ! Application • Data fusion, distributed signal processing, source coding, positioning, tracking, etc. ECOFAC’2010 ! Towards a true energetic autonomy 13 ECOFAC’2010 Autonomous nodes ? Agenda ! A WSN node is limited by the total energy it can store or scavenge from the environment ! WSN node architecture 14 • HW Platform • Processor and radio transceiver performance • Need a drastic reduction in the total consumed energy (radio + processing) ! Protocols • MAC and network • SW Stack ! Power models and estimation ! Energy optimization • • • • [CEA Leti, Managy] ECOFAC’2010 15 Cross-layer (MAC/LINK) FPGA co-processing Architectural and circuit-level optimization MIMO Cooperation ECOFAC’2010 16 Generic architecture of a node Generic architecture of a wireless node • Main tasks: sense, process, store, communicate, power management • Other features (e.g. location finding systems) Generator TION APPLICA NET DC/DC conv. Battery LINK SW e ructur t s a r f n I Sensor A/D Processor Coprocessor MAC Radio PH Y RAM Flash ture rastruc f n I W H ECOFAC’2010 17 Typical WSN platform ! What are the main sources of energy consumption ? • Radio: 30-70mW • Processor: 5-10mW DC/DC 18 Typical WSN platform ! Modern low-power microcontrollers Timer uController Radio ! Micro-controller ! Transceiver in Rx ! Transceiver in Tx Power consumption of a typical WSN platform ECOFAC’2010 ECOFAC’2010 Device Year Arch. Vdd (V) Ram Flash (kB) (kB) Active1 Sleep Wake (mA) (uA) (uS) Atmel AT128L AT256l 2002 2005 RISC 8b 2.7-5 1.8-5 4 8 128 256 0.95 0.9 5 1 6 6 Freescale HCS08 2003 MC13213 2007 8b 2.7-5 2-3.4 4 4 60 60 7.4 6.5 1 35 10 10 Jennic JN5139 2007 32b 2.2-3.6 192 128 3 3.3 2500 TI MSP430 F1612 F5437 2004 2008 RISC 16b 1.8-3.6 5 16 55 256 0.5 0.28 2.6 1.7 6 5 TI CC2430 2007 8051 2-3.6 128 5.1 0.5 4 NXP LPC1114 2010 CortexM0 ARM 32b 1.8-3.6 8 32 0.25 6 - 1Active 19 current at 3V and 1MHz ECOFAC’2010 8 #6mW@8MHz [Dutta, Sensys, 2008] 20 Typical WSN platform PowWow HW Platform (PWNode) PowWow: power optimized hardware/software framework for wireless motes ! Radio transceivers ! Open source hardware developed at IRISA/Cairn • IEEE 802.15.4 compatible Device Year Vdd (V) Atmel RF230 2006 1.8-3.6 Freescale Jennic • Mother board with MSP430 • Daughter boards for RxSens TxPwr (dBm) (dBm ) Rx Tx (mA) (mA) Sleep Wake (uA) (mS) -101 +3 15.5 16.5 0.02 1.1 MC13212 2005 2-3.4 -92 +3 37 30 1 7-20 JN5139 2007 2.2-3.6 -95 +0.5 37 37 2.8 >2.5 TI CC2420 CC2520 2003 2.1-3.6 2008 1.8-3.8 -95 -98 -20 to 0 18.8 -20 to 5 18.5 17.4 25.8 1 0.03 0.58 0.5 Leti Letibee 2008 1.2 -85 -3 9 - - 6 o V1: CC2420, Sensors o V2: FPGA, DVFS ! FPGA for hardware acceleration ! Voltage and frequency scaling ! Power management (sleep, wake-up) http://powwow.gforge.inria.fr #50mW ECOFAC’2010 21 PowWow HW Platform (PWNode) 22 Agenda ! Motherboard ! WSN node architecture • TI MSP430 low-power microcontroller • HW Platform • Processor and radio transceiver performance o MSP430F1612 version, 8 MHz clock o 55KB of flash memory, 5KB of on-chip RAM o 330uA at 1 MHz and 2.2 V in active mode o 1.1uA in standby mode ! Protocols • MAC and network • SW Stack • JTAG, RS232, and I2C interfaces ! Radio daughterboard ! Power models and estimation ! Energy optimization • TI CC2420 RF transceiver • single-chip 2.4 GHz IEEE 802.15.4 compliant • digital direct sequence spread spectrum baseband modem • • • • o spreading gain of 9 dB, data rate of 250 kbps o hardware support for • packet handling, data buffering, burst transmissions, data encryption, data authentication, clear channel assessment, link quality indication and packet timing information ECOFAC’2010 ECOFAC’2010 23 Cross-layer (MAC/LINK) FPGA co-processing Architectural and circuit-level optimization MIMO Cooperation ECOFAC’2010 24 MAC Protocols MAC Protocols ! MAC protocol determines next node to use the medium ! Sensor-MAC (S-MAC) ! Carrier Sense Multiple Access (CSMA) ! Timeout-MAC (T-MAC) • Simple, scalable • e.g. 802.15.4 • CSMA with fixed active/sleep duty cycle • Synchronization of active times ! Time Division Multiple Access (TDMA) • Avoid collision • e.g. GSM • Listen always • Collisions • Make duty cycle dynamic • End active time with time-out " Complexity of scheduling " Synchronization needed ECOFAC’2010 25 ECOFAC’2010 26 MAC Protocols Power Measurements on PowWow HW ! Asynchronous Rendezvous ! Wake-up and channel sensing • Asynchronous scheme initiated by receiver • RICER (Receiver-Initiated CyclEd Receiver) Rx Tx Rx Receiver wait Node j (transmitter) Wake-up period (T) All measurements realized with Agilent N6705A DC Power Analyzer Data transmission Node k (receiver) Wake-up frames Waiting time Acknowledge Radio transceiver in reception/listening mode Radio transceiver in transmission mode ECOFAC’2010 [Lin, ICC, 2005] 27 ECOFAC’2010 28 Power Measurements on PowWow HW Power Measurements on PowWow HW ! Wake-up and channel sensing ! Wake-up and channel sensing with collision 6, 8, 16, 32, 64, 128 packet size in bytes ECOFAC’2010 29 ECOFAC’2010 30 Routing Protocols Transmission Modes ! Multi-hop routing ! Geographical routing ! Broadcast • Direct transmission to {neighbors}, no ACK ! Flooding • Each node has (x,y) coordinates • Next node for hop transmission is chosen in the neighbors as the nearest to the destination • Broadcast a packet to all network nodes, no ACK ! Direct Hop with/without ACK o in the sense of Euclidian distance • Direct transmission to a specific neighbors • Neighbor table management o with or without ACK o A neighbor is a node in the radio range of another node o Set of neighbors is discovered power-up and on regular time period ! Robust Multi-Hop • Multi-hop transmission to a specific node in the network • Each hop is with ACK • Uses node address ! Many other protocols! ECOFAC’2010 31 ECOFAC’2010 32 PowWow SW Stack PowWow SW Stack ! HAL, PHY, LINK, MAC, NETW layers and API ! Open source software developed at IRISA/Cairn • • • • • http://powwow.gforge.inria.fr ! Event-driven, multi-threaded, C code • Based on Protothread library and Contiki FEC/ARQ, geographical routing, positioning, Tx power and low-power mode management Modes: broadcast, flooding, direct/multi-hop with/without ACK Configurable packet structure ! Analytical power estimation based on software profiling and power measurements of a set of scenarios ! Memory efficiency • 6 Kbytes (protocol layers) + 5 Kbytes (application) ! Over-the-air re-programmation (and soon reconfiguration) ! Comparison with 802.15.4/ZigBee • More than 12 less power for the same application scenario o Temperature sensing, 1-10-30s sensing period, TI MAC vs. PowWow ECOFAC’2010 33 Agenda ECOFAC’2010 34 Energy Model ! WSN node architecture • HW Platform • Processor and radio transceiver performance ! Protocols • MAC and network • SW Stack o PTB/PRB: power in baseband digital signal processing circuit of transmitter and receiver (mW) o PTRF/PRRF: power in front-end circuit of transmitter and receiver (mW) o PA: power of the Power Amplifier (mW) o PL: power of the Low Noise Amplifier (mW) ! Power models and estimation ! Energy optimization • • • • Cross-layer (MAC/LINK) FPGA co-processing Architectural and circuit-level optimization MIMO Cooperation ECOFAC’2010 35 ECOFAC’2010 36 Energy Model Energy Model ! Total power consumption ! Relation between PT(d) and the required RF output power for reliable transmission PTx(d) • Tx: PT(d) for a transmission of distance d • Rx: PR • • • • • d ECOFAC’2010 37 Chipcon Radio Tranceivers Gt, Gr: transmitter and receiver antenna gain PRx: desired receive power at destination !: drain efficiency !: carrier wavelength L: system loss factor ECOFAC’2010 38 Energy Consumption ! CC2420 radio transciever d=10m, SISO ECOFAC’2010 39 ECOFAC’2010 d=100m, SISO 40 Energy Estimation Energy Estimation ! How to estimate the energy more precisely ? ! How to estimate the energy more precisely ? ! Total energy dissipated in the WSN E = V . (Ep + Ert.Nrt + Ewuc.Nwuc + Ec.Nc) ! Mixed estimation method • • • • • • • Profiling of code execution or platform power measurement o Ep Ert V = {Vi,j}: volume of data between i and j Nrt: mean number of retransmissions Nwuc: mean number of wake-up collisions Nc: mean number of collisions Ep: energy of a packet transmission Ert Ewuc Ec: energy of a retransmission, a wake-up collision and a collision [Thèse Cartron, 2006] ECOFAC’2010 • Application and network simulation o {Vi,j} • Analytical performance models o Nrt 41 Scenario-based power estimation PP2 TxOK RxEr Rx Too many errors detected that can not be corrected TxC RxC Wake-up collision at Rx PP1/2: Process Packet Phase ! RxEr: Rx Packet with error ECOFAC’2010 TxOK RxOK PP2 Nc 42 Scenario-based power estimation Energy consumption of event cycles [J] • e.g. MAC events PP1 Nwuc ECOFAC’2010 ! Analytical approach based on software profiling and power measurements of a set of scenarios Tx Ewuc Ec RxOK TxOK Wake-up collision of a Tx with another Tx CBT T WUR WUC DC TIM Rx soft 5.3e-8 5.3e-8 1.2e-4 1.2e-4 5.3e-8 5.3e-8 Tx soft 4.1e-8 1.2e-3 1.0e-8 4.1e-8 6.9e-4 4.1e-8 clock 5.5e-7 5.5e-7 5.5e-7 5.5e-7 5.5e-7 5.5e-7 LINK 4.8e-7 0 4.8e-7 0 0 0 NETWORK 4.8e-7 0 0 0 0 0 Req_neighb 0 0 0 0 0 0 Ans_neighb 0 0 0 0 0 0 positionning. 0 0 0 0 0 0 listen target 0 0 0 0 0 0 scheduler 3.7e-7 3.7e-7 3.7e-7 3.7e-7 3.7e-7 3.7e-7 Tx HF 0 2.32e-3 2.32e-3 0 0 0 Rx HF 0 5.2e-2 3.1e-3 3.1e-3 5.2e-2 0 CBT: Calculation Before Transmission WUR: Wake Up with Reception DC: Data Collision TxOK/RxOK: Normal Tx/Rx TxC/RxC: Tx/Rx with collision 43 ECOFAC’2010 T: Transmission WUC: Wake Up with Collision TIM: Timer 44 Scenario 1 Scenario 2 ! Periodic sensing and data transmission to BS ! Periodic sensing with data transmission in case of alarm 1 Periodic sensing 3 2 4 1 2 7 Data transmission 4 4 Periodic sensing 1 2 9 5 1 2 7 6 4 7 6 5 6 7 7 3 5 2 3 4 1 Alarm 1 2 4 MSdesc V (7) Alarm data E(i) : number of alarm at node i V : matrix of messages i!j Tobs: simulation time Fech: sensing frequency MSdesc(i) : matrix of data transmitted from node i 45 ECOFAC’2010 46 45 ECOFAC’2010 46 Scenario 3 Results on scenario 2 ! On-demand sensing ! Periodic sensing with data transmission in case of alarm 0,343 0,139 No event 1,86 3,744 Wake-up interval: 2 s Query F(i) : number of queries to node i Radio Rx Digital 0,29 0,432 Available energy: 3 AA batteries Sensing MSasc(i) : matrix of queries transmitted to node i P = 3 mW Auton. = 126 days Data 0 Radio Tx 2,688 Total energy: 10.75 W.h 1,256 soft Rx soft Tx clock soft link layer soft network request neighb. answer neighb. thread positioning thread sensing scheduler radio Tx radio Rx 47 ECOFAC’2010 47 ECOFAC’2010 48 Results on scenario 2 Simulation: WSim+WSNet ! Periodic sensing with data transmission in case of alarm ! Open source software developed at INSA Lyon ! WSim: hardware platform simulation 0,337 0,355 0,274 1,758 10 events per second Wake-up interval: 2 s Available energy: 3 AA batteries 0,409 0 Digital Radio Rx 1,187 2,957 13,296 Radio Tx P = 5.7 mW Auton. = 66 days • • • • soft Rx soft Tx clock soft link layer soft network request neighb. answer neighb. thread positioning thread sensing scheduler radio Tx radio Rx ! WSNet: event-driven simulator for wireless networks • • • • Node simulation Environment simulation Radio medium simulation Extensibility ! http://wsim.gforge.inria.fr ! http://wsnet.gforge.inria.fr Total energy: 20.57 W.h ECOFAC’2010 Cycle accurate simulation Several models (processors, transceivers, peripherals) Interaction with WSNet for distributed network simulation Application final binary 49 Simulation: WSim+WSNet ECOFAC’2010 [Chelius, 2006] 50 Agenda ! WSN node architecture • HW Platform • Processor and radio transceiver performance ! Protocols • MAC and network • SW Stack ! Power models and estimation ! Energy optimization • • • • ECOFAC’2010 51 Cross-layer (MAC/LINK) FPGA co-processing Architectural and circuit-level optimization MIMO Cooperation ECOFAC’2010 52 Main Goals Power optimization of a wireless node ! How to design and optimize an energy-efficient software and hardware platform for wireless sensor networks ? TION APPLICA NET ! (1) Decrease transmission (Tx) power • Power-aware signal processing • Error detection and correction LINK SW e ructur t s a r f n I ! (2) Optimize radio activity and MAC MAC PH Y ! (3) Power optimization of the hardware ture rastruc f n I W H ECOFAC’2010 53 ECOFAC’2010 54 Results on MAC parameter optim. Results on MAC parameter optim. ! Wake-up period influence ! Wake-up period influence Short WUP • Application dependent Short WUP T 0,2 s 19,45 mW 19,3 Days E R More collisions Useless wake-up T R R Optimal WUP Optimal WUP T 1,6 s 5,64 mW 66,5 Days T R R R E Long WUP Long WUP Rx power increase R R R ECOFAC’2010 8,0 s 12,17 mW 30,8 Days Rx power increase T T E More collisions Useless wake-up Transmission Mode Receiver Mode Transmission Mode Receiver Mode 55 ECOFAC’2010 [Thèse Cartron, 2006] 56 Output Radio Power Management Power optimization of a wireless node TION APPLICA NET SW e ructur t s a r f n I LINK MAC PHY ture rastruc f n I W H ECOFAC’2010 57 ECOFAC’2010 58 Performance/energy joint modelling Output Radio Power Management ! Energy per successfully transmitted bit ! Energy per successfully transmitted bit Energy per successfully transmitted bit[J] Energy per successfully transmitted bit[J] D=10 m, Pnoise=-90 dBm, 53 bytes packets d CC1020 transceiver • as a function of distance and Tx power • without error correction D [m] Transmission Power (PTx) [dBm] ECOFAC’2010 " Dynamic adaptation of output radio power depending on channel conditions, distance, etc. " RSSI or CRC as quality metrics [Sentieys, DASIP, 2007] 59 ECOFAC’2010 PTx [dBm] 60 HW Platform Energy Optimization Power optimization of a wireless node ! ! ! ! TION APPLICA NET SW e ructur t s a r f n I LINK (1) (2) (3) (4) Co-processing Dynamic Voltage Scaling Power Gated FSM Dynamic Precision Scaling, etc. Generator DC/DC conv. Battery MAC Sensor PHY A/D Processor Coprocessor RAM Flash ture rastruc f n I W H ECOFAC’2010 61 PowWow HW Platform (PWNode) Radio ECOFAC’2010 62 Co-Processing with a Low Power FPGA ! FPGA co-processing, Power Mode Management (PMM) and Dynamic Voltage and Frequency Scaling (DFVS) daughterboard • Low-power Igloo FPGA from Actel o AGL125: 130nm, 125 kgates, 32kbits on-chip RAM, 1 kbits Flash, PLL for clock management. o Supply voltages 0 to 1.65V o Power consumption: 2.2uW, 16uW, 1-30mW in sleep, freeze, run modes Watch Dog Vdd scaling Wake-Up CC2420 o e.g. Viterbi implemented for error correction: 5mW Px ECOFAC’2010 DC/DC 63 ECOFAC’2010 Igloo FPGA Control Data MSP430 Sensors Co-processing mode 64 Dynamic Voltage Scaling (1/3) Dynamic Voltage Scaling (2/3) DVS (and frequency) Power of MSP430 1.6V, 1.8V, 2V, 2.5V, 3V, 3.3V 1.6V, 1.8V, 2V, 2.5V, 3V, 3.3V ECOFAC’2010 65 Dynamic Voltage Scaling (3/3) ECOFAC’2010 66 PowWow HW Platform (PWNode) ! FPGA/DFVS daughterboard (cont.) • Power Mode Management o Low-Power Programmable Timer Power of MSP+CC2420 • MAX6370, 8uA • Wake-up period 1.6V, 1.8V, 2V, 2.5V, 3V, 3.3V • DVFS o Programmable Clock • LTC6930, 490uA • 8MHz divided by 1 to 128 o Programmable DC/DC conv. • TPS62402/TPS61030 ECOFAC’2010 67 ECOFAC’2010 68 HW Platform Energy Optimization ! ! ! ! (1) (2) (3) (4) Power Gated Controllers (1/6) Co-processing Dynamic Voltage Scaling Power Gated FSM Dynamic Precision Scaling, etc. Generator Sensor Battery A/D ! Power Gating Principle DC/DC conv. Processor Coprocessor RAM Flash Radio ECOFAC’2010 69 Power Gated Controllers (2/6) ! Task graph to gated FSM ECOFAC’2010 70 Power-gated Micro-Task (3/6) ! HW specialization ! Leakage reduction ! Micro-Task • Customized FSM + minimalistic data-path Task A Task B Task C ECOFAC’2010 [Pasha, ISCAS 2009] 71 ECOFAC’2010 72 Power Gated Controllers (4/6) Automatic generation flow (5/6) ! Power gain versus MSP430 software execution ! C to VHDL compilation flow for the automatic hardware task generation o Gains w.r.t. the power and energy consumptions of an MSP430F21x1 (datasheet) and an open core MSP430-clone (without memory) o Operating frequency of 16 MHz • Based on GeCoS compiler/HLS infrastructure ECOFAC’2010 [Pasha, DAC, 2010] 73 Automatic generation flow (6/6) ECOFAC’2010 74 Dynamic Precision Scaling (1/3) ! System-level DSL ! Energy consumption reduction by fixed-point adaptation of the data-path wordlegth • depending on observed error rate (BER) and signalto-noise ration (SNR) System System outputs System inputs f fp ( p) Fixed-point specification selection p Metric p measurement ECOFAC’2010 75 ECOFAC’2010 [Thèse H.N. Nguyen] 76 Dynamic Precision Scaling (2/3) Dynamic Precision Scaling (3/3) ! Range estimation: determine the minimal integer word-length which guarantees no overflow ! Precision Analysis: determine the minimal fractional word-length which guarantees the performance criterion ! 40% of energy savings between 0dB and –25dB ECOFAC’2010 • CDMA receiver, data-path energy 77 Radio transceiver optimization [Nguyen, ISCAS, 2009] ECOFAC’2010 78 Power optimization of a wireless node ! LetiBee chip (CEA LETI) • Expected power consumption (2nd release) Function RX (mA) TX RF 0.5 2.73 LO 4 7 PLL 0.35 0.35 Analog 0.2 0.4 Digital 0.5 0.25 Biasing 1.5 0.5 TION APPLICA NET SW e ructur Infrast o Tx # 13.5 mW @ -2 dBm o Rx # 8.5 mW @ -85 dBm MAC PH Y ! Trends ture rastruc f n I W H • Wake-up radio, Ultra-Wide Band ECOFAC’2010 LINK [Bernier, ESSIRC, 2008] 79 ECOFAC’2010 80 Cooperative techniques Alamouti scheme ! Context ! Model • Physical layer (with impact on MAC/NET layers) • Cooperative strategies between wireless nodes s(1) -s*(2) y(1) y(2) Source o take advantage of channel spatial and temporal diversity to decrease the radio output power s(2) Destination s*(1) ! Protocol ! Objectives • Optimize different Cooperative MIMO techniques • Compare and associate them with Relay techniques • Consider the energy consumption to determine the optimal selection scheme ECOFAC’2010 [Alamouti 1998] ECOFAC’2010 81 Performance of Space-Time Codes 82 Cooperative MIMO ! Three phases of C-MIMO communications • Phase 1: Local data exchange and space-time coding • Phase 2: Virtual MIMO transmission • Phase 3: Cooperative reception d Nr S D MIMO transmission dm Nt Space-Time Codes bring us better performance ECOFAC’2010 dm<<d 83 ECOFAC’2010 dm = 1..10 m 84 Energy consumption of C-MIMO Cooperative MIMO ! Cooperative MIMO technique is more energy efficient than SISO and multi-hop SISO techniques for long distance transmission ! Total energy = Transmission Energy + Digital Energy + Cooperation Energy d=100m, MISO 2-1 ECOFAC’2010 [Nguyen, ICC, 2008] 85 Relay Technique d=100m, SISO ECOFAC’2010 86 Performance of relay technique ! Diversity gains by sending additional copies of the signal through relays ! Less delay, simpler processing ! Two main types • Amplify and Forward (AF) • Decode and Forward (DF) Performance of relay model is better than that of SISO ECOFAC’2010 87 ECOFAC’2010 88 MIMO relay model r1(1) yr1(1) MIMO simple cooperative relay model r1(1) r1(2) yr1(2) yr1(1) s(1) -s*(2) r2(1) yr2(1) R2 r2(2) yr2(2) s(1) -s*(2) y(1) y(2) t1 y(3) s(2) S1 s(1) -s*(2) S2 s(2) s*(1) R1 yr1(1) yr1(2) R2 yr2(1) yr2(2) D y(1) y(2) t3 Transmit r1(1) t5 t4 y(3) ECOFAC’2010 t6 s*(1) Time Slot Transmit r1(2) Transmit r2(1) y(4) y(5) Transmit r2 (2) y(6) 89 R2 yr2(2) r2(2) t1 t2 r2*(1) t3 t4 S1 s(1) -s*(2) S2 s(2) s*(1) R1 yr1(1) yr1(2) r1(1) -r1*(2) R2 yr2(2) yr2(2) r2(2) r2*(2) D y(1) y(2) y(3) y(4) ECOFAC’2010 90 Performances Energy simulation ! AF protocol ! Optimal model choice to minimize energy relative distance: d(source-relay) d(source-dest) BER (Bit Error Rate) y(4) D y(4) y(5) y(6) s*(1) t2 y(3) y(2) Source yr2(1) Time Slot R1 yr1(2) y(1) D Source s(2) -r1*(2) SNR (Signal to Noise Ratio) ECOFAC’2010 92 ECOFAC’2010 [Thèse V. Tran] 93 Cooperative MIMO and Relay Summary ! Advantage ! Energy minimization in WSN • Complex cross-layer problem • Power/performance models • Energy efficiency for o long-range transmission (WSN) o fading channels ! Reduction of Tx Power ! Trade-off • Signal processing, error correcting code • Extra circuit consumption ! Reduction of Rx activity o MIMO digital signal processing • MAC, routing • Delay of cooperative local data transmission or relay • Distance between cooperating nodes dm much smaller than transmission distance d ! Power optimization of heterogeneous platforms • Power management, dedicated hardware, power gating, etc. ! Challenges • Link with MAC and route protocols • Choice of the optimal cooperating strategy ECOFAC’2010 ! Power optimization of analog and radio 94 Design challenges Signal Processing Control ! Channel coding ! MIMO, relay … ! Modulation ! MAC ! Routing … Communications ECOFAC’2010 [Min02] R. Min et al., Power-aware Wireless Microsensor Networks, in Power-aware Design Methodologies, 2002. [Cui04] S. Cui, A. Goldsmith, Energy_efficiency of MIMO and cooperative MIMO Techniques in Sensor Networks, IEEE JSAC, 2004. [Cartron 2006] M. Cartron, Vers une plate-forme efficace en énergie pour les réseaux de capteurs sans fil, PhD Thesis, University of Rennes 1, 2006. [Li07] Y. Li, B. Bakkaloglu and C. Chakrabarti, A System Level Energy Model and Energy-Quality Evaluation for Integrated Transceiver Front-Ends, IEEE Trans. on VLSI, 2007. [Nguyen07] Tuan-Duc Nguyen, Olivier Berder and Olivier Sentieys, Cooperative MIMO schemes optimal selection for wireless sensor networks, IEEE VTC-Spring, 2007. [Sentieys07] O. Sentieys, O. Berder, P. Quemerais and M. Cartron, Wake-up Interval Optimization for Sensor Networks with Rendez-vous Schemes, Workshop on Design and Architectures for Signal and Image Processing (DASIP), 2007. [Wang06] Q. Wang, M. Hempstead and W. Yang, A Realistic Power Consumption Model for Wireless Sensor Network Devices, IEEE SECON, 2006. [Pasha09] M. A. Pasha, S. Derrien, and O. Sentieys. Ultra low-power fsm for control oriented applications. IEEE International Symposium on Circuits and Systems, ISCAS 2009, pages 1577 – 1580, Taipei, Taiwan, May 2009. [Pasha10] M. A. Pasha, S. Derrien and O. Sentieys, A Complete Design-Flow for the Generation of Ultra Low-Power WSN Node Architectures Based on Micro-Tasking, Proc. of the IEEE/ACM Design Automation Conference (DAC) Anaheim, CA, USA, June 2010. [Nguyen08a] T. Nguyen, O. Berder, and O. Sentieys, Impact of transmission synchronization error and cooperative reception techniques on the performance of cooperative MIMO systems, IEEE ICC, 2008. [Nguyen08b] T. Nguyen, O. Berder, and O. Sentieys, Efficient space time combination technique for unsynchronized cooperative MISO transmission, IEEE VTC-Spring, 2008. [Lin05] E.Y Lin, J. Rabaey, S. Wiethoelter, and A. Wolitz. Receiver Initiated Rendez-vous Schemes for Sensor Networks. In Proc. of IEEE Globecom 2005, 2005. [Lin04] E.Y. Lin, J. M. Rabaey, and A. Wolisz. Power-Efficient Rendez-vous Schemes for Dense Wireless Sensor Networks. In IEEE International Conference on Communications ICC 2004, 2004. [Menard08A] D. Menard, R. Rocher, O. Sentieys, and O. Serizel. Accuracy Constraint Determination in Fixed-Point System Design. EURASIP Journal on Embedded Systems, 2008. [Menard08B] D. Menard, R. Rocher, and O. Sentieys. Analytical Fixed-Point Accuracy Evaluation in Linear TimeInvariant Systems. IEEE Transactions on Circuits and Systems I, 55(1), November 2008. ! Verification ! Distributed computing ! Embedded software ! Middleware ! Operating Systems … WSN 95 Bibliography Computer Science ! Hybrid systems ! Networked control ! Source coding ECOFAC’2010 ! Micro-architecture ! CAD tools ! Digital ! Analog ! RF … Microelectronics 96 ECOFAC’2010 97 Questions ? Thanks for their contributions to: Thomas Anger, Jérôme Astier, Arnaud Carer, Olivier Berder, Duc Nguyen, Vinh Tran, Adeel Pasha, Steven Derrien, Hai-Nam Nguyen, Daniel Ménard, Vivek T.D., Mahtab Alam and others