DISTRIBUTED SYS & SENSORS
DISTRIBUTED SYS & SENSORS ECSE 6962
Popular in Course
Popular in ELECTRICAL AND COMPUTER ENGINEERING
This 35 page Class Notes was uploaded by Miss Damien Crooks on Monday October 19, 2015. The Class Notes belongs to ECSE 6962 at Rensselaer Polytechnic Institute taught by Staff in Fall. Since its upload, it has received 7 views. For similar materials see /class/224777/ecse-6962-rensselaer-polytechnic-institute in ELECTRICAL AND COMPUTER ENGINEERING at Rensselaer Polytechnic Institute.
Reviews for DISTRIBUTED SYS & SENSORS
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 10/19/15
Power Control and Cross Layer Design in Ad Hoc and Sensor Networks ECSE 6962 Di Wang 1 1072005 Outline Overview Design Principles for Power Control Power Control Protocols Unintended Consequences Control Theory Based Approach Conclusion Reference a Overview Why is Power Control Important Limited resources of energy Aiming to bring better performances Throughput Delay Why is Power Control a Cross Layer Design Problem Affect the physical layer quality of the signal Affect the network layer range of transmission Affect the transport later magnitude of the interference Overview Multi dimensional Effect Mac Layer Performance contention for the medium Topology Control Problem Connectivity of the network Effect on several important metrics Energy Consumption Throughput Capacity End to End Delay Impact on protocols in existence Create unidirectional links Affect MACrouting protocols Distributed Bellman Ford RTSCTS handshake in IEEE 80211 Design Principles For Power Control To increase network capacity it is optimal to reduce the transmit power level For transmit range r The area of interference is proportional to r2 The relaying burden is proportional to 1r Then The area consumed by a packet is proportional to r Design Principles For Power Control Reducing the transmit power level reduces the average contention at the MAC layer For any given point in the domain An average of ch transmitters within range Traffic flowing through each node is proportional to 1r then The net radio traffic in contention range is proportional to r Design Principles For Power Control The impact of power control on total energy consumption depends on the energy consumption pattern of the hardware Terms PRXelec the power consumed in the receiver electronics for processing PTxelec the power consumed in the transmitter electronics for processing PTxRadp Power consumed by the power ampli er to transmit a packet at the power level p PIdle I PSleep The impact of power control on total energy consumption If the energy consumed for transmission PTxRadp Dominates Using low power level is broadly commensurate with energy efficient routing for commonly used inverse 0th law path loss models with 022 Energy efficient routing seeks to minimize Z quot2 quot Can get the graph consisting of edges lying along some power optimal route between any pair of nodes l 3940 39 49 E VI A A vs 5 39 The impact of power control on total energy consumption i WK 1quot ii iii 39l i i W quotWe EV A i q NV E 07 i iMAM 4i 1 MW IA 39 v g 9 W I i H i iv i 1 iii 1 4 4 WM A 5191301 U aGUElSUM 103cc mm X distance in meters 5am 4u03 The graph of links lyjllg along pourer optimal routes 1 Connections only with nearby nodes and no intercections The impact of power control on total i energy consumption Suppose two edges cross For o2 can find an anglej lt 90 xi x12 xl xj2 ltxi xj2 10 The impact of power control on total energy consumption When PSleep is much less than PIdle turning the radio off whenever possible becomes an important energy saving strategy Estimates show that usually PIdle gt ZOPSeep Power management protocols seeking to put nodes to sleep while maintaining the network connectivity 11 The impact of power control on total energy consumption When a common power level is used throughout the network There exists a critical transmission range rcrit below which transmissions are sub optimal with regards to energy consumption Given two nodes with distance d the energy consumed for transmitting one packet d a r PRxelec PTXeleC C7 Which can be minimized at 3 Dillpjquotvclcr PI J39clcc crzr IIII CLx i 1 12 The impact of power control on end to end delay Power level and Traffic load jointly determine the end to end delay A Under high load a lower power gives lower delay Under low load a higher power gives lower delay packet experiences Propagation delay neglectable Processing delay time taken in receiving decoding and retransmitting inversely proportional to range r Queuing delay can be shown it increases super linearly with the power level p 13 The impact of power to end delay 13 gtP2gtP1 133 ll p2 3911 Network dn ougllp ut Delay A ua tative sketch of me expected delayethroughput Curves at xfferent povrer 1eve15 180 NEW delay in msecs D 0 Del simulations control on end Delay iThroughput so nodesv 1714 OER Conns 100 150 200 glegaLe Tluupul in KUilssdc e lrollghput curves at different power levels obtained through 14 Design Principles For Power Control Power control can be regarded as a network layer problem In fact it impacts multiple layers Numerous approaches attempt to solve it at MAC Layer Adjust the transmit power level to make the SINR just enough for receiver to decode the packet Only a local optimization Network layer power control is capable of a global optimization 15 Power Control Protocols COMPOW Protocol Design Strategies Choose a common power level Set this power level to the lowest value which keeps the network connected Keeps the energy consumption close to minimum while restricting the lowest admissible power level to rcrit Implementation Running multiple proactive routing protocols at each power level and find out the routing table with lowest p Appealing feature Provides bidirectional links 16 Power Control Protocols CLUSTERPOW Protocol COMPOW is not energy efficient when there are outlying node Design Strategies Select n different power levels to form a n level hierarchical structure Implementation Building routing table for each power level Transmitting packet at the smallest power level p such that the destination can be found on the p routing table 17 i CLUSTERPOW Protocol Routing by CLUSTERPOVV in a typical non homogeneous nenvon39k 18 g CLUSTERPOW Protocol CLUSTERPOW is loop free Still can be further improved o A quota 39 X r Y my D Suppose there is a loop on the path P om S to D Dashed lines indicate paths colIslstiIlg of many hops 19 Power Control Protocols Recursive Lookup Schemes 029mg 3 a r39 wquot N 1 N 7 Modlfyiug the CLUSTERPOVV protocol so that 13911 100 111quot hop from S to N1 can be replaced by ve hops of 1 39 113 m UV ach 20 s Recursive Lookup Schemes may not be loop free n1 V squot x The recursive lockup schexue is not free of in nite loops Solution Tunnelled CLUSTERPOW 21 Power Control Protocols Tun nel led CLUSTERPOW Protocol When doing recursive lookup for an intermediate node encapsulates the packet with the address of the node I In N 1 S 1 0 mvv N 1 0 1 0 11an D olt o o quot 1 N2 quot 111VV quoteenreee r N3 Tululelled CLUSTEELP OVV protocol resolves the nl mfe rolltmg loop of me neru39ork in Fxg 9 The headers added to the packet as n travels alollg me more are also shown 22 Power Control Protocols MIN POW Protocol Design Objective Provide a globally optimal solution with respect to total power consumption Implementation Proactively sends hello at multiple transmit power levels Only the hello packets at the Pmax contain routing updates For each link computes the power consumption per packet PTXtotal PTxelec PTxradp at all power level and take the minimum as the link cost in the distance vector algorithm Feature a globally optimal solution for power consumption but may not be the optimal solution for network capacity 23 Power Control Protocols Simulation Results E Q a 99 1 EQ a a 8 lt9 53 6 63 6 m m a a 3 a a 9 O 9 ewe r 63 G e a Q Q 69 ens ea zy 3 Q 3 399 e lt9 9 has E39 9 Single outlying node 3 6 3 8 1 A Illstered top310 gy 24 Simulation Results RELE V AANT S ILIULALTION P R4kIE TERS TC P TKAFFIC FOR THE SINGLE OUTLX ING NODE CALSE Agg Throughput Sn Devmrmn of Delay E X kbps 2 ms 1 V 4 23 1113 9056 1115 I 338 26431 1115 I 91 kbps 9 kbps I 1598 nus 3266 1115 Lu FT 039 T m U 25 1 nenalmr aroelzw EMWM My Mm am earn rim 4 v n amox aan nv r3417 759m wrnghnL39 Fauer mono 77 so mmugnpul Pam Mm raw a 21 mm mad H mm Constant bxt rate UDP traf c in a Clustered topology of so nodes thh randoxuly selected 16 nun arcluster and 4 intezgtcluster connections 26 Unintended Consequences Power Control can be addressed as Multi dimensional Optimization Usually one objective is achieved at the expense of one another Cross Layer Optimization Should not ignore the interactions between different layers 27 g Unintended Consequences Example the MINPOW Power Control Protocol Compared with MHRP80211 solution Min Hop Routing MHRP80211 A gtB and E gtD can happen A concurrently MINPOW quot A has to resort to C to send packets to B Then E gtD cannot happen 28 gControl Theory Based Approach Channel Model It is simple to use the a inverse 0 law path loss model is E 7 It will be rather complicated 710 5quot 39 when taking into account the time variance of the channel gain 5 1n 5 Travelled distance in The poxvcr gain 9t is moclclod as the DIJJLII of three colnponents path loss gp t shadow fading LIEU and iuultipaLli fadjiig t Here this is illustrated when uloV iiig on a referellce point and away froi 9 mm Er n t he t rallsiljit t e r 29 C0ntrol Theory Based Approach Receiver rr 1 39 Translllltter O 911tA 104 Ell vircul Figllre 3 Block diagl alll Of the receiver trallslnitter pair 739 xvhel elnploy39illg a gelleral SIRibasel 13C YEI control algoritlnlL In operation the controller result in a closed local loop 30 gControl Theory Based Approach Feedback based Power Control quot2 0 Psf 1 Gaff 14 Can Derive the closed loop system Rq 6quot qpwlt1 Me Time delay can be compensated for using the Smith predictor Predict the power gain to improve the reactions so as to decrease the disturbulance 31 iControl Theo ry Based 015solid TS T 05dashed S With Smith Predictor dark Without Predictor light Approach 32 A 1 1 b m m 1 and r 2 I V KJ LJn 9 1O 20 30 40 SD GD 70 SCI 90 1 0 50 7 5 0 7 5 Predictive control irnproxre the perfor Figure rnance by 1edicting the disturbances a colltroller Vitll c stllrljztllce preclictioll solid and a 1ninj1nunlvariance c011troller cl151161 b T119 sillglei bit error repr95911tatior1 and 7 control with tlle Snlith predictor alld distur bance predictiOIl 33 Conclusion Power Control can be addressed as a cross layer design problem which involves a multi dimensional optimization Introduced the impact of power control on a variety of parameters and phenomenon and then presented fundamental design principles Introduced power control protocols achieving successful power saving but sometimes at the expense of a reduction in the sense of other metrics Put power control algorithms into a control theory context 34 Reference Kawadia V Kumar PR Principles and protocols for power control in wireless ad hoc networks Selected Areas in Communications IEEE Journal on Volume 23 Issue 1 Jan 2005 Pages76 88 Krunz M Muqattash A Sun Ju Lee Transmission power control in wireless ad hoc networks cha enges solutions and open issuesNetwork IEEE Volume 18 Issue 5 Sept Oct 2004 Pages8 14 Fredrik Gunnarsson Fredrik Gustafsson Power control in Wireless Communications Networks From a Control Theory Perspective Cautionary Aspects of Cross Layer Design ConteXt Architecture and Interactions httQiwwweasasuedujunshanICCKumarICCQdf S Narayanaswamy V Kawadia R S Sreenivas and P R Kumar Power control in ad hoc networks Theory Architecture Algorithm and implementation of the COMPOW protocol in European Wireless Conference 2002 35