opportunistic routing for wireless ad hoc and sensor networks: present and future directions

IEEE Communications Magazine • December 2009 1030163-6804/09/$25.00 © 2009 IEEE

INTRODUCTION

Routing is crucial in wireless ad hoc as well assensor networks. The tasks of routing includeroute selection and packet forwarding. Routeselection is to select one or more routes con-necting a pair of nodes. Packet forwarding makesa one-hop decision on which neighbor should bechosen for forwarding a packet along the select-ed routes. The highly dynamic and lossy natureof the wireless medium makes routing in wire-less networks a challenging problem.

Traditional routing protocols for wireless net-works perform best path routing that preselectsone or more optimized fixed routes before trans-missions start and uses a fixed neighbor to for-ward a packet in each hop. This strategy doesnot adapt well to the dynamic wireless environ-ment where transmission failures occur frequent-ly, which would trigger excessive link-levelretransmissions, waste of network resources, oreven system breakdown.

Recently, opportunistic routing (OR) has

received a great deal of attention and researchwork. The key idea behind OR (also calledopportunistic forwarding) is to overcome thedrawback of unreliable wireless transmissionby taking advantage of the broadcast nature ofthe wireless medium such that one transmis-sion can be overheard by multiple neighbors.The forwarding task can continue as long as atleast one neighbor along a route receives thepacket. It has been shown that OR improvesperformance over that of traditional best pathrouting.

As an example, a directed graph in Fig. 1represents a wireless network in which a link(x,y) has a delivery probability P(x,y). The tra-ditional routing achieves only 20 percent end-to-end delivery probability for any possiblerouting path from node s to node t. However,an OR approach called relay-based opportunis-tic forwarding could achieve a delivery proba-bility of (1 – (1 – 20 percent)5) ≈ 67 percent ifall five neighbors of s are selected as relay can-didates. As another example, Fig. 2 illustrateshow opportunistic forwarding can affect anentire routing path. For clarity, the deliveryprobabilities for some links are not shown inFig. 2. It should be clear that each of links(s,B), (B,D), (D,t) has a 60 percent deliveryprobability, and (s,C), (C,t) each has a 40 per-cent delivery probability. A packet from sources may follow different paths to reach destina-tion t . Traditional best path routing wouldalways choose the most reliable link to forwarda packet to the next hop, which results in thepath s → A → B → C → D → E → t, which hasan (80 percent)6 ≈ 26.2 percent end-to-enddelivery probability with six hops. With OR, ifwe restrict OR to route data packets via pathswith at most three hops, there are four pathsmeeting this requirement: s → C → t, s → C →D → t, s → B → C → t, and s → B → D → t,which form an interleaved forwarding directedmesh connecting the s-t pair. The former twopaths have a successful delivery probability of

ABSTRACT

Opportunistic routing has recently attractedmuch attention as it is considered a promisingdirection for improving the performance of wire-less ad hoc and sensor networks. With oppor-tunistic routing, intermediate nodes collaborateon packet forwarding in a localized and consis-tent manner. Opportunistic routing greatlyincreases transmission reliability and networkthroughput by taking advantage of the broadcastnature of the wireless medium. In this article wefirst illustrate the basic idea behind opportunisticrouting, and then categorize current researchwork based on different criteria. We illustratehow different protocols work, and discuss theirmerits and drawbacks. Finally, we point outpotential issues and future directions in oppor-tunistic routing for wireless ad hoc and sensornetworks.

AD HOC AND SENSOR NETWORKS

Haitao Liu and Baoxian Zhang, Chinese Academy of Sciences

Hussein T. Mouftah, University of Ottawa

Xiaojun Shen, University of Missouri — Kansas City

Jian Ma, Nokia Research Center

Opportunistic Routing for Wireless Ad Hoc and Sensor Networks:Present and Future Directions

MOUFTAH LAYOUT 11/18/09 12:05 PM 03

IEEE Communications Magazine • December 2009104

PsC × (1 – (1 – PCt) × (1 – PCD × PDt)) = 40percent × (1 – (1 – 40 percent) × (1 – 80 per-cent × 60 percent)) ≈ 27.5 percent. Similarly,the last two paths have a successful deliveryprobability of 60 percent × (1 – (1 – 60 percent× 60 percent) × (1 – 80 percent × 40 percent))≈ 33.9 percent. The overall successful deliveryprobability by the above four paths is therefore1 – (1 – 27.5 percent) × (1 – 33.9 percent) ≈52.1 percent. In large wireless networks, theexample in Fig. 1 illustrates the question ofhow each sender should select a good for-warder set to move a packet downstream, andthe example in Fig. 2 illustrates the question ofhow to build an efficient forwarding meshstructure to improve the performance of OR inend-to-end packet delivery.

The key design issues in opportunistic routinginclude forwarder set selection and prioritiza-tion, and duplicate transmission avoidance/sup-pression. For instance, in our examples theforwarder sets for source s in Figs. 1 and 2 are{A, B, C, D, E} and {B, C}, respectively. In Fig.1 each forwarder candidate has the same priori-ty, while in Fig. 2 node C has higher priority iffaster end-to-end delivery is preferred. In thisarticle we discuss in detail how each of the aboveissues is tackled using different strategies.

CLASSIFICATION OF OPPORTUNISTICROUTING PROTOCOLS

Fundamental issues in the design of OR proto-cols include forwarder set selection, prioritiza-tion, and duplicate forwarding avoidance/suppression. Resolving these issues needs toconsider routing efficiency, protocol overhead,

compatibility with existing medium access con-trol (MAC) protocols, use of network stateinformation, location information, and use ofcoding function. By taking these factors as wellas different design strategies into consideration,we categorize the current major OR protocols asfollows.

END-TO-END VS. HOP-BY-HOPFORWARDER SET SELECTION

By end-to-end forwarder set selection, the setof forwarders can be determined once and forall (forwarder-set-related information can becarried by data packets [1–3], kept at interme-diate nodes [4], or determined on the fly on aper packet basis using certain criteria [5, 6]).End-to-end forwarder set selection may lead toduplicate transmissions since non-neighboringforwarders can make inconsistent decisions onpacket forwarding. With hop-by-hop forwarderset selection, each packet holder (including thesource) independently determines its own for-warder set along the path to the intended des-tination. A selected forwarder will againcontinue this process until the packet reachesits destination. Example protocols of this typecan be found in [7–9]. In general, end-to-endstrategy outperforms hop-by-hop strategy sincethe former can optimize the selection of a for-warder set with more network state informa-tion gathered. However, the hop-by-hopstrategy is easy to implement and scales well.The size of the forwarder set is also a factoraffecting the performance of an OR protocol.A larger size can increase the one-hop success-ful delivery probability, but may introducesome forwarder candidates with small progresstoward the destination.

METRICS FOR PRIORITIZATIONAfter a forwarder set has been selected, we needto assign priorities for these forwarders. Select-ing a good metric has a large impact on networkperformance. Forwarder candidates can be pri-oritized based on expected transmission count(ETX) [1, 2], hop count [6, 7], geo-distance [8,9], coding chances [3], and so on. Utilization ofhop count or ETX needs an underlying routingprotocol (either reactive or proactive) to gathersuch information. Geo-distance requires theavailability of location information of nodes. Acoding approach can be used to minimize thenumber of transmissions. How these metrics areintegrated into protocol design can greatly affectthe routing performance. Moreover, the accura-cy of a metric depends on the proper measure-ment of link quality and timely dissemination ofsuch information.

DISTRIBUTIVE COORDINATION FORPRIORITIZATION

In some mechanisms [1, 2, 4] the decision onpriorities is made by distributive coordinationamong forwarder candidates that, upon receivinga data packet, compete with each other to selectthe best one to continue the forwarding taskwhile the rest resign. This deferred choice giveseach transmission several opportunities to make

Figure 1. An illustration of relay-based opportunis-tic forwarding.

20%

20%

20%

20%

100%

100%

100%

100%

100%

A

B

C 20% s t

D

E

Figure 2. An illustration of path-based opportunistic forwarding.

tE D B

40%

C

60%

80%

s A

MOUFTAH LAYOUT 11/18/09 12:05 PM Page 104

IEEE Communications Magazine • December 2009 105

progress. However, it can introduce certain extrapacket delivery latency. Some other mechanisms[7, 8] may use explicit control packet(s)exchanged immediately before or after a datatransmission for the distributive coordination.For example, the request-to-send/clear-to-send(RTS/CTS) scheme can be used to select thebest among potential forwarders that respond ina priority order [8]. Upon receiving the firstCTS, the sender immediately starts to transmitdata to that node. This strategy can avoid dupli-cate transmissions. Another strategy is to enableforwarder candidates that, upon receiving a datapacket, send back a short acknowledgment(ACK) in a predetermined order. The first nodereturning an ACK wins and the others resign.This strategy requires that forwarder candidatesbe neighbors of each other such that the trans-mission of an ACK can be overheard by all ofthem. Among existing work, there are also somemechanisms [2, 4–6] that require no coordina-tion among forwarder candidates. In these mech-anisms data packets are broadcast to the air,which can largely ease the design of the MACprotocol at the penalty of increased duplicatetransmissions seen at destination nodes.

DETERMINISTIC VS. PROBABILISTIC FORWARDER SELECTION

Whether a forwarder candidate receiving a pack-et will be a true forwarder can be decided ineither a deterministic or probabilistic manner.Example protocols of the former type can befound in [3, 4, 8]. The Opportunistic Routing inDynamic Ad Hoc (OPRAH) networks protocol[4] requires each node in the prebuilt forwardingmesh structure to definitely serve as a forwarder.Here, a forwarding mesh structure is a collectionof nodes that form partially overlapped multiplepaths connecting a source-destination pair.Alternately, in [3, 8] the forwarder candidatewith the highest priority is selected as the nextforwarder. In contrast, by probabilistic forwarderselection [5, 6], upon receipt of a data packet, aforwarder candidate will independently decide aprobability with which it chooses itself as a for-warder. Probabilistic selection requires no coor-dination among forwarder candidates and ismore resilient to highly lossy and uncertain wire-less environments. However, duplicate transmis-sions can be significantly too many if the size ofthe forwarding set and forwarding probabilityare not properly chosen.

LOCATION-BASED VS.TOPOLOGY-BASED SELECTION

By location-based selection [8, 9], the forwardersand their priorities are determined with little orno global topology-based network state informa-tion. In contrast, topology-based selections [1–3]require network state information (either localor global) to enable efficient opportunistic rout-ing. With more information, a sophisticated butefficient metric can be used to help forwarderset selection and prioritization. In general, topol-ogy-based selection has better routing perfor-mance while location-based selection has betterscalability for large-scale wireless networks.

CODING VS. NON-CODING

Integration of OR with an appropriate codingstrategy can greatly improve the routing perfor-mance. For example, forward error correction(FEC) can reduce the packet loss ratio of links,which consequently affects forwarder set selec-tion and prioritization. Hybrid Automatic RepeatRequest (ARQ) [9] can increase the transmis-sion range and link layer transmission reliability.Integration of network coding and OR canreduce the total amount of transmissions byguiding packets to next hops with more codingchances [3]. In general, design of efficient cod-ing-aware OR protocols needs to consider topol-ogy, traffic pattern, and reduction of extraoverhead.

SURVEY ON OPPORTUNISTICROUTING PROTOCOLS

In this section we introduce some major ORprotocols. We explain how each of them works,and discuss their merits and drawbacks.

EXTREME OPPORTUNISTIC ROUTINGExtreme Opportunistic Routing (ExOR) [1] is astate-of-the-art OR protocol for wireless multi-hop networks and has been implemented on theRoofNet testbed at Massachusetts Institute ofTechnology (MIT). ExOR integrates routing andMAC protocols. It improves routing perfor-mance by utilizing long-range but lossy links.ExOR is designed for batch forwarding. Thesource node includes a forwarder list in eachpacket, prioritized by ETX distance to the desti-nation: the shorter the distance, the higher thepriority. Only those nodes that are closer to thedestination than the source are included in theforwarder set. Each packet has a BITMAPoption, which marks those packets that havebeen received by the sending node or nodes withhigher priorities. All packets are broadcast. Aforwarder transmits a packet only if no for-warder with higher priority has explicitlyacknowledged receipt of it, as indicated in theBITMAP position for this packet.

ExOR has good routing performance. How-ever, it also has the following drawbacks. First, itreduces spatial reuse because it enforces globalcoordination among forwarders. Second, for-warders connected with low-quality links or nolinks can make inconsistent decisions on packetforwarding because a forwarder may not hearthe acknowledgment from other forwarders withhigher priorities, which causes duplicate trans-missions.

OPPORTUNISTIC ANY-PATH FORWARDINGIn ExOR, the condition for a node to be select-ed into a forwarder set is simply that its ETXvalue to the destination is shorter than that ofthe source. This simple condition may cause datapackets to take low-quality routes. To overcomethis problem, Opportunistic Any-Path Forward-ing (OAPF) [10] introduces an expected any-path count (EAX) metric such that it recursivelycalculates the near-optimal forwarder set at eachpotential forwarder to reach the destination. The

Integration of OR

with an appropriate

coding strategy can

greatly improve the

routing performance.

For example,

Forwarding Error

Correction (FEC) can

reduce packet loss

ratio of links, which

consequently affects

forwarder set

selection and

prioritization.



extra cost paid is the high computational over-head and more network state information gath-ered.

MAC-INDEPENDENT OPPORTUNISTICROUTING AND ENCODING PROTOCOL

MAC-Independent Opportunistic Routing andEncoding Protocol (MORE) [2] integrates ORand intra-flow network coding and is targetedfor enhancing ExOR. Figure 3 illustrates howExOR can cause duplicate transmissions. In thisexample, after source s sends out packets a, b,Node B receives b, and node A receives a and b.Node B transmits first because it is closer to tthan A. Node A has the following three choices:forwarding a, b, or both a and b. There is 2/3probability of duplicate transmissions. In MOREnode A can forward a coded packet a ⊕ b. Thendestination t performs an XOR operation on thetwo received packets: a ⊕ b ⊕ b = a. Thus, noduplicate transmission occurs at t.

MORE works as follows. When the source isready to send, it keeps creating coded packetsvia random linear combination of the K nativepackets in the current batch. The source keepssending such coded packets out until the wholebatch is acknowledged by the destination, atwhich time the source proceeds to the nextbatch. In MORE, data packets are always codedand carry a list of forwarders and a code vectorrecording how the native packets are combined.Upon receiving such a coded packet, a node inthe forwarder list first checks for the innovative-ness of the packet (i.e., if it is linearly indepen-dent of the packets previously received). Aforwarder only stores innovative packets. Fur-thermore, each forwarder keeps a TX counter,which is calculated by a distributed algorithmbased on the concept of ETX. When a forwarderreceives an innovative packet from an upstreamnode, it increments the counter by its TX credit(0 < TX < 1), and when the MAC allows thenode to transmit, the node checks whether thecounter is positive. If yes, the node creates acoded packet, broadcasts it, then decrements thecounter by one. If the counter is zero or nega-tive, the node does not transmit. Once the desti-nation receives K innovative packets, it can

decode the whole batch. It then sends an ACKback to the source to allow it to move to thenext batch.

Different from ExOR, MORE uses the con-cept of innovative packets to judge whether areceived packet brings new information insteadof using duplicate packets as in ExOR. Moreover,it uses a TX counter at each forwarder to fur-ther reduce the amount of transmissions. Simu-lation results showed that MORE can greatlyreduce the total number of transmissions com-pared to ExOR.

INTEGRATION OF AOMDV AND ANYCASTJain and Das developed an anycast MAC proto-col to perform channel-state-based next hopselection among multiple next hop candidates[7]. The designed protocol is an extension of theIEEE 802.11 MAC layer. A sender first multi-casts an RTS to the multiple next hop candi-dates, which contains the addresses of all thereceivers. CTS transmissions are staggered intime in order of the priorities of the receivers,which can be based on hop count or queuelength at the receivers. Upon receipt of a CTS,the sender transmits DATA to the sender of theCTS after a short interframe space (SIFS) inter-val, which notifies the remaining receivers tostop sending their CTSs. In [7], this anycastMAC protocol is further integrated with a multi-path extension to the Ad Hoc On Demand Vec-tor (AODV) routing protocol, referred to asAOMDV. Experimental and simulation resultsshow that the integrated protocol can improvenetwork performance in time-varying radio envi-ronments.

OPRAHOPRAH [4] builds a braid multipath set betweensource and destination via on-demand routing tosupport opportunistic forwarding. For this pur-pose, OPRAH allows intermediate nodes torecord more subpaths back to the source andalso those subpaths downstream to the destina-tion via received Route Requests and RouteReplies. OPRAH can increase end-to-end for-warding reliability. In OPRAH the destinationmay receive duplicate data packets because abuilt route set may contain spatially disjointpaths or partially disjoint paths from an interme-diate node down to the destination.

RESILIENT AND OPPORTUNISTICROUTING SOLUTION FOR MESH NETWORKS

To increase the forwarding resilience and relia-bility in highly lossy multirate wireless networks,a resilient and opportunistic routing solution formesh networks (ROMER) [5] builds a forward-ing mesh on the fly and on a per packet basis. InROMER each packet carries an option indicat-ing how far it is allowed to deviate away fromthe shortest path from source s to destination t,denoted ExtraCost (ExtraCost ≥ 0). Let dis-tance_so_fari represent the distance a packet hasgone thus far from its source to the current nodei, and shortest_distance(x,y) the shortest distancefrom node x to node y. The precondition underwhich an intermediate node i overhearing atransmission is allowed to further transmit the

Figure 3. An illustration of potential duplicate transmissions in ExOR.

b

A

B

ts

b b a

b a



packet is as follows: distance_so_fari +shortest_distance(i,t) ≤ shortest_distance(s,t) +ExtraCost, which (approximately) creates anellipse-style mesh structure, centered on thelong-term shortest path from source to destina-tion. The mesh provides enough flexibility ofrich and interleaved paths to accommodate theshort-term radio channel dynamics and transientoutages. In [5], to suppress duplicate copies, theforwarding probabilities at intermediate nodesnot-on-shortest-path are adjusted based on theirdownstream links’ instantaneous throughput.Moreover, the integration of AOMDV and any-cast, OPRAH, and ROMER can support nodemobility due to the high resilience of their creat-ed forwarding structures.

DIRECTED TRANSMISSION ROUTING PROTOCOLDirected Transmission Routing Protocol(DTRP) [6] works in a way similar to ROMERfor creating a forwarding mesh and also on a perpacket basis but adjusts the forwarding probabil-ity at potential forwarders in a different way. InDTRP, rules include:• Nodes sitting on the shortest path from the

source to the sink forward each packet withprobability of one.

• Nodes not sitting on the shortest path for-ward a received packet with a probabilitydetermined by the minimum extra cost (i.e.,extra distance).The more the extra cost, the smaller the prob-

ability. In this way a high probability of end-to-end data delivery can be achieved. Theadvantages of ROMER and DTRP are theirsimplicity and high reliability in data delivery.Both DTRP and ROMER can provide redun-dant data copies on the forwarding mesh struc-ture, which is suitable for highly lossy wirelessenvironments.

GEOGRAPHIC RANDOM FORWARDINGGeographic Random Forwarding (GeRaF) [8] isa geographical forwarding protocol. It selects aforwarder set and prioritizes them using locationinformation. In GeRaF each packet carries thelocations of the sender and destination. Onlythose neighboring nodes closer to the destina-tion than the sender can be forwarder candi-dates. Moreover, these eligible candidates rankthemselves based on their geo-distances to thedestination. In this way the forwarder set andprioritization can easily be implemented via anRTS-CTS dialog at the MAC layer, which alsoensures that a single forwarder can be chosen.GeRaF adopts hop-by-hop forwarder set selec-tion. It is targeted for relatively dense networks.GeRaF is simple to implement. However, thecost for acquiring location information may betoo high to implement, at least in the nearfuture.

HYBRID ARQ-BASEDINTERCLUSTER GEOGRAPHIC RELAYING

Hybrid ARQ-Based Intercluster GeographicRelaying (HARBINGER) [9] is a combinationof GeRaF with hybrid automatic repeat request(ARQ). In GeRaF, when nodes work with ran-dom sleep scheduling for battery saving, each

transmission may need multiple handshakings atthe MAC layer before a packet relay is made;thus, GeRaF degrades for sparser networks. Thekey difference between GeRaF andHARBINGER is as follows. When there is noforwarder within the range of a sender, withGeRaF, everything must start over again suchthat the sender will make a new transmissionattempt. In contrast, in HARBINGER, hybridARQ is used for a receiver to combine the infor-mation accumulated over multiple transmissionsfrom the same sender. Specifically, when a nodedecides to receive packets, it keeps each packetit receives from the same sender so that oldinformation can be combined with fresh infor-mation gained after each new ARQ transmis-sion. Eventually, this receiving node will be ableto decode the packet. Accordingly, the moretransmissions made by a sender, the moreextended the communication range that can beachieved. HARBINGER provides a better ener-gy-latency trade-off than GeRaF, in particularwhen node density is low.

CODING-AWARE OPPORTUNISTICROUTING MECHANISM

Coding-Aware Opportunistic Routing Mecha-nism (CORE) [3] is an integration of localizedinterflow network coding and opportunistic rout-ing. Existing localized network coding mecha-nisms only adopt best path routing and passivelywait for the appearance of coding chances onthe path. This leads to inefficient use of networkresources. By integrating localized network cod-ing and opportunistic forwarding, CORE enablesa packet holder to forward a packet to the nexthop that leads to the most coding changes amongits forwarder set. Such recursive hop-by-hopoperations can greatly improve the end-to-endcoding gain in packet delivery with little extraprotocol overhead. Table 1 summarizes all theabove discussed protocols based on the criterialisted in the previous section.

CONCLUSION ANDFUTURE DIRECTIONS

In this article we have presented a survey on ORprotocols for wireless ad hoc and sensor net-works. Existing work shows that OR can largelyimprove the performance of wireless networks.Among the surveyed protocols, it is seen thatintegrating OR with network coding is an effec-tive strategy for improved performance. Further-more, localized coordination and locationinformation can ease the implementation of anOR protocol. However, research topics in ORare far from exhausted. Next, we list some poten-tial issues in this area.

NEW METRICS FOR ORAs stated earlier, the choice of metric has greatimpact on the performance of an OR protocoland also its complexity. Existing work has mainlyfocused on using ETX, hop count, codingchances, and related facets as primary metrics.In [11] Cui et al. introduced power consumptionas a primary metric in optimizing the design of

Existing work shows

that OR can largely

improve the

performance of

wireless networks.

Among the surveyed

protocols, it is seen

that integrating OR

with network coding

is an effective

strategy for

improved

performance.



an energy-efficient coding-aware OR protocolfor wireless networks. Wu et al. [12] proposed autility-based OR protocol, which uses the utilityof each packet as a routing metric. In [12] thesuccessful delivery of a packet brings benefit,and the utility of a packet equals its benefitvalue minus the transmission cost to reach thepacket destination.

JOINT SCHEDULING ANDOPPORTUNISTIC ROUTING

Transmission scheduling is to determine whichnode should transmit and also which packetshould be transmitted at an instant. The decisionshould consider the instant radio conditions atdifferent receivers. In contrast, route selectioncan affect how packets are routed through a net-work. Cui et al. [11] formulated the issue of jointopportunistic routing and scheduling when net-work coding is used as an optimization tech-nique.

CROSS-LAYER ORBesides transmission scheduling and route selec-tion, there are still some other factors such asmodulation techniques, packet size, transmissionrates, and transmission ranges that can largelyaffect the quality of wireless links and then, inturn, affect the selection of forwarder set andprioritization. Therefore, designing cross-layerOR protocols is an effective strategy to achieve adesirable trade-off among different measure-ments.

OPPORTUNISTIC MULTICAST ROUTINGMulticast is an important communicationsparadigm in wireless networks. The availabilityof multiple destinations in a multicast tree canmake the selection of forwarder candidates, dis-

tributed coordination among them, and relatedprioritization complicated. Thus far, little workhas been done in this topic.

OR IN MULTICHANNELMULTIRADIO NETWORKS

This involves the choice of an appropriate chan-nel and/or radio between a sender and potentialreceivers. In this sense, a sender may have dif-ferent forwarder candidates on different chan-nels/radios, and their prioritization can bedifferent as well. OR in multichannel multiradionetworks can potentially improve network per-formance and deserves in-depth investigation.

In summary, opportunistic routing is apromising direction to improve the performanceof wireless networks. Much work still needs tobe done in this area to achieve high performanceand low complexity for dynamic wireless net-works.

ACKNOWLEDGMENTThis work was supported in part by the Knowl-edge Innovation Program of the Chinese Acade-my of Sciences (no. KGCX2-YW-120) and NSFof China (no. 60702074). B. Zhang’s work waspartially supported by SRF for ROCS, SEM. X.Shen’s work was partially supported by K.C.Wong Education Foundation, Hong Kong.

REFERENCES[1] S. Biswas and R. Morris, “Opportunistic Routing in

Multi-hop Wireless Networks,” Proc. ACM SIGCOMM,Aug. 2005, pp. 133–44.

[2] S. Chachulski et al., “Trading Structure for Randomnessin Wireless Opportunistic Routing,” Proc. ACM SIG-COMM ‘07, Oct. 2007, pp. 169–80.

[3] Y. Yan et al., “Practical Coding-Aware OpportunisticRouting Mechanism for Wireless Mesh Networks,” Proc.IEEE ICC 2008, May 2008, pp. 2871–76.

Table 1. Classification of opportunistic routing protocols.

Protocols/properties

Forwarder setselection

Metrics forprioritization

Forwardercandidatescoordination

Deterministic orprobabilisticforwarder selection

Location ortopology-based

Codingor not

ExOR [1] End-to-end ETX ACK-based Deterministic Topology-based ×

OAPF [10] Hop-by-hop EAX ACK-based Deterministic Topology-based ×

MORE [2] End-to-end ETX N/A Deterministic Topology-based √

AOMDV+Any-cast [7]

Hop-by-hop Hop count RTS-CTS Deterministic Topology-based ×

OPRAH [4] End-to-end Hop count N/A Deterministic Topology-based ×

ROMER [5] End-to-end Hop count N/A Probabilistic Topology-based ×

DTRP [6] End-to-end Hop count N/A Probabilistic Topology-based ×

GeRaF [8] Hop-by-hop Geo-distance RTS-CTS Deterministic Location-based ×

HARBINGER [9] Hop-by-hop Geo-distance ACK-based Deterministic Location-based √

CORE [3] Hop-by-hop Code chances Data-based Deterministic Topology-based √



[4] C. Westphal, “Opportunistic Routing in Dynamic Ad HocNetworks: The OPRAH protocol,” Proc. IEEE MASS ‘06,Oct. 2006, pp. 570–73.

[5] Y. Yuan et al., “ROMER: Resilient Opportunistic MeshRouting for Wireless Mesh Networks,” Proc. IEEEWiMesh ‘05, Sept. 2005.

[6] M. S. Nassr et al., “Scalable and Reliable Sensor Net-work Routing: Performance Study from Field Deploy-ment,” Proc. IEEE INFOCOM ‘07, May 2007, pp. 670–78.

[7] S. Jain and S. Das, “Exploiting Path Diversity in the LinkLayer in Wireless Ad Hoc Networks,” Proc. IEEE WoW-MoM ‘05, June 2005, pp. 22–30.

[8] M. Zorzi and R. R. Rao, “Geographic Random Forward-ing (GeRaF) for Ad Hoc and Sensor Networks: MultihopPerformance,” IEEE Trans. Mobile Comp., vol. 2, no. 4,Oct.–Dec. 2003, pp. 337–48.

[9] B. Zhao, R. I. Seshadri, and M. C. Valenti, “GeographicRandom Forwarding with Hybrid-ARQ for Ad Hoc Net-works with Rapid Sleep Cycles,” Proc. IEEE GLOBECOM‘04, Nov.–Dec. 2004, pp. 3047–52.

[10] Z. Zhong et al., “On Selection of Candidates forOpportunistic AnyPath Forwarding,” ACM SIGMOBILEMobile Comp. Commun. Rev., Oct. 2006, pp. 1–2.

[11] T. Cui, L. Chen, and T. Ho, “Energy Efficient Oppor-tunistic Network Coding for Wireless Networks,” Proc.IEEE INFOCOM ‘08, Apr. 2008, pp. 361–65.

[12] J. Wu, M. Lu, and F. Li, “Utility-Based OpportunisticRouting in Multi-Hop Wireless Networks,” Proc. IEEEICDCS ‘08, June 2008, pp. 470–77.

BIOGRAPHIESHAITAO LIU is currently a full professor with the ShanghaiInstitute of Microsystem and Information Technology ofthe Chinese Academy of Sciences, where he is vice presi-dent. He received his Ph.D. degree from the University ofTechnology and Science of China in 1998. He is the generaldesigner in the sensor network field of the Chinese Acade-my of Sciences. He is Secretary-General of the Sensor Net-work Standardization Working Group of China. He is amember of the National Key Program Expert Group onnext-generation broadband mobile communication net-works. He is a recipient of the National Science & Technol-ogy Progress Prize of China (Second Class) (2005) and theOutstanding Science & Technology Achievement Prize ofthe Chinese Academy of Sciences (2005). He has publishednearly 100 technical papers in archival journals and confer-ence proceedings. His research interests cover variousaspects of sensor network systems.

BAOXIAN ZHANG [M] is currently a full professor with theGraduate University of the Chinese Academy of Sciences,Beijing. He is vice director of the Key Laboratory of Wire-less Sensor Network and Communications, Chinese Acade-my of Sciences. He received his Ph.D. degree from NorthernJiaotong University (now Beijing Jiaotong University),China, in 2000. He has served as Guest Editor of specialissues for IEEE JSAC, Wiley Wireless Communications andMobile Computing, and ACM Mobile Networks and Appli-cations. He served as Co-Chair for ChinaCom ’08 Advancesin Internet Symposium. He has served on technical pro-gram committees for many international conferences and

symposia such as IEEE GLOBECOM and ICC. He has pub-lished over 70 refereed technical papers in archival journalsand conference proceedings. His research interests covernetwork protocol and algorithm design, and wireless adhoc and sensor networks.

HUSSEIN T. MOUFTAH [F‘90] ([email protected]) joinedthe School of Information Technology and Engineering(SITE) of the University of Ottawa in September 2002 as aTier 1 Canada Research Chair Professor, where he becamea University Distinguished Professor in 2006. He was previ-ously with the Department of Electrical and ComputerEngineering at Queen’s University (1979–2002), where hewas prior to his departure a full professor and departmentassociate head. He served as Editor-in-Chief of IEEE Com-munications Magazine (1995–1997) and IEEE Communica-tions Society Director of Magazines (1998–1999), Chair ofthe Awards Committee (2002–2003), a member of theBoard of Governors (1997–2000 and 2006–present), andDirector of Education (2006–present). He is founding Chairof two IEEE Communications Society Technical Committees(TCs). He is the author or coauthor of six books, 30 bookchapters, more than 850 technical papers, 10 patents, and138 industrial reports. He is the joint holder of eight Out-standing/Best Paper Awards. He is a Fellow of the Canadi-an Academy of Engineering (2003) and EngineeringInstitute of Canada (2005), and Fellow RSC: The Academiesof Canada (2008).

XIAOJUN SHEN [SM] received his B.S. degree in numericalanalysis from Tsinghua University, Beijing, China, in 1968,and his M.S. degree in computer science from NanjingUniversity of Science and Technology, China, in 1982. Hereceived his Ph.D. degree in computer science from theUniversity of Illinois at Urbana-Champaign in 1989. He isa full professor with the School of Computing and Engi-neering, University of Missouri-Kansas City. He has pub-l ished over 60 refereed technical papers in archivaljournals and conference proceedings. His current researchinterests include computer algorithms, parallel processing,telecommunications, computer networking, and sensornetworking such as switching, routing, and packetscheduling.

JIAN MA received his B.Sc. and M.Sc. in 1982 and 1987,respectively, from Beijing University of Posts and Telecom-munications, and his Ph.D. degree in 1994 from theDepartment of Electronics Engineering, Helsinki Universityof Technology, Finland. He is now a principal member ofresearch staff atn Nokia Research Center, Beijing, and isresponsible for university cooperation in China and otherAPACs, Nokia Postdoc Working Station, and Nokia-TsinghuaJoint Research Laboratory. He has 11 patents granted, andis author or co-author of more than 150 publications injournals and conferences, as well as a couple of books andbook chapters. He is also adjunct professor of Beijing Uni-versity of Posts and Telecommunications, Graduate Univer-sity of the Chinese Academy of Science, Tongji University,and Beihang University. His current research interests areconsumer services on wireless sensor networks, mobileInternet applications, pervasive computing, and ad hoc andp2p networking.


opportunistic routing for wireless ad hoc and sensor networks: present and future directions

Documents