Cross-layer Optimal Decision Policies for Spatial Diversity Forwarding in Wireless Ad Hoc Networks

October 11 2006IEEE MASS 2006A. A. Abouzeid

Cross-layer Optimal Decision Policies for Spatial Diversity Forwarding in

Wireless Ad Hoc Networks

Prof. Alhussein Abouzeid

Rensselaer Polytechnic Institute

Joint work with Jing Ai & Zhenzhen Ye


Hussein Abouzeid

Jing Ai

(if we were here)

Zhenzhen Ye


Outline

• Spatial Diversity Forwarding– Motivation

– Related work

• Problem formulation of Optimal Stopping Relaying

• Implementation of OSR

• Evaluations of OSR– Analytical results

– Simulation results

• Summary


Spatial Diversity Forwarding: Motivation & Related Work

• Motivation: Exploiting spatial diversity inherent in multi-hop wireless networks– Observation: There naturally exist alternate next-hop relays with high

probability that one of them has favorable channel condition at any given instant!

• The key challenge is design of strategy for next-hop selection. Prior work can be classified into two classes:

1. FSR (First Stopping Relaying) [JD05, WZF04, etc.]– Selects the first relay that replies to the forwarding node. (Problem: though

minimum overheads incurred, the selected relay may be poor in quality.)2. LSR (Last Stopping Relaying) [SM05]

– Collecting CSI from all candidate relays and then selecting the best one. (Problem: though the selected relay might be (if CSI is not outdated) the best, it incurs the maximum delay.)

We propose OSR: Exploiting the range between these extreme cases.


Problem formulation of optimal stopping relaying (OSR)

• Optimal stopping relaying: investigate and design spatial-diversity forwarding policies based on a formally defined stochastic decision framework.

• Mapping next-hop selection problem to a sequential optimal stopping problem– The Decision maker: the forwarding node which has a packet to be

forwarded– Every time step, the decision maker observes a random variable which

is the state (e.g. channel quality) of the next available candidate. It also computes the reward up to that point in time.

– Action: whether to “stop” at a candidate next-hop relay and forward the packet to it, or “continue” observing other candidates

– Policy: The goal of the problem is to derive an optimal policy (i.e. a rule for deciding which action to take at every decision instant) in order to maximize the expected reward.


Problem formulation of OSR (cont.)• A “conceptual” decision making procedure of OSR

Channel Quality: 2

Channel Quality: 7

Channel Quality: 8

Channel Quality?

2

2 is low, continue!

Channel Quality?

7

7 is good enough, stop!

AB


Problem formulation of OSR (cont.)

• Given: a forwarding node ns which intends to forward a packet toward its destination and a set of L candidate next-hop relays {n1, n2, . . . , nL} known at the routing layer, which can be characterized by L independent discrete random variables (rewards) {Θ1,Θ2, . . . ,ΘL}.

• Problem: what is the optimal policy at the forwarding node ns to select the next-hop relay to which the packet is to be forwarded so as to maximize the expected reward E{Θ}?

• Θ is defined as d*Rdata, a generalization of Information Efficiency (IE)

• Solution: the optimal policy is a threshold-based policy– easy to implement


Implementation of OSR

• Physical layer: rayleigh fading channels– determined the channel quality statistics if known long-

term average

• MAC layer: an extended MAC anycast scheme based on IEEE 802.11– perform the “optimal stopping” decision-making procedure

• Network layer: Greedy geographic routing [BMSU99,BK00]– take charge of selecting a set of candidate next-hop relays

at a forwarding node


MAC layer anycast

• Motivation– transform “costly” sequential decision-making procedure at the

forwarding node to a relatively “cheap” parallel decision-making procedure on the relay side.

• MRTS-CTS dialogue– Multicast RTS (MRTS) carries the threshold-based policy– CTS comes from the relay selected by the optimal stopping policy

• Feedbacks collision resolution– it is possible that more than one candidate next-hop relay qualified to

relay the packet by performing the threshold-based policy issued by the forwarding node

– prioritized the responses of relays in an order pre-assigned by the forwarding node

– with CSMA, the response of a higher-priority relay can suppress the responses of lower-priority relays


MAC layer anycast (cont.)

• A sample timeline of the OSR scheme for three candidate next-hop relays

CTS1

NAV(MRTS)

NAV(CTS2)

MRTS

CTS2

DATA

ACK

Tx

Rx1

Rx2

Rx3

Others

CTS3

NAV(DATA)

SIFS

2SIFS

SIFS

SIFS

SIFS

Random back-offafter DIFS


OSR v.s. FSR: Analytical Results

• Scenario– X-axis represents the average

SNR– Y-axis represents the gain

OSR over FSR in terms of IE– L homogenous candidate

next-hop relays

• Main observation– FSR can not utilize more than

two candidate relays as efficiently as OSR, especially when channels quality is average


LSR v.s. OSR: Simulation Results (Qualnet)

• Comparison of end-to-end performance metrics in Qualnet [QUA]– Grid topology (8X8 grid), a single TCP flow

• throughput

– Random topology, multiple TCP flows (not included due to space limit)

– Static random topology, multiple UDP flows• packet delivery ratio• end-to-end delay• jitter

– Mobile topology, multiple UDP flows (not included due to space limit)


Impact of Fading Velocity (vm) on Average FTP Throughput

• Note: M is a protocol parameter in geographic routing that specifies the maximum order of spatial diversity that can be utilized by a forwarding node • Observation: In OSR, larger M means better performance (which is what we want). Not true for LSR (due to overheads)


Impact of Traffic Load on CBR flows


Summary & Future Work

• Formulated the next-hop relay selection problem as a sequential decision problem and derived the Optimal Stopping Relaying (OSR) policies for improving spatial diversity gain in wireless ad hoc networks

• Implement OSR in a realistic protocol stack• Both analytical and simulation results reveal that OSR

outperforms FSR and LSR in terms of IE/end-to-end performance metrics

• Enable nodes to learn the fading channel characteristics online instead of relying on a channel model

• Extend such a decision framework to flow/service differentiation schemes.


References

• [BK00] B. Karp and H T. Kung, “GPSR: greedy perimeter stateless routing for wireless networks,” in Proc. ACM/IEEE Mobicom 2000.

• [BMSU99] Prosenjit Bose, Pat Morin, Ivan Stojmenovic and Jorge Urrutia, “Routing with guaranteed delivery in ad hoc wireless networks,” in Proc. DIALM, 1999.

• [JD05] S. Jain and S. R. Das, “Exploiting path diversity in the link layer in wireless ad hoc networks,” in Proc. IEEE WoWMoM, 2005.

• [JZF04] J. Wang, H. Zhai and Y. Fang, “Reliable and efficient packet forwarding by utilizing path diversity in wireless ad hoc networks,” in Proc. IEEE Milcom, 2004.

• [QUA] “Qualnet 3.7 user’s guide.” http://www.scalable-networks.com/.

• [SM05] M. R. Souryal and N. Moayeri, “Channel-adaptive relaying in mobile ad hoc networks with fading,” in Proc. IEEE SECON, 2005.


Thank You!

• Questions?

[email protected]

Cross-layer Optimal Decision Policies for Spatial Diversity Forwarding in Wireless Ad Hoc Networks

Documents

Transcript of Cross-layer Optimal Decision Policies for Spatial Diversity Forwarding in Wireless Ad Hoc Networks