TRANSMISSION POWER CONTROL FOR AD HOC WIRELESS NETWORKS: THROUGHPUT, ENERGY AND FAIRNESS Lujun Jia; Xin Liu; Noubir, G.; Rajaraman, R.; Wireless Communications and Networking Conference, 2005 IEEE. Presenter: Han-Tien Chang 1

Outline  Introduction  Network Models And Assumptions  δ -PCS: A Class of Power Control Schemes  Simulation Results  Implementation Issues  Conclusion And Future Work  Comments 2

Introduction  Introduce a new power control scheme  Combines collision avoidance and spatial reuse  Significant improvements for network throughput and energy efficiency simultaneously  Adhere to the single-channel, single-transceiver design rule  Solve the fairness problem 3

Introduction (cont’d)  Drawbacks of IEEE  uses maximum transmission power P max regardless of the distance between the transmitter and the receiver  Spatial channel reuse in IEEE is not optimized One on-going transmission may unnecessarily block multiple nearby sessions by transmitting at P max  Fairness problem IEEE delivers more packets for short distance traffic pairs than for long-distance traffic pairs When the network load increases, the ratio of delivered short distance traffic to long-distance traffic increases 4

Introduction (cont’d) 5  Related power control schemes [13] may suffer from  The scheme improves energy-efficiency but not network throughput, or increase the throughput at the expense of energy consumption  Extra hardware and spectrum availability are required, i.e., multiple wireless channels and transceivers  Strong assumptions on MAC or physical layers are imposed, which are often difficult to implement [13] A. Muqattash and M. Krunz. A single-channel solution for transmission power control in wireless ad hoc networks. In ACM MobiHoc, May 2004.

Introduction (cont’d) 6  δ -PCS  A novel transmission power function P(t) to compute an appropriate transmission power, so that a better spatial channel reuse is achieved Unlike POWMAC [13], no collision avoidance information is explicitly advertised in our scheme. Instead, nodes choose a transmission power level based on its traffic distance d, and an estimate of the interference level it experiences.

Introduction (cont’d) 7 [8] E.S. Jung and N.H. Vaidya. A power control MAC protocol for ad hoc networks. In ACM Mobicom, September [11] J. Monks, V. Bharghavan, and W. Hwu. A power controlled multiple access protocol for wireless packet networks. In IEEE INFOCOM, April 2001.

Network Models and Assumptions 8  P (r) = c · P (t) /d α where c is a constant that depends on the antenna gains and heights, and carrier frequency, d is the distance that the signal travels, and α is the power attenuation factor. The typical value of α ranges from 2 to 4. For our simulation study, we adopt the standard two-ray ground model that sets α to be 4 for long-range distances and 2 for short-range distances.

Network Models and Assumptions (cont’d) 9  The bit-error rate of a transmission depends on the noise power and the interference level at the receiver Let {X k, k ∈ T} be the set of nodes simultaneously transmitting at any time instant. Let X j be the receiver of a transmitter X i ∈ T For the transmission by X i to be successfully received by X j : (1) the received power P (r) at node X j must exceed the receiving threshold, RX th (2) the SINR at X j must exceed the SINR threshold, SINR th

Network Models and Assumptions (cont’d) 10  We use traffic distance or traffic length to denote the distance between the transmitter and the receiver.  The transmission range of a certain power level P (t) denotes the maximum distance at which the received signal is right above RX th  Beyond the transmission range, the signal can still be detected (but not decoded) if its strength is above the carrier sensing threshold, CS th. Typically, CS th is several dBs lower than RX th. Thus, the sensing range defined by CS th is larger than the transmission range

δ -PCS: A Class of Power Control Schemes 11  Overview of δ -PCS  1. Given the maximum transmission power level, P max The corresponding maximum transmission range d max satisfies c · P max /d max α = RX th.  2. Given the distance d between the transmitter and receiver, the minimum necessary transmission power, Pmin(d), for a transmission to be successful satisfies c · P min /d α = RX th.

 Let parameter δ be any constant between 0 and α our power function is the following  By substituting RX th with c · P max /d max α Let d T be the transmission range of P (t). By the definition of transmission range, the received signal power at d T is equal to the receiving threshold, i.e., P (r) = c · P (t) /(d T ) α = RX th. We then have that the transmission range of P (t) is d T = d δ / α d max 1 − δ / α. Note that d ≤ d T ≤ d max, for any δ between 0 and α. δ -PCS: A Class of Power Control Schemes 12, δ ∈ [0, α ]

δ -PCS: A Class of Power Control Schemes 13 How the transmission power changes as a function of the traffic distance d, under different δ -PCS The transmission range δ -PCS that lies between d and d max Main Goal: identify a ”good” δ value such that the corresponding power control scheme yields performance improvement in network throughput, energy efficiency and fairness simultaneously

δ -PCS: A Class of Power Control Schemes 14  Analysis of δ -PCS  analyze the fairness behavior of δ -PCS, the fairness for different traffic distance is closely related to the aggregate throughput  Assumptions Each source node is located independently and uniformly in the Euclidean plane For each source, its destination is located at a distance chosen independently and uniformly at random from [0, d max ]. N = 0 in our following derivation for simplicity

δ -PCS: A Class of Power Control Schemes 15  Consider an on-going transmission from a node X i to node X j For any other transmitting node X k, k≠i, let X k’ be the receiver. In order for the transmission from X i to X j to be successful, we need SINR j to exceed SINR th. We determine the value δ that minimizes E[1/SINR j ]. We note that minimizing E[1/SINR j ] is only an approximation for maximizing the aggregate throughput.

δ -PCS: A Class of Power Control Schemes 16  Lemma 3.1: The minimum of this equation is achieved at δ OPT (d ij )= 1/ (ln d max − ln d ij ) − 1 We can thus infer the following (1) This equation is a decreasing function of δ on [ − ∞, δ OPT ], and an increasing function of δ on [ δ OPT,∞] (2) δ OPT is an increasing function of d ij (3) There exist threshold distances d’, d’’ with δ OPT (d’) = 0 and δ OPT (d’’) = α

δ -PCS: A Class of Power Control Schemes 17  Three observations ( 討論函數的增減性 ) Observation 1: For 0 < d ij ≤ d’, δ OPT (d ij ) < 0, which implies that E[1/SINR j ] increases as δ increases from 0 to α Thus, the throughput of the traffic pairs with 0 < d ij ≤ d’decreases when δ increases from 0 to α Observation 2: For d’<d ij ≤ d’’, 0 ≤ δ OPT (d ij ) ≤ α, which implies that E[1/SINR j ] first decreases then increases when δ increases from 0 to α. This indicates that the throughput of the traffic pairs with d’≤ dij ≤ d’’ first increases then decrease when δ increases from 0 to α. Observation 3: For d’’ α, which implies that E[1/SINR j ] decreases when δ increases from 0 to α. This indicates that the throughput of the traffic pairs with d ij > d’’ increases when δ increases from 0 to α.

Simulation Results 18  Simulation model  GloMoSim-2.03 Physical layer Application layer CBR traffic model (pkt size: 512 bytes) ParameterValue Bandwidth2Mbps Receiving Threshold-64dBm Carrier sensing threshold-71dBm P max 25dBm d max 250m Carrier sensing range550m

Simulation Results (cont’d) 19  Topology and traffic Generate random topologies with 200 stationary nodes distributed on a 2000 × 2000m 2 area Select randomly located single-hop transmitter-receiver pairs also referred to as traffic pairs Select random traffic pairs from 0 to 250m  Performance metrics Aggregate and normalized throughput Energy efficiency (Mb/Joule) Throughput achieved in different destination ranges

Simulation Results (cont’d) 20  Random topologies with varying number of traffic pairs Aggregate throughput under varying number of traffic pairs with fixed data rate of 1.0Mbps. Normalized throughput under varying number of traffic pairs with fixed data rate of 1.0Mbps.

Simulation Results (cont’d) Distance in m 1.The preference is slowly inversed when δ increases from 0 to 4, with 4-PCS showing a strong preference for long traffic pairs. 2.for destination range m, the achieved throughput first increases then decreases 3.for destination range m, the achieved throughput increases 4.When δ is equal to 2 or 2.5, the achieved throughputs on each range are close to the average, thus a fair allocation of the channel capacity is observed

Simulation Results (cont’d) 22 Normalized bit-meter/sec measurement under varying number of traffic pairs Data delivered per unit energy under varying number of traffic pairs

Simulation Results (cont’d) 23  Random topologies with varying data rate Aggregate throughput under varying data rate. The total number of pairs is 30 Achieved throughput on different destination ranges (30 traffic pairs at data rate 600Kbps)

Simulation Results (cont’d) 24 Normalized bit-meter/sec under varying data rate, where the total number of pairs is 30. Data delivered per energy unit under varying data rate. The total number of pairs is 30.

Implementation Issues 25  Power function implementation  Within the RTS/CTS handshake protocol the communicating nodes can estimate the channel gain. We propose to convert the distance variable of the power function into a gain variable.  RTS/CTS power level update In a mobile network the channel characteristics change as the nodes move, therefore the transmission power levels have to be updated. We propose to use a technique similar to the closed-loop power control used in CDMA cellular systems.

Conclusion And Future Work 26  Propose δ -PCS  a class of power control schemes for ad hoc wireless networks, based on a novel transmission power function  Compared with IEEE achieves up to 40% throughput increase improves energy efficiency by a factor of 3 shows better fairness with respect to the traffic length distribution.  Future work  plan to integrate our power control scheme within a resource efficient multi-hop routing protocols

Comments 27  Experiment Discussion  We should know what we want to find out in this experiment or prove the hypothesis  Analysis  Experiment