HKUST Combined Cross-Layer Design and HARQ for TDD Multiuser systems with Outdated CSIT Rui Wang & Vincent K. N. Lau Dept. of ECE The Hong Kong University of Science & Technology

HKUST Outline  Model of Multiuser System with HARQ-IR  Cross-Layer Problem Formulation  Cross-Layer Scheduler Design  Discussions & Simulation Results  Conclusions

HKUST HARQ Generally speaking, HARQ (Hybird Automatic Retransmission reQuest) is a retransmission technique to improve the packet receiving. There are two schemes of HARQ retransmission: –Chase Combining (CC): the retransmit packets are exactly the same, while the receiver combines the multiple copies of the packet to obtain a higher post-combining SNR. –Incremental Redundancy (IR): the information is first encoded into a long mother code, and then punctured into multiple blocks where the blocks will be sent in subsequent retransmission. We consider the scheme of incremental redundancy in this paper. InformationCoded Packet 2 nd Tx3 rd Tx1 st Tx4 th Tx MRC Combining Encoder: Tx: Rx:Achieve larger receiving SNR InformationSubpacket 2 nd Tx3 rd Tx1 st Tx4 th Tx Encoder: Tx: Rx: Subpacket Protect the packet by redundancy

HKUST System Model We consider downlink transmission of a multiuser system with one BS and K mobile users. –At each packet transmission, BS should select one mobile as target receiver. –HARQ-IR is supported, hence, if the target receiver cannot decode the packet, a NAK will be feed back to the BS. –The BS should retransmit the failed packet until the ACK feedback is received or the maximum number of transmission is reached. Let X be the transmitted symbol, the receiver symbol of the user k is given by the following equation. –We consider slow fading channel where the channel gain is quasi-static within a transmission event (the duration one packet transmission). BS User1 2 3 Received symbol Channel gain Noise

HKUST Model of Outdated CSIT --- Practical Issues In order to perform scheduling in the downlink transmission, the BS should have the knowledge of channel state information (CSIT). However, the CSIT estimated at the BS is usually outdated. A general model of CSIT error is given below: CSIT Error Actual CSI Estimated CSIT

HKUST Packet Error Model --- Subsequent of Outdated CSIT In slow fading channels, there are two reasons of packet error –Finite block length of channel coding [channel noise effect] –Transmitted data rate exceeding the instantaneous mutual information of the channel [channel outage] By applying strong channel coding (e.g. LDPC) with reasonable block length (e.g. 2k byte), it can be shown that Shannon’s limit can be achieved to within 0.05dB for a target FER of 10^{-2}.  the effect of channel noise can be ignored with strong coding. Yet, the second factor (channel outage) is systematic and will be the major contributor of packet error (esp when strong coding is used). Hence, we assume Packet Error Rate = Pr [r > mutual information]. To account for penalty of packet errors, we shall consider system goodput (b/s/Hz successfully delivered to the mobiles) as our optimization objective.

HKUST Average System Goodput The instantaneous throughput of a transmission event is The average system throughput of a transmission event is Since each transmission event may contain multiple channel use, we use the normalized average throughput (named as average goodput) as the system optimization objective. Average number of transmissions per transmission event Date rateI[] is 1 when the event is true and 0 otherwise.Summation due to IR retransmission

HKUST Outline  Model of Multiuser System with HARQ-IR  Cross-Layer Problem Formulation  Cross-Layer Scheduler Design  Discussions & Simulation Results  Conclusions

HKUST Diagram of Scheduler Cross-Layer Scheduler CSIT Data rate Power for each transmission We shall formulate this box as an optimization problem Selected user

HKUST Policies --- Actions of Scheduler The average system goodput is a function of the user selection policy A, power allocation policy P and rate allocation policy R. User selection policy A: determine the active user for each packet transmission according to the CSIT Power allocation policy P: determine the transmit power for active users according to the CSIT. Rate allocation policy R: determine the transmit data rate for active users according to the CSIT.

HKUST Problem Formulation The optimal user selection policy A*, the optimal power allocation policy P* as well as the optimal rate allocation policy R* are given by: –Subject to the following constraint: PER constraint: the packet error probability after the maximum number of transmissions should be ε. Average power constraint: the average transmit power cannot be large than P 0. Maximum number of transmission Power of j-th transmission Packet error rate after j-1 transmissions Average power per transmission event

HKUST Outline  Model of Multiuser System with HARQ-IR  Cross-Layer Problem Formulation  Cross-Layer Scheduler Design  Discussions & Simulation Results  Conclusions

HKUST Cross-layer Scheduler Design The asymptotical optimally scheduler design for sufficiently small target outage probability and sufficiently large SNR on each transmission is given by: –User selection: –Power allocation (power for j-th transmission): –Rate allocation: Outage probability of the j-1 th transmission Power constraint Constant related to L Target outage probability Estimation error

HKUST Outline  Model of Multiuser System with HARQ-IR  Cross-Layer Problem Formulation  Cross-Layer Scheduler Design  Discussions & Simulation Results  Conclusions

HKUST Discussions The average system goodput is given by –Goodput vs. Number of users K: The average system goodput scales in the order of ln(K), for small number of users K. –Goodput vs. Maximum number of transmissions L: The average system goodput scales in the order of ln(L). O(lnL) O(lnK)

HKUST Simulations

HKUST Conclusions In this paper, we study the combined design of cross-layer scheduling and HARQ for TDD multiuser systems with outdated CSIT in slow fading channel. We obtain the closed-form expressions for the average system goodput. –average system goodput scales in the order of O(ln L) at small K and target PER ε. – the average system goodput also scales in the order of O(ln K) for small K.

HKUST Thank You !