Non-uniform pilot-symbol allocation for closed-loop OFDM

We present a new perspective on the use of channel state information (CSI) at the transmitter of orthogonal frequency-division multiplexing (OFDM) systems. While conventional closed-loop OFDM systems attempt to pre-equalize the channel through power loading, we propose an average signal- to-noise ratio (SNR) maximizing strategy based on the allocation of pilot locations. Our derivation for the average output SNR of OFDM systems with interpolated channel estimation reveals a combinatorial optimization problem in terms of the pilot- symbol locations. Using this framework, we arrive at the SNR gain achieved in using uniform and non-uniform pilot locations. Further, we present simple sub-optimum solutions based on decoupled optimization and show how minor adjustments of these solutions can also lead to very favorable error rate performances. The proposed solutions are simulated in a wide range of Doppler frequencies and compared to uniform pilot allocation in terms of average symbol error rate.