Abstract
Generalised orthogonal frequency-division multiple-access (OFDMA) has been one of key air-interfaces for next-generation mobile communication systems. Allocated pilot symbols play a key role for the channel estimation and prediction that is essential for the design of OFDMA transceivers. In the past years, pilot-assisted channel estimation (PACE) has received extensive investigations particularly for the downlink. Many classical PACE schemes have been developed together with the pilot optimisation that minimises the noise-induced channel estimation error. However, in uplink communications, where polynomial interpolation schemes are usually employed, the interpolation error induced by the mismatch between interpolation models and actual channels often exists besides the noise-induced error. Conventional PACE approaches assume that the interpolation-error can be ignored. However, throughout the investigation of this thesis, it was found the interpolation error could not be compensated by increasing signal-to-noise-ratio (SNR), meanwhile it could be the dominative factor to affect the PACE performance in high-SNR or high-mobility environments. Therefore, investigation for uplink PACE by considering the interpolation error becomes an interesting issue, and it is addressed from the following two aspects in this thesis: First, it was found that interpolation errors could be minimised via pilot placement optimisation. Therefore, the optimum pilot placement for OFDMA uplink was investigated for chunk-based systems, where linear interpolation is often employed for PACE. Two interesting results were further found: 1. the pilot placement minimising the linear interpolation error is independent of channel environments under certain conditions; 2. the optimum pilot placement is a function of the SNR. However, the best SNR-independent pilot placement offering the near-optimum performance for practical systems were proposed. Second, the interpolation error can also be reduced by carefully designed interpolation algorithms. So, new PACE approaches that performed the linear interpolation for the amplitude and phase respectively were proposed. Another investigated issue in this thesis is the waste of channel-gain resource for OFDMA downlink when an OFDMA system is rich in pilots. In this scenario, an adaptive approach for joint placement of pilot and payload symbols was proposed to effectively improve the overall system performance with the pay of channel estimation performance. Key words: OFDMA, Chunk, Channel estimation, Interpolation error, Pilot placement, Link adaptation, Linear interpolation.