Kalman Smoothing for Irregular Pilot Patterns; A Case Study for Predictor Antennas in TDD Systems

A Semi-Linear Approximation of the First-Order Marcum Q-Function With Application to Predictor Antenna Systems

“…The feasibility of PA setups, which are of interest particularly in public transport systems such as trains and buses, but potentially also for the more design-constrained cars, has been previously shown through experimental tests [43]- [49]. Particularly, as shown in testbed implementations, e.g., [45], [48], with a two-antenna PA setup a normalised mean square error of around -10 dB can be obtained for speeds up to 50 km/h, with measured predictions horizons up to three times the wavelengths.…”

“…Moreover, the European project deliverables, e.g., [57,Chapter 2], [58, P. 107] and [59,Chapter 3], have well addressed the feasibility of the PA concept in network-level design. Finally, different works have analyzed the PA system in both frequency division duplex (FDD) [45], [46], [48] and time division duplex (TDD) [44], [49] systems, with some developments on addressing the system challenges such as antenna coupling [47], [50], spatial mismatch [50], [51], and spectrum underutilization [52], [53].…”

“…Such inaccurate channel state information (CSI) estimation will affect the system performance considerably at moderate/high speeds [44], [51]. One way to compensate for this mismatch problem is by interpolating channel samples at the base station (BS) [44], [45], [49]. Using all RAs as PA in the first time slot, [44] proposes an interpolation-based scheme for multiantenna systems with beamforming.…”

“…Then, [45] shows that an important part of the implementation of the PA is to identify the sample that is closest in space, and if necessary, interpolate between previous estimates. Also, in [49] Kalman smoother is used to interpolate larger distances that might occur in TDD. However, interpolation comes with additional overhead due to the increased number of pilots/PAs, and this overhead becomes more severe under FDD setup.…”

“…This may lead to performance losses, in terms of end-to-end throughput. Also, studies in [44], [45], [49] have not included analytical analysis for the PA system.…”