PDA multiple model approach for joint channel tracking and symbol detection in MIMO systems

The optimal channel estimation approach for multiple-input multiple-output (MIMO) systems, which runs filters for all possible symbol combinations, requires exponentially growing resources with time and number of transmit antennas. The conventional suboptimal channel estimation approaches are in a decision-directed manner, i.e. estimating the channels using the symbols already detected. The drawback of this kind of method is that possible symbol-detection errors are not fully accounted for in the channel estimation. In the paper, a sub-optimal joint channel tracking and symbol detection method is proposed based on probabilistic data association (PDA) and generalised pseudo Bayesian (GPB) algorithms. The PDA principle is applied to reduce the model size at every time instant and the first order GPB algorithm (GPB1) is used to control the size of the filtering tree by combining the estimation result from different models at every time instant. Simulation results demonstrate that the proposed multiple model channel estimation algorithm (PDAMM) performs better than the conventional decision based single model channel estimation algorithm (PDAKal).