Power scheduling for MSE minimization with peak and average power constraints

This paper considers the problem of precoding design and power scheduling in a multiple-input multiple-out (MIMO) system, assuming channel state information (CSI) availability at both the receiver and the transmitter. We derive an optimal power scheduling policy in time and space to minimize the ergodic summation of the mean squared error (MSE) subject to both peak (short-term) and average (long-term) power constraints. By applying a rotation matrix to the MIMO precoder, we also propose a suboptimal power scheduling policy to improve the bit error rate (BER) performance. We then obtain a closed-form solution to the power scheduling strategy for identically and independently distributed (i.i.d.) Rayleigh fading channels. Numerical results shows that the peak power constraint does not have a great effect on the power scheduling policy at the mid-signal-to-noise region. Numerical results also show the improvement in terms of MSE and BER performance by the proposed power scheduling policy over allocation schemes in space only.