After the work on CSI compression was completed, as we planned in our proposal, we started working on other problems, like CSI estimation, and novel data modulation techniques. We are making progress in multiple fronts, and have had a success in CSI estimation. To be more specific, we have found a method to estimate the CSI of multiple transmitters by using the fact that the CSI can be approximated as a linear combination of base sinusoids.
Our method, called ParEst, can immediately replace the existing CSI estimation methods in Wi-Fi and LTE/5G NR. For example, in LTE, to estimate the channel states, the node may transmit a peak on each antenna in the De-Modulation Reference Signal (DMRS). A cutoff method can be used, i.e., the signal around each peak can be carved out for each transmitter. However, the tail of each peak will be cutoff, leading to errors. In contrast, given the same received signal, ParEst estimates the CSI from all antennas jointly by solving an optimization problem. The complexity is very low, because the CSI is approximated as the linear combination of complex sinusoids on constant frequencies, and the coefficients of the base sinusoids are found by minimizing the approximation error. As the base sinusoids are on fixed frequencies, many steps, such as matrix inversion, can be precomputed.
ParEst is much more accurate and efficient than the existing methods for estimating the CSI from multiple antennas. As MIMO and MU-MIMO have become increasingly more important in Wi-Fi and cellular networks such as LTE and 5G NR, ParEst is expected to have a significant impact in wireless commutation and networking.