Stochastic Gradient Descent Based MIMO Detection with Interference Rejection Combining

Multiple-input multiple-output (MIMO) technology promises high spectrum utilization and system capacity and has been seen as a key technology for modern wireless communication systems. With the rapid increase in the demand for system capacity, the interference between adjacent cells cannot be ignored, and the antenna scale is constantly increasing, which makes it increasingly difficult for uplink MIMO receivers to achieve the cubic complexity of the MMSE-IRC algorithm. Some methods based on series or approximate solutions provide quadratic complexity but have difficulty achieving interference suppression. To solve this issue, we proposed a method for calculating the equalization matrix based on the stochastic gradient descent algorithm. We theoretically proved that it can converge to the equalization matrix of the MMSE-IRC algorithm. After that, we experimentally demonstrated that its performance loss is within 0.2dB and the complexity does not exceed the quadratic, which provides a potential solution for reducing the complexity of uplink receivers in massive MIMO systems.