Convergence improvement in repeating weighted boosting search algorithm for channel estimation

The fifth generation mobile communication systems are expected to support the Internet-of-Things (IoT) applications. One of the problems in massive connections is the assignment of pilot sequences. For supporting a large number of the IoT devices, non- orthogonal sequences have to be assigned and a receiver in a base station needs to estimate channel impulse responses (CIRs) by solving a joint optimization problem. In this paper, a new channel estimation scheme for the IoT devices is proposed. The proposed scheme is based on a repeating weighted boosting search algorithm and modifies the generation process of candidate CIR vectors for faster convergence in search iterations. Instead of calculating the sum of weighted candidate CIR vectors, the proposed scheme generates a candidate CIR vector that is assumed to be closer to the global optimum. In this case, the proposed scheme improves the convergence rate as compared to the conventional scheme though the mean square errors (MSE) of the solutions is worse. If the number of subcarriers is limited, the MSE of the estimated CIR vectors with the proposed algorithm is equivalent to that with the conventional scheme. Thus, the proposed scheme is less complex and suitable for massive connections. Numerical results obtained through computer simulation have shown that the proposed scheme achieves faster convergence by about 17-35% as compared to the conventional scheme.