Self-Organizing Scheme for Machine-to-Machine Networks Based on Cross-Entropy Method

In the near future, thousands of machines would be deployed and large-scale interconnections between these machines would construct a massive network. In this study, a clustered network structure is considered for which the machines in a macro cell are divided into clusters, and the machines belonging to the same cluster communicate to the cluster head which then aggregates the traffic and relays to the macro base station. The clustered M2M network spatially reuses the radio resource. Thus it causes inter-cluster interference. How to properly configure transmission power to mitigate inter-cluster interference is an important issue. In addition, since humans also share the same radio resource with machines, the occupied resource of machines should be as fewer as possible under the given quality of service requirements. However, the amount of machines is huge. It is difficult to obtain parameters of all clusters to mitigate inter-cluster interference by heuristic methods. The cross-entropy (CE) method is adopted to search optimal parameters. The CE method was proposed to estimate the probability of rare events. Afterwards, it was applied to solve combinatorial optimization problems. Since the evaluations of CE method can be individually performed in distributed machines and then be collected to a center, it is suitable for M2M networks. From simulation results, the proposed scheme can obtain sub-optimal parameters of transmission power to minimize the occupied radio resource within few iterations.