Minimizing the impact of prediction errors during anticipatory resource allocation

By predicting changes in service requirements and future evolution of network states, anticipatory networking aims at improving the performance of decision making and optimization algorithms. In this paper, we propose a reliability-aware scheduling model which exploits information of upcoming users' channel states to proactively optimize the resource allocation. Along with the optimal formulation, proved to be NP-hard, a heuristic is also presented. The evaluation of the proposed optimization model and corresponding heuristic is carried out assuming both a perfect prediction and an error-prone prediction of the achievable users' channel states and show that both are robust to errors.