MPC with Machine Learning Applied to Resource Allocation Problem Using Lambda Architecture

Abstract

The resource allocation problem is the process of allocating limited resources for a vast amount of tasks. Within this problem there are several important variants such as the stochastic time-variant resource allocation problem. This problem is relevant within environments where the distribution of resources varies with time, bringing difficulties to forecasting. Related research generally address the problem by using model predictive control (MPC) techniques or machine learning (ML) algorithms. However, both can be applied together in order to improve the tasks prioritization and forecasting. Therefore, this paper proposes a solution using the concept of lambda architecture in order to tackle the time-variant and the distinct input information. First results show that the integration between MPC and ML prioritizes the resource allocation and a Markov chain model is capable of forecasting tasks, optimizing a strategic binary control. We analyze a case study of a real problem and show how the proposal was built and its advantages over the traditional method.