Optimal Assignment of Resources for Distributed Computing in Real-Time Applications

Distributed real-time applications often require an optimal assignment of resources to improve performance. This paper proposes a methodology to optimally assign system resources that enables to minimize the processing and communication time. In this study, we defined a case of study partitioned into six subsystems to be simulated with four available processing units. We used undirected graphs to obtain a representation of the system and subsequently solved the resulting NP-hard problem as a mixed-integer quadratic program (MIQP). We also implemented a comprehensive search as groundwork and compared both methods using computational and global time as metrics. Numerical simulations showed that our methodology obtained both a better assignment of computational resources and significant solution time reduction than the comprehensive search. Moreover, the solution increased the rates of shared information between units during the reconciliation process. This methodology can thus be used in applications like distributed state estimation, distributed control or co-simulation.