Path Structured Multimarginal Schr\"odinger Bridge for Probabilistic Learning of Hardware Resource Usage by Control Software

TL;DR

This paper introduces a novel probabilistic learning method based on path structured multimarginal Schrödinger bridge problems to predict hardware resource usage by control software, demonstrating rapid convergence and broad applicability.

Contribution

It applies recent algorithmic advances in structured MSBPs to learn and predict hardware resource distributions in control software, with guaranteed linear convergence.

Findings

Rapid convergence to accurate resource usage predictions

Effective in a model predictive control case study

Broad applicability to cyber-physical systems

Abstract

The solution of the path structured multimarginal Schr\"{o}dinger bridge problem (MSBP) is the most-likely measure-valued trajectory consistent with a sequence of observed probability measures or distributional snapshots. We leverage recent algorithmic advances in solving such structured MSBPs for learning stochastic hardware resource usage by control software. The solution enables predicting the time-varying distribution of hardware resource availability at a desired time with guaranteed linear convergence. We demonstrate the efficacy of our probabilistic learning approach in a model predictive control software execution case study. The method exhibits rapid convergence to an accurate prediction of hardware resource utilization of the controller. The method can be broadly applied to any software to predict cyber-physical context-dependent performance at arbitrary time.