A Scheduling Model of Battery-powered Embedded System

TL;DR

This paper develops a dynamic hybrid model for battery-powered embedded systems, integrating task scheduling and battery capacity prediction to enable real-time analysis of system behavior.

Contribution

It introduces a unified hybrid modeling framework combining scheduling dynamics and battery capacity estimation for single processor systems.

Findings

The model predicts task scheduling behavior within finite time windows.

It accurately estimates remaining battery capacity based on discharging current.

The framework unifies CPS components under hybrid systems theory.

Abstract

Fundamental theory on battery-powered cyber-physical systems (CPS) calls for dynamic models that are able to describe and predict the status of processors and batteries at any given time. We believe that the idealized system of single processor powered by single battery (SPSB) can be viewed as a generic case for the modeling effort. This paper introduces a dynamic model for multiple aperiodic tasks on a SPSB system under a scheduling algorithm that resembles the rate monotonic scheduling (RMS) within finite time windows. The model contains two major modules. The first module is an online battery capacity model based on the Rakhmatov-Vrudhula-Wallach (RVW) model. This module provides predictions of remaining battery capacity based on the knowledge of the battery discharging current. The second module is a dynamical scheduling model that can predict the scheduled behavior of tasks within any finite time window, without the need to store all past information about each task before the starting time of the finite time window. The module provides a complete analytical description of the relationship among tasks and it delineates all possible modes of the processor utilization as square-wave functions of time. The two modules i.e. the scheduling model and the battery model are integrated to obtain a hybrid scheduling model that describes the dynamic behaviors of the SPSB system. Our effort may have demonstrated that through dynamic modeling, different components of CPS may be integrated under a unified theoretical framework centered around hybrid systems theory.