Component-based computation-energy modeling for embedded systems

Abstract

Computational energy-efficiency is a critical aspect of many modern embedded devices as it impacts the level of autonomy for numerous scenarios. We present a component-based energy modeling approach to abstract per-component energy in a dataflow computational network executed according to a given scheduling policy. The approach is based on a modeling tool and ultimately relies on battery state to support a wider range of energy-optimization strategies for power-critical devices. CCS Concepts• Hardware → Power estimation and optimization; • Computing methodologies → Modeling and simulation; • Computer systems organization → Embedded systems; Multicore architectures.

Mechanical and Computational Energy Estimation of a Fixed-Wing Drone

“…The computation can be adjusted dynamically in the sense that the system can evaluate how different configurations will affect the energy consumption, and by implementing a scheduling strategy accordingly (we presented an example of a component-based scheduling strategy in our previous work [2]). The computational energy model could furthermore be integrated into a component-based framework, such as the ROS middleware, or the TeamPlay components toolchain [22].…”

“…Both object detection and encryption were further merged arithmetically, to model battery SoC of using the two components concurrently (Figure 4). In a concrete implementation of the computational energy model, a layer 2 model can be automatically derived through powprofiler for either separate components that are subsequently analyzed (an approach that we investigated in our previous work [2]) or for components running in parallel. In the latter, the modeling time is higher since components combinations are evaluated empirically.…”

“…The information about SoC can be also used to evaluate the energy cost of different components, or components executing in a larger component-based framework, like the one outlined in our previous work [2]. Moreover, SoC is of particular interest for mobile robots where it can be used to assess dynamically energy efficiency of software featuring autonomous tasks, and thus to correlate the battery state to the level of autonomy.…”

“…The information can be propagated to dynamically adjust component configurations to meet specific energy targets, as the drone might limit its QoS in order to meet specific mission-dependent criteria. To this extent, we presented a modification of a real-time scheduling algorithm [3] in our previous work [2], that can be used to assess the energy cost of software components running in a larger, constrained network where one software component input might depend from the others. The scheduler is statically generated in this context, which allows meeting real-time requirements of many mobile robot scenarios.…”

“…Our approach is implemented in a profiling tool named powprofiler. 1 The tool, written in C++ and distributed under MIT license, has been further extended from [2], [4] to support the mechanical energy model described in this paper, as energy estimation of mobile robots often requires an expected energy value as a function of the mission time. The current implementation allows deriving an energy model that includes, but is not limited to, SoC from an arbitrary number of software components and a given flight mission.…”