Joint Optimization of DNN Inference Delay and Energy under Accuracy Constraints for AR Applications

TL;DR

This paper proposes an iterative scheduling algorithm to optimize DNN inference delay and energy consumption in mobile edge computing for AR, while respecting accuracy constraints, addressing high computational and energy demands.

Contribution

It introduces a novel joint optimization framework for delay and energy in AR inference tasks, with an iterative algorithm for optimal offloading policy under accuracy constraints.

Findings

Optimized offloading policy reduces delay and energy consumption.

Demonstrated trade-offs between accuracy, delay, and energy in AR inference.

Proposed method improves efficiency in MEC-based AR applications.

Abstract

The high computational complexity and high energy consumption of artificial intelligence (AI) algorithms hinder their application in augmented reality (AR) systems. This paper considers the scene of completing video-based AI inference tasks in the mobile edge computing (MEC) system. We use multiply-and-accumulate operations (MACs) for problem analysis and optimize delay and energy consumption under accuracy constraints. To solve this problem, we first assume that offloading policy is known and decouple the problem into two subproblems. After solving these two subproblems, we propose an iterative-based scheduling algorithm to obtain the optimal offloading policy. We also experimentally discuss the relationship between delay, energy consumption, and inference accuracy.