Computation Offloading: Minimizing Energy Consumption

Limited Computing Power as a Challenge

The rapid growth of compute-intensive mobile applications – such as real-time object recognition, augmented reality and interactive media – places heavy demands on the processing power and energy supply of modern smartphones. Their limited resources often reach their limits. For this reason, researchers have long explored offloading computational tasks to Edge or Cloud servers to reduce energy consumption. Previous studies show that offloading often saves more energy than local execution [1–3], but they mostly rely on unrealistic assumptions, such as executing single tasks, fixed relationships between data size and computational complexity, or ideal network conditions. It remains unclear whether these benefits hold under more realistic scenarios.

Offloading as a Sustainable Solution

To address these gaps, Wei Geng's team conducts empirical experiments under previously unexplored conditions and demonstrates that offloading still saves energy when the team relaxes these assumptions. They analyze not only single tasks but also sequences of tasks, where naively offloading all tasks can be suboptimal due to the long-tail energy consumption [4] of modern networks. Since smartphones can also execute tasks locally on the Central Processing Unit, the Graphics Processing Unit or the Neural Processing Unit, the team develops a strategy that decides which tasks the device offloads and which it executes locally.

The team proposes a novel graph-based scheduling algorithm that computes near-optimal execution plans for task sequences, significantly reduces energy consumption and outperforms existing approaches in extensive empirical evaluations.

These findings have important societal implications: increased energy efficiency in mobile computing reduces battery drain, extends device lifespan and lowers electronic waste. By promoting sustainable use of computational resources in mobile and Edge infrastructures, this work directly contributes to energy savings and reducing the CO₂ footprint of digital services.