As Internet of Things (IoT) devices continue to generate massive amounts of data, edge computing must strike a balance between performance, energy efficiency, and adaptability. A recently defended PhD thesis explores how Approximate Computing (AxC) can optimize edge device performance by integrating inexact arithmetic into Coarse-Grained Reconfigurable Arrays (CGRAs).
By designing novel runtime-configurable inexact adders and multipliers, Hans Jakob Damsgaard’s research demonstrates that edge devices can dynamically adjust computational accuracy to save power—reducing consumption by up to 14.9% in key applications while keeping hardware overhead minimal. The work also introduces open-source libraries that lower the barrier to AxC research, enabling more efficient design exploration for future edge architectures.
This thesis paves the way for smarter, energy-efficient edge computing, ensuring IoT devices remain powerful yet sustainable.
Read the thesis here: https://trepo.tuni.fi/handle/10024/161786
