Battery-powered wireless vision sensors can reduce installation cost and enable deployments where external power is impractical. But once we add edge AI for on-device image processing—both to protect privacy and to minimize bandwidth—the power budget becomes the central design constraint and the main source of technical risk. In this talk, we’ll share a practical approach to designing battery-powered edge AI sensors, starting with how we build and validate a power budget: choosing battery chemistry, estimating usable capacity, characterizing performance over temperature and age and forecasting consumption across operating modes. We’ll then cover key architectural decisions that determine feasibility, including sensor and regulator selection, memory trade-offs (RAM vs. flash vs. MRAM vs. ReRAM), NPU/accelerator choices, memory sizing, switched power domains, battery end-of-life behavior and wireless connectivity. Attendees will leave with a repeatable checklist for reducing risk and avoiding late-stage power surprises.

