What question did this study set out to answer?

The research aims to address the challenge of balancing performance and power constraints in Edge AI devices powered by solar energy.

January 18, 2026Open Access

Dynamic Power Management in Edge AI: A Sustainable Self-Adaptive Approach

View Full Paper

Ask AI

Bookmark

The research aims to address the challenge of balancing performance and power constraints in Edge AI devices powered by solar energy.
Conducted a parameter study on Raspberry Pi hardware to quantify power consumption and detection confidence.
Developed a custom Reinforcement Learning environment utilizing Kernel Density Estimation and physics-based models.
Trained a Proximal Policy Optimization agent within a Constrained Optimization framework to maximize performance.
The PPO agent outperformed static and random baselines in managing the trade-off between resource use and detection confidence.
Achieved up to 1.5 hours longer operational time compared to power-hungry baselines.
Provided over 40 percentage points more satisfaction of service level agreements than the least power-hungry static policy.

No takes yet. Share an insight, caveat, or question.

This thesis demonstrates a sustainable self-adaptive method for power management in Edge AI devices using a reinforcement learning agent.

Julia Oberauner (2025) studied this question.

View Full Paper

Ask AI

Bookmark