What the study found
The study found that a Double Deep Q-Network (DDQN), a reinforcement-learning method, can be used to adjust CPU frequency in edge computing when renewable energy supply is uncertain. The authors report that this approach learned near-optimal frequency adjustment policies that balanced computation latency, energy consumption, and renewable energy availability.
Why the authors say this matters
The authors say this matters because renewable energy variability makes it difficult to maintain efficient and sustainable Internet of Things ecosystems powered by edge servers. They conclude that real-time CPU control under energy constraints is needed for sustainable edge computing.
What the researchers tested
The researchers combined convex optimization with a DDQN framework and formulated CPU control as a stochastic Markov Decision Process, a decision-making model that accounts for randomness over time. They tested adaptive CPU frequency scaling for edge computing under renewable energy uncertainty through simulation.
What worked and what didn't
According to the simulation results, the proposed DDQN policy reduced prediction error by 35%. The abstract also reports an optimal balance at 1.8 GHz with a lower energy-latency product, and says the method improved energy storage utilization, CPU throughput, and edge resource efficiency. The abstract presents conventional convex optimization methods as less able to handle stochastic, time-varying renewable energy supply and dynamic workloads.
What to keep in mind
The evidence described here comes from simulation results, so the abstract does not report real-world deployment. The abstract does not provide detailed limitations beyond noting that traditional convex optimization methods struggle with stochastic, time-varying conditions.
Key points
- A DDQN-based method was used to adapt CPU frequency in edge computing under uncertain renewable energy supply.
- The study reports a 35% reduction in prediction error.
- The abstract identifies 1.8 GHz as the point with the best balance and a lower energy-latency product.
- The authors say the approach improved energy storage utilization, CPU throughput, and edge resource efficiency.
- Conventional convex optimization methods are described as less effective for stochastic, time-varying renewable energy and workload conditions.
Disclosure
- Research title:
- DDQN improved CPU frequency control under renewable energy uncertainty
- Authors:
- Adeb Salh, Mohammed A. Alhartomi, Ghasan Ali Hussain, Saeed Alzahrani, Ahmed Alzahmi, Fares Suliaman Alromithy, Hock Guan Goh, N. M. Shah
- Institutions:
- Tun Hussein Onn University of Malaysia, Universiti Tunku Abdul Rahman, Universiti Tunku Abdul Rahman, University of Kufa, University of Tabuk, University of Tabuk, University of Tabuk, University of Tabuk
- Publication date:
- 2026-04-06
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with scholarly research without reading academic papers — one filtered digest, every Friday.