Research Areas
Our research group is dedicated to advancing intelligent systems through the integration of deep learning, reinforcement learning, semantic ontologies, IoT-enabled data infrastructures and edge computing.
A WORK TO INNOVATION AND SUSTAINABILITY
Innovation and Sustainability
Smart Lab is a community of engineers, students, and academics working together to develop innovative solutions for smart buildings through IoT and reinforcement learning technologies.
Reinforcement Learning for Smart HVAC Systems
Deep Reinforcement Learning (DRL) techniques optimise how HVAC systems operate to balance thermal comfort, energy consumption, and indoor air quality (IAQ).
- Multistage Control: Optimising air conditioning and industrial fans together.
- Algorithmic Advances: Testing algorithms like DQN, PPO, and SAC in Sinergym-EnergyPlus environments to improve accuracy.
Transfer Learning & Meta-RL
Transfer from simulation to the real environment is a major challenge (Sim-to-Real). We use the Model-Agnostic Meta-Augmentative Meta-Reinforcement Learning (MAML-RL) framework to enable rapid adaptation to different climatic conditions. This allows our models to understand relationships between devices rather than just training on raw data.
Accelerated DRL Training via LSTM-Based Surrogate Simulation
To tackle the slow training cycles of physics-based simulators (like EnergyPlus), we introduced an LSTM-based surrogate simulator. This enables scalable experimentation and prototyping, achieving up to 57.6% reduction in RL training time.
Semantic Interoperability & Ontology
By developing a pipeline that generates RDF knowledge graphs from EnergyPlus outputs, we have enabled semantic modelling compatible with SAREF. This structure makes CO2 levels, occupant presence, and HVAC decisions semantically queryable, turning the relationship between parameters into mathematical data.
REAL Lab
Bridging the gap between simulation and reality using Reinforcement Learning.
Quick Links
Contact Us
A-403, Dept of EEE, METU
Ankara, Turkiye