AI+Smart City
AI+Smart City
Embedded Files
As a Co-PI, I am leading the project to develop Reinforcement Learning based optimization algorithm to minimize energy consumption of building environment.
We are developing our idea to be applicable other domains in control problem specifically for building smart city.
As a Co-PI, I am leading the project to develop Reinforcement Learning based optimization algorithm to minimize energy consumption of building environment.
We are developing our idea to be applicable other domains in control problem specifically for building smart city.
CTRL-C (Conditioning buildings with Transferable Reinforcement Learning based Controllers)
Palo Alto Research Center, Inc. (PARC) with Oak Ridge National Laboratory (ORNL) proposes Conditioning buildings with Transferable Reinforcement Learning based Controllers (CTRL-C): a physics-guided deep reinforcement learning (DRL)-based automated building control system to learn interpretable, stable and transferrable decision-making rules (policies) to control HVAC by an iterative process through collective buildings in which it is deployed. The four overall goals of the effort are: (1) minimize total energy consumption from buildings, (2) maximize comfort of occupants of buildings, (3) make control policies human readable, and (4) remove cumbersome modeling requirements from the process.
CTRL-C (Conditioning buildings with Transferable Reinforcement Learning based Controllers)
Palo Alto Research Center, Inc. (PARC) with Oak Ridge National Laboratory (ORNL) proposes Conditioning buildings with Transferable Reinforcement Learning based Controllers (CTRL-C): a physics-guided deep reinforcement learning (DRL)-based automated building control system to learn interpretable, stable and transferrable decision-making rules (policies) to control HVAC by an iterative process through collective buildings in which it is deployed. The four overall goals of the effort are: (1) minimize total energy consumption from buildings, (2) maximize comfort of occupants of buildings, (3) make control policies human readable, and (4) remove cumbersome modeling requirements from the process.
Publications in preparation (* Equal contribution)
Sookyung Kim*, Chihyeon Song*, Weizhe Chen* et al. "Conditioning buildings with Transferable Reinforcement Learning based Controllers (CTRL-C)". Planned to submit to AAMAS-2023
Page updated
Google Sites
Report abuse