Reinforcement Learning (RL) is the science of decision making. It is about learning the optimal behavior — how to map situations to actions — in an environment to maximize the total reward. The algorithm is not told which actions to take, but instead must discover which actions yield the most immediate and subsequent reward by trying them and observing how the environment responds. This discovery process is quite similar to children exploring the world around them and learning actions that help them achieve a goal.
As it has a capacity to seek new and creative ways in unseen dynamics without the help of a supervisor, RL is an unique tool that has the potential to transform our world, especially to mitigate climate change. For instance, when we look at 36 solutions to halve emissions by 2030 in the Exponential Roadmap by J. Falk, what’s fascinating is that a proper optimization and control system powered by RL can be used for each of those items.
Hereby a few examples,
Turn the electricity system into a game. The board is the electricity grid, our pieces are energy-intensive devices on the grid, and our goal is to minimize the emissions produced by these devices whilst maintaining a quality of service to society.
RL algorithms can reduce the energy consumption by interacting directly with the heating/cooling system. It can adapt its learning continuously to the controlled environment using real-time data collected on-site (external temperature, energy prices, occupancy, etc) without the need to access a thermo-energetic model of the building.
The proliferating RL solutions in transportation are already facilitating the transition.
For cultivated seafood companies, RL controls the environment parameters for optimum growth. This reduces the need to add external growth factors into the tissue, which is how most cultivated meats are grown today, water pollution, etc.
In a next blog post, we’ll explain in detail the case Heating-Ventilation-Air Conditioning (HVAC) control system with RL.
Main image source: high tech artificial intelligence for climate change in a solar punk utopia (MidJouney)