Archinect's Lexicon focuses on newly invented or adopted vocabulary within the architectural community. For this installment, we're featuring a term relevant to the recent wave of generative AI tools released for public use in the AEC sector and beyond.

"Reinforcement Learning" (RL) is a subset of machine learning where an agent learns to make decisions by taking actions in an environment to maximize some notion of cumulative reward. The agent learns from the consequences of its actions, rather than from being explicitly taught, and adapts its future behavior based on what it learns.

The agent's learning process involves exploring its environment, taking actions, and getting rewards or penalties. Over time, it builds a policy, which is a strategy to choose actions based on the state it is in, aiming to maximize the total reward over a period of time.

Reinforcement learning has potential relevance to building design in multiple ways:

• Optimized Design Solutions: Reinforcement learning can be used to optimize the design of buildings. By modeling different design choices as actions and the desirability of the resulting building design as rewards, a reinforcement learning algorithm could learn to make design choices that lead to more desirable outcomes. This could be particularly useful in areas such as energy efficiency, where the algorithm could learn to make design choices that minimize energy use.
• Autonomous Building Systems: Reinforcement learning can be applied to autonomously manage and control building systems like HVAC, lighting, and energy systems. By taking actions and learning from their results, such a system could learn to operate a building's systems in ways that maximize energy efficiency and occupant comfort.
• Structural Optimization: Reinforcement learning can be used to automate the process of structural design, wherein the algorithm can learn the most optimal arrangement of materials and structures that balance between safety, functionality, and cost.

This article is part of the Archinect In-Depth: Artificial Intelligence series.