Paper Conference

Proceedings of BSO Conference 2022: 6th Conference of IBPSA-England


Multi-agent learning of incremental housing development strategies for solar utilisation in Peru

Sergio Edgar Mauricio Poco-Aguilar, Parag S Wate, Darren Robinson,
The University of Sheffield

Abstract: Incremental housing, whereby an initial core unit is constructed and occupied by a family, who then progressively enlarge it during its lifetime, is a common form of housing provision in Peru, and many other countries in the Global South. However, these complex enlargement decisions, which place a significant financial burden on homeowners, are likely to be sub-optimal in terms of energy and cost effectiveness. To address this, we have developed a new computational workflow, combining geometry generation, multi-agent reinforcement learning and energy modelling to support incremental housing owners to optimise their enlargement decisions. In this paper, we describe this new prototypical workflow and its application to two use cases: a single housing unit, surrounded by a static neighbouring scene and a single housing unit that dynamically interacts with a changing neighbouring scene. In both cases, we arrive at stable solution that maximise solar energy availability at least cost.
Keywords: Multi-agent learning, Incremental housing, Parametric modelling