Automated aggregation of dwelling units and traffic cores in high-rise residential floor plans using genetic algorithm and multi-agent cooperative deep Q-network

Abstract

With the increasing demand for high-rise residential buildings (HRBs), traditional manual design processes involving multiple revisions and expertise encounter design efficiency challenge. Although several approaches have been proposed to automatically generate HRB standard floors using predefined component libraries, adaptive solutions across diverse design scenarios remain limited. This paper presented an automated aggregation method for dwelling units and traffic cores to construct complete standard floor plans using a genetic algorithm (GA) and multi-agent cooperative deep Q-network (MACDQN). First, dwelling units and traffic cores are represented using information masks and vectors. Then, GA is introduced to optimize the orientation of dwelling units. Finally, MACDQN is proposed to automatically aggregate dwelling units and traffic cores while meeting various design objectives. A comprehensive empirical study confirms the efficiency of the proposed method in producing practical and innovative layouts from authentic designs for various objectives, highlighting its potential to advance HRB floor plan design automation.