To address the challenge of synergistically optimizing multiple conflicting objectives, energy efficiency, low carbon emissions, and occupant comfort, in indoor design, this study proposes an improved multi-objective particle swarm optimization (IMOPSO) framework. Tailored to the mixed heterogeneous nature of design variables, the method introduces a probability-driven discrete-variable update mechanism, integrated with adaptive parameter tuning and a dual-archive elitist guidance strategy, enabling automated, iterative coupling with building energy simulation tools. A case study based on a typical office space demonstrates that, compared with NSGA-II, MOEA/D, and standard MOPSO, the Pareto front obtained by IMOPSO achieves superior performance in generation distance (0.019) and hypervolume (0.745), with solution-set uniformity improved by approximately 13%. Analysis of optimized solutions reveals that, while maintaining equivalent thermal comfort, operational energy consumption can be reduced by 31.2% and whole-life-cycle carbon emissions by 22.5%. This study provides a quantitative decision-support tool for performance-driven, low-carbon indoor design.