An open-source multi-agent ecosystem for household decision-making simulation, connecting scientists, developers, users, and contributors through reusable AI agents, datasets, tools, and governance.
HDSim will build a secure, extensible, and community-driven platform for modeling household decisions that shape transportation, disaster resilience, economic growth, energy systems, and public health. Led by Xilei Zhao with co-PIs Yushun Dong, Allison Reilly, and Yuanyuan Lei, the project will turn the existing PEMAND prototype into reusable research infrastructure for transparent model development, rigorous evaluation, STEM training, and sustained community innovation.
A scientist and developer module helps contributors create, evaluate, and integrate new household decision-making applications. An external user module provides accessible simulation capabilities for stakeholders through specialized agents coordinated by an orchestrator. PEMAND supplies the core methodology for persona modeling, structured negotiation, and validation.
This window embeds the full previous PEMAND Simulation Replay, including scenario selection, playback speed, Run Simulation, reset, phase progression, agent cards, negotiation streams, moderator scoring, process logs, and replay notes.
HDSim will make household decision simulation more cumulative: every new application, dataset, behavioral theory, annotation protocol, agent skill, or domain module can strengthen the shared infrastructure.
Static demographic data becomes household-aware personas grounded in behavioral theory, then agents negotiate collective outcomes.
Trip planning, residential mobility, and post-disaster relocation will provide reusable templates for future contributors.
Human-calibrated moderation, annotation rubrics, and benchmark workflows support transparent and reproducible validation.
Tutorials, workshops, developer challenges, and backbone partnerships will support adoption beyond the funding period.
HDSim will grow through coordinated roles: researchers bring theories and data, software contributors improve infrastructure, students and early-career researchers build reusable skills, and external users turn practical questions into simulation workflows.
Open-source the HDSim infrastructure, governance model, documentation, and contribution pathway around the existing PEMAND prototype.
Develop trip planning, residential mobility, and post-disaster relocation applications as reusable templates for future domains.
Promote adoption through tutorials, workshops, NHERI events, AI Days developer challenges, and open-source backbone partnerships.