Participatory Urban Digital Twins with Fairness Aware Optimization for Equitable Quality of Life Improvements in Multi Objective Planning Contexts

Abstract

This study presents a participatory urban digital twin that embeds fairness-constrained multi-objective Bayesian optimization to co-design budget-feasible investments that raise quality of life (QoL) while bounding neighborhood shortfalls. The research addresses the gap between technically capable urban digital twins and decision processes that deliver distributionally equitable outcomes. The objective was to quantify how equity-aware optimization, paired with transparent explanations, alters citywide QoL trade-offs under fiscal limits. A city-scale twin for Riyadh integrated environmental sensing (PM₂.₅, temperature, wind, noise), 15-minute accessibility modeling from OpenTripPlanner, and Highway Safety Manual empirical-Bayes safety prediction. District-level QoL indices were constructed for thermal comfort, clean air, accessibility, and street safety; fairness was encoded as average shortfall constraints relative to the city median. Gaussian-process surrogates with q-Expected Hypervolume Improvement guided portfolio search; SHAP attributions and integer-constrained counterfactuals supported stakeholder deliberation. Under a 246.0-million SAR plan, median UTCI exceedance hours fell by 160 (−13.1%; 95% CI −14.9% to −11.2%; p<0.001), normalized accessibility rose by +0.08 (95% CI +0.06 to +0.10; p<0.001), and EB-adjusted injury crashes declined by −2.7 per 10,000 residents (−11.1%; 95% CI −3.4 to −2.0; p<0.001), while PM₂.₅ decreased modestly by −1.1 µg·m⁻³ (−4.8%; p=0.03). Inequality narrowed (Atkinson 0.154→0.112; average shortfall 0.071→0.048). The optimizer outperformed evolutionary baselines (hypervolume 0.618 vs 0.552; fairness-feasible share 0.68 vs 0.42). SHAP and counterfactuals coincided with higher stakeholder trust (+0.8 points; Z=3.80; p<0.001) and majority selection of fairness-constrained portfolios. These findings indicate that equity-formalized optimization within a digital twin can translate complex trade-offs into implementable plans with measurable QoL gains; future extensions should incorporate emissions-control actions and hybrid surrogate–mechanistic models to strengthen air-quality responsiveness.