Urban Infrastructure Enhancement Using Multi-Objective Genetic Algorithm in Sustainable City Planning
Annotatsiya
Sustainable urban infrastructure planning requires balancing objectives that cannot be reduced to a single scalar score without an implicit, often unexamined, value judgement about their relative importance: capital cost, environmental sustainability, commute efficiency, climate resilience, and social equity each matter, and improving one frequently requires sacrificing another. NSGA-II (Non-dominated Sorting Genetic Algorithm II), the multi-objective evolutionary algorithm of Deb et al. (2002), avoids collapsing these objectives prematurely by evolving a population toward the Pareto front—the set of solutions for which no objective can be improved without worsening another preserving the full trade-off structure for human decision-makers to evaluate, rather than committing to a single weighting before the search even begins. This paper applies NSGA-II to a 288-variable infrastructure-allocation problem (48 city zones × 6 intervention types: green space, public transit, bicycle/pedestrian infrastructure, stormwater management, EV/clean-energy infrastructure, and affordable-housing density bonuses) spanning a city of 840,000 residents and a $480 million infrastructure budget, optimising simultaneously across five objectives (capital cost, sustainability score, mean commute time, climate resilience score, and social equity score). Across 300 generations (population 200), NSGA-II converges to a final Pareto front of 84 non-dominated solutions and a hypervolume of 0.684, exceeding weighted-sum scalarisation (0.524; +30.5%, which by construction returns only a single solution regardless of population size), greedy zone-by-zone allocation (0.424; +61.3%), and random allocation (0.284; +140.8%). The city council selected a balanced solution from the front (capital cost $284 million) that improves sustainability score from a baseline of 42.4 to 74.8 (+76.4%), reduces mean commute time from 38.4 to 29.6 minutes (-22.9%), raises climate resilience from 38.8 to 68.8 (+77.3%), and raises social equity score from 52.4 to 68.4 (+30.5%), with low-income zones (35.0% of city population) receiving 43.0% of allocated budget—a genuinely progressive allocation rather than a merely nominal equity score. A Year 1 interim review of phased implementation, a gentrification-risk screen, an objective-weighting sensitivity analysis, and a post-hoc engineering feasibility audit each surface specific limitations bounding the model's planning-horizon validity and the council's selection process.
Hali tarjima qilinmagan