Toward Resilient and Equitable Mobility Systems with Dynamic Synthetic Populations

Research Project, 2026 – 2030

We aim to achieve beyond state-of-the-art static synthetic-population snapshots: dynamic, scenario-conditioned synthetic agents with generative multi-day activity–travel sequences that quantify urban resilience and equity across people, places, and time to inform sustainable spatial planning. We achieve this by coupling generative sequence AI with dynamic multi-source data fusion and lead-time-aware stress testing within a mobility digital twin. We test whether dynamic synthetic populations outperform static baselines in disrupted scenarios, whether warning lead times shift behavior systematically across groups, and whether policy levers can reduce unserved mobility and recovery times under equity constraints. Theoretically, we build on activity-based and time geography principles and agent-based modelling. Methodologically, we fuse multi-source data (mobile phone geolocations on mobility, census, transit, land use, networks) under privacy-by-design; train generative sequence models that produce feasible, multi-day plans respecting temporal and spatial constraints; and stress-test policies within a mobility digital twin with explicit resilience and equity metrics. A key innovation is lead-time-aware resilience: acute, little-to-no-warning events trigger immediate protective behaviours, e.g., evacuation, whereas forewarned hazards (e.g., heat waves) induce anticipatory adaptations (schedule shifts, trip chaining, remote activity). We calibrate these responses using before/after mobility evidence and draw on international case studies to obtain behavioral priors and external validation before carefully transferring them to Swedish contexts. Expected outputs include: (i) new knowledge on constructing updatable, privacy-preserving synthetic populations; (ii) methodological advances in high-fidelity micro-simulation of activity–travel chains; (iii) validated approaches for lead-time-aware behavior modeling with equity-sensitive indicators; and (iv) open benchmarks and reproducible pipelines that enable cumulative learning and cross-country comparison. Societal impact follows from enabling agencies to stress-test planning/policy interventions, protect access during shocks, and improve distributional outcomes, aligned with Sweden´s goals for climate adaptation, preparedness, and sustainable mobility. Ethics and privacy are ensured via data minimisation, secure environments, statistical disclosure control, and the release of only privacy-safe artefacts.