How AI can strengthen urban resilience in real time

Disparities in data availability and monitoring capacity have left many cities exposed to infrastructure failures they cannot predict in advance. While some urban centers deploy advanced analytics, others rely on limited or fragmented information as climate risks intensify.

A new study published in Frontiers in Artificial Intelligence examines these gaps. Titled Multimodal AI Fusion for Infrastructure Resilience: Real-Time Urban Analytics Framework Aligned with SDG-9, demonstrates how AI models can deliver reliable resilience insights even in data-scarce urban settings.

Moving beyond reactive infrastructure management

Traditional models treat infrastructure as static assets that degrade predictably over time. Maintenance schedules, inspections, and repairs are often planned months or years in advance, with limited ability to adapt to sudden shocks. According to the authors, this approach fails in modern cities where infrastructure systems are tightly interconnected and exposed to dynamic risks.

Transport networks depend on drainage systems, which in turn interact with weather patterns, land use, and population density. A failure in one subsystem can quickly propagate across others. The study argues that resilience must therefore be modeled as a dynamic property that evolves in real time, rather than as a fixed design threshold.

To address this challenge, the researchers propose a multimodal AI fusion framework that integrates multiple streams of urban data. These include temporal data such as traffic flow, rainfall, and sensor readings, as well as spatial data that capture how infrastructure elements are connected across a city. The goal is to move from isolated predictions to system-level awareness.

Under the hood, the framework features a combination of Long Short-Term Memory neural networks and Graph Neural Networks. LSTM models are used to capture temporal patterns and anomalies, such as sudden changes in traffic congestion or drainage capacity. Graph Neural Networks, meanwhile, model the spatial relationships between infrastructure nodes, allowing the system to understand how stress in one location may affect others.

This hybrid approach reflects how cities actually function. Infrastructure failures are rarely confined to a single point. Instead, they spread along physical and operational connections. By embedding these connections into the learning process, the AI system can anticipate cascading risks before they fully materialize.

Real-time resilience scoring and decision support

The researchers develop a dynamic Resilience Scoring Index that translates AI outputs into actionable resilience scores for individual infrastructure components. These scores reflect the likelihood of failure, current stress levels, and the potential impact on surrounding systems.

This scoring mechanism is designed to support real-time decision-making. City authorities can use it to prioritize inspections, allocate maintenance resources, and plan interventions before failures occur. The authors argue that this represents a shift from reactive crisis management to anticipatory governance.

Crucially, the framework does not assume perfect data availability. The study acknowledges that many cities, particularly in the Global South, lack dense sensor networks or consistent historical records. To test robustness, the researchers evaluated the model in three contrasting urban environments: Singapore, Chennai, and Rotterdam.

Singapore represents a highly instrumented smart city with extensive real-time data. Chennai reflects a data-scarce city facing acute infrastructure stress and climate exposure. Rotterdam serves as an example of a city with mature flood resilience systems and long-standing investment in infrastructure planning.

Across all three cases, the multimodal AI framework demonstrated strong predictive performance. The hybrid LSTM–GNN model consistently outperformed traditional machine learning and statistical forecasting methods. Notably, the performance gap widened in data-scarce environments, where spatial learning through graph structures helped compensate for missing or noisy data.

The findings suggest that advanced AI models, when designed to reflect real-world system interdependencies, can reduce the digital divide in urban resilience planning. Rather than benefiting only data-rich cities, AI-based resilience analytics may offer particular value where traditional monitoring is weakest.

Aligning AI with sustainability and governance goals

The study draws focus to policy alignment and governance relevance. The framework is explicitly designed to support the United Nations Sustainable Development Goal 9 (SDG 9), which focuses on building resilient infrastructure, promoting inclusive industrialization, and fostering innovation.

According to the authors, many AI applications in urban planning remain disconnected from sustainability objectives. They produce technical insights without clear pathways to policy action. By embedding SDG-9 targets directly into the analytics pipeline, the proposed framework links AI outputs to internationally recognized resilience and development benchmarks.

This alignment has practical implications. Infrastructure investments are increasingly evaluated against sustainability criteria by governments, development banks, and international donors. A system that quantifies resilience in real time and ties it to SDG targets can support evidence-based funding decisions and regulatory compliance.

The study also addresses governance challenges associated with AI deployment. It argues that real-time resilience analytics must be transparent, interpretable, and adaptable to local contexts. Rather than replacing human judgment, the system is designed to augment it, providing planners and engineers with timely, system-level insights.

The authors caution against viewing AI as a standalone solution. Institutional capacity, data governance, and inter-agency coordination remain critical. AI can identify risks, but effective response depends on organizational readiness and political will. The framework is therefore positioned as an enabling tool within broader governance structures.

Furthermore, the study focuses on scalability. Urban infrastructure systems vary widely in size, complexity, and regulatory context. The researchers argue that the modular design of their framework allows it to be adapted across cities without requiring uniform data standards or infrastructure layouts. This flexibility is essential for global applicability.

Overall, infrastructure failures disproportionately affect vulnerable communities, who often live in high-risk areas and have limited access to services. A real-time resilience system can help identify these vulnerabilities and support more equitable allocation of resources.

The study acknowledges ethical and operational challenges. Data privacy, algorithmic bias, and uneven access to digital infrastructure remain concerns. The authors call for responsible deployment practices, including stakeholder engagement and continuous evaluation.