The Urban Intelligence & IoT Lab (UI²) at KFUPM turns city-scale sensing and data into safe and secure decisions that benefits public safety and the environment - with the verification, validation, and safety assurance to deploy AI in cities at scale.
UI² is a research and development laboratory at the Interdisciplinary Research Center for Smart Mobility and Logistics, focused on the safe prototyping, deployment, and scaling of the intelligence, sensing, and safety-assurance layer of urban mobility, transportation, logistics, and infrastructure systems.
We develop the methods that turn IoT and connected-infrastructure sensing, urban data fusion, digital twins, and learning-based prediction into decision-making that operates safely alongside human users. Our particular emphasis: how AI-enabled systems are verified, validated, and safely operated at scale.
Working closely with partnering labs, research centers, and other stakeholders who build the autonomy stack itself, UI² provides the simulation environments, instrumented testbeds, benchmarks, and assurance evidence aligned with emerging standards such as ISO/PAS 8800, UL 4600, and the NIST AI Risk Management Framework.
To become a leading regional and internationally recognized hub for scalable and trustworthy urban intelligence — providing the methods, tools, and evidence that allow AI and IoT systems to be deployed safely, reliably, and accountably across mobility, logistics, and critical infrastructure.
Distributed and event-driven sensing for roads, transit, utilities, and logistics networks — the data layer of the intelligent city.
High-fidelity urban digital twins fusing live sensor data with simulation for scenario testing, benchmarking, and rehearsal before deployment.
Data-driven forecasting and optimization for traffic, freight, and infrastructure operations — designed for the constraints of real deployments.
Sequential and risk-aware decision-making for systems that share streets, skies, and networks with people.
Rare-event and long-tail evaluation, uncertainty quantification, runtime monitoring, and the assurance evidence needed for certification and regulatory acceptance.
How does a small perception error become a citywide incident? We trace failure paths of AV stacks across various testbeds simulators, including high-fidelity perception, urban network-scale traffic, and fast closed-loop studies for mitigating rare cascade failures, quantifying their likelihood, and validating runtime monitors and failsafe interventions before anything reaches a public street.
Teacher-guided reinforcement learning for lane keeping and full-loop driving in a Duckietown-style environment, the software stack for our upcoming physical Duckiebot city, where algorithms graduate from simulation to hardware.
Safety-aware deep RL for quadrotor flight in confined, GPS-denied spaces, tunnels, pipes, tight urban corridors, with barrier-inspired constraints on clearance and motion, built on gym-pybullet-drones; plus MDP benchmarks for aerial surveillance planning.
Real streets have rusty, faded, occluded signs. We build data augmentation and evaluation suites that stress perception against real-world infrastructure degradation, and point toward IoT-based monitoring of the infrastructure itself.
The city itself can sense: a traffic-light camera watches vehicles queue, cross, and clear a CARLA Town10 junction while pedestrians use the crosswalks, the first eye of an instrumented-city stack feeding digital twins and city-scale decision-making.
When agents share a network, failures spread like contagion. We model adversarial and benign cascades across multi-agent systems, derive detection bounds, and design topologies that contain the spread.
From open-pit haulage fleets to multi-echelon supply chains: simulation-driven dispatch, forecasting, and bullwhip-effect analysis for the logistics systems that keep cities and industry supplied.