eXplainable AI Twins for Resilient Cities
Digital twins (digital replicas of cities) are a key innovation in shaping the future of mobility – in the face of growing challenges such as major disruptions from climate change, pollution and equity. Such twins, coupled with AI, are crucial in providing support to mobility system decision-makers. The ‘black box’ nature of competitive AI approaches, however, hinders adoption since AI recommendations are not transparent.
The XAIT Lab develops novel approaches that focus on the explainable aspect of AI for mobility. We incorporate explainability into three core modules within mobility decision support systems: state estimation, prediction and optimization. This means integrating multimodal data, contextual information and novel AI techniques into a single holistic digital twin platform. The result is augmented digital twinning methodologies with explainable AI techniques, creating a new class of AI twins for cities. Such approaches provide tangible and concrete decision support that helps resilient cities to tackle urban problems such as evacuation, energy, long-term planning and infrastructure operations.
The XAIT Lab is part of the TU Delft AI Labs programme.