Weather-Aware and Perception-Driven Digital Twin for Urban Robot Navigation

Urban autonomous robots require navigation systems that can adapt to real-time environmental conditions and ensure safe path planning across complex road networks. Existing routing algorithms perform well in static settings using traditional shortest-path methods, but they often fail to account for dynamic weather conditions and integrate real-time perception, limiting their reliability during adverse scenarios when safetycritical decisions are essential. Therefore, this work proposes a digital twin framework that combines YOLOv11-based object detection trained on the KITTI dataset, live meteorological data from the Open-Meteo API, and adaptive routing applied to road networks generated via OSMnx. The system models roads as weighted directed graphs, where Dijkstra’s algorithm computes optimal paths using scenario-specific multiplicative penalties ranging from 8% for light rain to 50% for severe precipitation. An interactive IoT dashboard provides real-time monitoring of conditions and routing outcomes. Experimental validation on the urban road network of Cambridge, England (3,294 nodes, 7,227 edges) shows adaptive travel times from 12.2 minutes in clear weather to 29.1 minutes in heavy rain. Vehicle classes were detected with a strong mAP50 of 0.854 and precision of 0.917, while route lengths were adjusted from 4.33 km in clear weather to 12.77 km under severe conditions. This integration of perception, meteorology, and routing enhances the safety and robustness of autonomous urban navigation.