Digital Twin Enabled Autonomous Vehicles for Obstacle Avoidance in Disaster Response

This is the authors' abstract. We don't add key points for this paper.

Unmanned Aerial Vehicles (UAVs) deployed in search and rescue missions must operate reliably across structurally diverse and unpredictable terrains. This paper proposes a novel digital twin based perception planning control framework that integrates the three dimensional Vector Field Histogram Plus (3D VFH) algorithm within the Field based Autonomous LiDAR Control for Obstacle Navigation (FALCON) architecture. The primary contribution is the establishment of a unified and consistent benchmarking baseline that isolates controller behaviour from environment specific tuning, enabling rigorous cross terrain performance evaluation without scenario dependent optimisation. The framework is validated in two geometrically distinct environments: a structured urban campus and an unstructured mountainous terrain. Performance is assessed using trajectory tracking accuracy, path length, mission time, and safety quantified through a 2 m Safety Envelope (SE). While all controllers exhibit comparable performance in the urban scenario, substantial differentiation emerges in rugged terrain. The manually tuned controller achieves the fastest completion with zero safety violations and a 6.21 m SE, whereas the PSO optimised controller also maintains zero violations with strong robustness characteristics. In contrast, GA and MA controllers experience violation rates of 14.3% and 28.6%, respectively, highlighting terrain induced sensitivity to gain optimisation strategies. The results demonstrate that controller robustness is strongly environment dependent, and that maintaining a fixed algorithmic baseline enables clear identification of perception control coupling effects. The proposed framework provides a scalable and reproducible testbed for evaluating environment aware UAV navigation strategies in safety critical applications.