Altyapısız göreli iç alan uzamsal kestirim çerçevesi için bir yöntem

This thesis proposes the Infrastructure-free Relative Indoor Spatial Estimation framework to set up a dynamic relative coordinate system for mobile agents in Global Navigation Satellite Systems denied environments. The proposed system runs without relying on pre-installed infrastructure. To address geometric ambiguities inherent in range-only localization, the research introduces a hybrid estimation strategy fusing Ultra-Wideband Time-of-Flight ranging with Pedestrian Dead Reckoning velocity estimates. The method employs a Multi-Hypothesis Tracking algorithm to resolve initial angular uncertainty and transitions to an Extended Kalman Filter for tracking upon convergence. Furthermore, a centralized Non-Linear Least Squares optimization using the Dogleg algorithm and Huber Loss function ensures global geometric consistency across the network by mitigating drift and outliers. Simulations resulted in a positioning Root Mean Square Error of 0.448 meters in movement scenarios having paths with continuous curvature and demonstrated scalability in networks of up to 11 agents.