A robust ray-tracing (RT) based single-site 3D localization method is proposed for urban non-line-of-sight (NLOS) scenarios. RT is utilized to model channel propagation in complex environments, with NLOS paths transformed into LOS paths through a generalized source (GS) approach. The weights of GS are determined using a multipath residual weighting scheme based on an assignment optimization algorithm. To address irregular NLOS noise, a robust iteratively reweighted least squares with initial weight (W-IRLS) method is developed. Simulation results demonstrate that the method achieves near-Cramer-Rao lower bound (CRLB) accuracy under severe NLOS conditions and maintains robustness against geometric model errors, providing critical support for integrated sensing and communications (ISAC) applications in complex urban environments.

This letter presents an innovative single-site angle of arrival (AOA) and time difference of arrival (TDOA) hybrid localization algorithm for non-line-of-sight (NLOS) scenarios. An advanced ray-tracing (RT) method is employed to evaluate channel propagation in complex environments, transforming NLOS paths between the base station (BS) and mobile station (MS) into equivalent LOS paths by leveraging multipath propagation. This process identifies generalized sources (GSs) in the BS's environment, including LOS, reflective, diffractive, and transmissive sources. A novel weighting algorithm based on an assignment optimization approach is proposed to determine the locations of TDOA reference stations and assign weights to the TDOA/AOA hybrid equations. Furthermore, considering the significant NLOS noise in the hybrid TDOA/AOA equations, the proposed method integrates a weighting-enhanced iterative reweighted least squares (W-IRLS) algorithm to enhance the robustness of the localization process. Simulation results demonstrate that the proposed algorithm achieves Cramer-Rao lower bound (CRLB)-level accuracy under severe NLOS multipath conditions.

Vehicle-to-vehicle (V2V) communications require highly accurate and real-time channel modeling techniques. The dynamic nature of V2V scenarios demands continuous scene updates while ray-tracing (RT) algorithms typically require substantial computational time to achieve accurate predictions. This letter proposes an OptiX-based GPU-parallel dynamic shooting and bouncing ray tracing (DSRT) method that combines the advantages of shooting and bouncing ray (SBR) and RT approaches. The method employs an enhanced adaptive ray tube splitting mechanism to improve accuracy. A level of detail (LOD)-based dynamic scene update method further optimizes performance, while OptiX and CUDA acceleration enable efficient RT. Channel measurements in urban environments demonstrate the high accuracy of our algorithm. Performance evaluation results show that DSRT achieves a parallelization degree exceeding 98.8% and delivers exceptional speedup ratios.