Photon Splatting: Real-Time Propagation and 4D Channel Modeling for Wireless Digital Twins

Modeling radio wave propagation in complex radio environments is essential for next-generation (NextG) wireless applications such as immersive computing and wireless digital twins, which require reliable, real-time predictions to optimize connectivity and ensure seamless performance. However, traditional ray-tracing algorithms are computationally demanding for real-time use. This paper proposes Photon Splatting, a novel neural propagator inspired by photon mapping and Gaussian splatting. The model delivers near real-time predictions of channel impulse responses. Numerical experiments demonstrate both scalability and accuracy in complex wireless environments.