In this paper, we propose a novel computational method, termed Monte Carlo Integration with Common Volume Sampling (MCI-CVS), which enables direct photon sampling within the common scattering volume. This approach significantly accelerates the convergence of optical wireless scattering channel simulations. Although each sample in MCI-CVS incurs approximately 3.74 times the computational cost of conventional methods, numerical results for ultraviolet scattering channels demonstrate that it reduces the required number of samples by a factor of 1,000 compared with Monte Carlo Integration with Importance Sampling (MCI-IS). As a result, MCI-CVS achieves an overall efficiency gain of approximately 267.3 times faster than MCI-IS and about 1,206 times compared to the baseline Monte-Carlo Simulation (MCS) method. Furthermore, the proposed method effectively addresses inefficient photon path sampling in infrared scattering channels, where low scattering and absorption coefficients degrade sampling efficiency and thus result in slow convergence. By incorporating photon path reuse, MCI-CVS also enables fast generation of two-dimensional scattering fields on a standard personal computer.

Considering the scattering characteristics of wireless ultraviolet (UV) channels, the double-scatter assumption is revealed to well characterize the scattering-enhanced channel over a range of less than 100 m. Based on the cuboid scatterers with representative shapes, we present a ray-tracing algorithm under scattering enhancement conditions and give the probability density function of the free path in a non-uniform atmosphere. Then, we propose a semi-analytical (SA) link gain model with scattering enhancement, via generating a set of coordinates for a series of conical integral paths using a rotation matrix. On this basis, a one-dimensional integral expression for the single-scatter of the proposed SA link gain model is obtained, and the expression for the double-scatter is further derived. The link gain evaluation results show that the presence of scattering region can improve the received power of the non-line-of-sight UV system at a receiver. For all considered transceiver geometry, the results of the proposed SA model match well with the counterparts of Monte-Carlo (MC) based numerical calculation, with a gap of around 2 dB for both narrow and wide beam. In terms of computational efficiency, the execution time of the proposed SA model for the single-scatter and double-scatter is significantly reduced to 0.065% and 9.69% of those of the MC simulations, respectively. We further establish experiments using a fog-filled quartz cubic cavity in an outdoor environment to verify the correctness of the proposed model. The link gain from the numerical evaluation assuming double-scatter can improve the agreement with the experimental results by 1 dB compared with that of the single-scatter assumption.

In this work, we develop a theoretical transmission model for optical wireless communication systems with receiver nonlinearity. To provide a framework for analyzing transmission rates, we derive lower and upper bounds on channel capacity under average and peak optical power constraints in both single-input single-output and multiple-input multiple-output systems using entropy power inequalities and dual expression. Numerical results verify the tightness of our derived capacity bounds and highlight the significant relationship between channel capacity and receiver nonlinearity.

Multi-radar product mosaic is a kind of common product in many applications and is a pre-requisites many weather-radar-related machine learning research. Two highly paralleled mosaic image generation frameworks are proposed. The first one is based on the MapReduce paradigm. In the map stage, each gate is mapped from radar coordinates to a radar-centric grid space; in the group stage, gates contribution to grid cells are collected; in the reduce stage, the final values are derived. The same logic applies again to combine radar-centric grids into the final domain. Another framework is replacing original reduce stage with RayTracing technology which allows generating products at multiple resolutions in one batch.