In this paper, we present a novel method that detects appearing targets and improves their detection probability in a known static environment. Ray tracing-based channel model is used to calculate the radio signal paths using a 3D model of the environment. This preliminary information and the measured radar image are delivered to the convolutional neural network (CNN) based autoencoder (AE). The output image provides an improved representation of appearing targets, as the redundancy coming from known static environment is modeled and cancelled from the radar image. The method does not need the traditional constant false alarm rate (CFAR) threshold setting, which is an advantage, especially in heterogenous environments. The functionality of the method is tested with radar measurements in a laboratory corridor. Our results show that environmental clutter decreases significantly in the radar image and new targets are more clearly distinguished.

A ray tracing simulator for urban and indoor environments is introduced. The simulator uses NVIDIA graphics processing unit (GPU) accelerated CUDA parallel computing platform and programming mode and the OptiX Ray Tracing Engine. As a use case, channel characteristics for Global Navigation Satellite System (GNSS) satellites are simulated and compared with measurements in an urban area. The speedup achieved by parallel processing allows computation of multiple relevant reflections characteristic of a satellite channel.

This paper proposes a semi-analytical performance prediction method for an iterative receiver that processes signals arriving through a frequency selective multiple-input multiple-output radio channel. Earlier results provide a distribution of the signal-to-interference-and-noise ratio at the output of a frequency-domain soft-cancellation minimum mean square error equalizer, which is subsequently used to approximate the distribution of the log-likelihood ratios (LLRs) of the demapped bits. Extrinsic information transfer (EXIT) charts establish a bijective relationship between the variance of the LLRs and mutual information. We use an approximate distribution of the equalizer/demapper output's LLR variance to estimate the probability of an intersection between the decoder and equalizer graphs. The probability provides a good estimate of the frame error rate.

This paper proposes a semianalytical performance prediction method for an iterative receiver that processes signals arriving through a frequency-selective multiple-input multiple-output radio channel. Earlier results provide a distribution of the signal-to-interference-plus-noise ratio at the output of a frequency domain soft interference cancellation minimum mean square error equalizer, which is used to approximate the distribution of the log-likelihood ratios (LLRs) of the demapped bits. Mutual information transfer charts establish a bijective relationship between the variance of the LLRs and the mutual information. We assume a 3rd Generation Partnership Project compliant turbo code, whose transfer charts are simulated in advance. By approximating the LLR variance distribution of the demapped equalizer output, we can evaluate the probability of an intersection between an equalizer chart associated with a random channel realization and a fixed decoder chart. This probability provides the frame error rate. Our proposed prediction method allows horizontally and vertically encoded spatial multiplexing schemes. The vertically encoded scheme is complemented by the inclusion of the Chase combining hybrid automatic repeat request protocol. The proposed method allows reasonable receive antenna correlation.