MACHINE LEARNING-BASED INTERFERENCE MITIGATION AND GHOST TARGET REDUCTION FOR ONE-BIT QUANTIZED SIMO FMCW AUTOMOTIVE RADARS

Date

2024-9

Author

Baytok, Musa Burak

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

198
views

88
downloads

Cite This

The use of automotive radars equipped with one-bit analog-to-digital converters offers a promising alternative to their high-precision counterparts, due to their cost-effectiveness and low power consumption. However, one-bit sampling can lead to the appearance of ghost targets in range-Doppler maps, potentially causing false detections. Furthermore, in scenarios involving radar-to-radar interference, the target-like appearance and high power of coherent interference exacerbates the ghost target problem, while non-coherent interference reduces target detectability by raising the noise floor. This thesis addresses the issue of ghost targets caused by coherent interference and the noise floor increase due to non-coherent interference in one-bit quantized frequency-modulated continuous-wave radars. It explores the use of machine learning methods to mitigate interference and reduce ghost targets in range-Doppler maps. Performance analysis of a neural network based solution is conducted based on the detection performance, the number of ghost targets in the network’s output, and the SINR metric. The results demonstrate that the proposed neural networks effectively eliminate the ghost targets caused by coherent interference, thus preventing false detections, and mitigate interference caused by non-coherent interference, thereby increasing SINR.

Subject Keywords

Automotive Radar, One-Bit Quantization, Ghost Target, Interference, Machine Learning

URI

https://hdl.handle.net/11511/111329

Collections

Graduate School of Natural and Applied Sciences, Thesis