Multipath Ghost Target Detection Using Space Processing for Automotive MIMO Radar

The paper presents an algorithm for separating ghost and real objects in automotive MIMO radar systems. Ghost objects are typically false targets generated by multiple reflections of radar signals from objects on the road. For real objects, the angles of arrival estimated using transmit, receive, and virtual MIMO antennas are generally consistent. However, in multipath scenarios, where signals reflect from multiple surfaces before returning to the radar, these angles of arrival exhibit distinct differences. The proposed algorithm leverages the angular discrepancy to distinguish between real objects and ghost targets. To validate its effectiveness, the algorithm was tested using an open dataset containing automotive radar measurements. The results demonstrate that the proposed method can reliably serve as a feature for separating real targets from multipath-induced ghost objects, making it a promising solution for improving the accuracy and reliability of automotive radar systems.