Trajectory-Based Physical Target Discrimination Method for 4D mmWave Vehicular Radar

In current Intelligent Transportation Systems, 4-dimensional (4D) millimeter-wave radar is gradually conquering more important roles. However, the rich multipath phenomena of radar waves can cause virtual images to be mistaken for actual targets. Most existing solutions aim to identify distinguishable features within a single signal frame to differentiate virtual targets from actual ones, but they fail to fully leverage temporal evolution information. Moreover, they are often limited to simple scenarios with low-order bounces and cannot handle extreme high-order bounce cases. To address this gap, in this paper we establish a research methodology based on topology and propagation graph theory, and propose a time-varying 4D trajectory-based physical target discrimination (TPTD) method. This method can handle multipath virtual images formed by arbitrary-order bounces. Its effectiveness and superiority over advanced methods have been qualitatively and quantitatively validated through traceable simulated and measured data. We also discuss in detail the scalability of the algorithm and its potential bottlenecks. The proposed method can be applied to fields such as precise radar sensing and channel modeling, laying the foundation for integrated sensing and communication and autonomous driving.