High-resolution hybrid TDM-CDM MIMO automotive radar

Abstract

This paper proposes a deep learning (DL)-based high-resolution hybrid time-division multiplexing (TDM) and code-division multiplexing (CDM) multiple input multiple output (MIMO) automotive radar to enhance the discrimination capabilities of the radar in a cluttered environment. The hybrid TDM-CDM approach is implemented by partitioning the transmit and receive arrays into subarrays, applying CDM across the subarrays, while TDM is used within each subarray. On the other hand, the DL-based scheme utilizes the SqueezeNet deep convolutional neural network (DCNN), which treats the angle, range, and Doppler estimations of the extracted targets as a multi-label classification problem. Compared to CDM-MIMO radars, this approach requires fewer spreading codes, alleviating the challenge of spreading and despreading over each element. Compared to TDM-MIMO radars, it requires fewer time slots, increasing the refresh rate. Our approach outperforms existing DL-based TDM-MIMO radar systems and performs similarly to DL-based CDM-MIMO radar systems but with reduced complexity. Simulation results show that an angular resolution of 0.25° was achieved using 12-element transmit and receive arrays, each partitioned into three subarrays.