MTL-based Joint DoA Estimation and AMC of Unknown Overlapped Signal

This paper proposes a multi-task learning (MTL)-based algorithm for joint direction of arrival (DoA) estimation and modulation classification to process overlapping unknown signals in multi-source communication environments. Conventional methods estimate each characteristic independently, resulting in redundant computations and low efficiency. To overcome these limitations, the proposed algorithm employs MoDANet, an MTL-based deep learning model that performs multiple independent tasks concurrently and integrates signal detection, separation, and feature estimation into a unified system. Simulation results show that the proposed algorithm achieves lower computational complexity and improved estimation performance compared with sequential approaches for individual tasks.