Diffusion Schrödinger Bridge-Based Channel Estimation for MIMO ISAC Systems

This work proposes a novel sensing channel estimation method for integrated sensing and communication (ISAC) systems based on diffusion models and the Schrödinger bridge. The channel estimation problem is formulated as an entropy-regularized optimal transport task. By leveraging stochastic differential equations induced by the diffusion Schrödinger bridge, the method models the optimal transport path between the least-squares (LS) estimation and the ground-truth channel. An ordinary differential equation (ODE)-based sampling scheme is adopted to enable efficient inference. Numerical results demonstrate that the proposed method achieves optimal estimation performance with only a single inference step across various signal-to-noise ratio (SNR) levels.