The author proposes a soft-constrained Schrödinger bridge (SSB) as a generalized stochastic control problem, allowing flexibility in terminal distributions while penalizing divergence. The main contribution is the theoretical derivation of the solution to SSB, showing its generalization of the original Schrödinger bridge.
The author proposes an efficient RADI-type method, ISC, to solve large-scale stochastic continuous-time algebraic Riccati equations by incorporating shifts and compressions.