2019/05/28 by Benjamin Gravell, Yi Guo, Gravell, Benjamin +3
Computer Science · #Neural Networks Stability and Synchronization
paper · pdf · doi:10.48550/arxiv.1905.13548
We give algorithms for designing near-optimal sparse controllers using policy\ngradient with applications to control of systems corrupted by multiplicative\nnoise, which is increasingly important in emerging complex dynamical networks.\nVarious regularization schemes are examined and incorporated into the\noptimization by the use of gradient, subgradient, and proximal gradient\nmethods. Numerical experiments on a large networked system show that the\nalgorithms converge to performant sparse mean-square stabilizing controllers.\n