Actor–Critic Learning Control With Regularization and Feature Selection in Policy Gradient Estimation | IEEE Journals & Magazine | IEEE Xplore