Active Perception With Initial-State Uncertainty: A Policy Gradient Method | IEEE Journals & Magazine | IEEE Xplore