Synthesizing Control Barrier Functions With Feasible Region Iteration for Safe Reinforcement Learning | IEEE Journals & Magazine | IEEE Xplore