ABSTRACT 

          For the safety-critical autonomous systems, analysis of the safe deep learning models is of great interest. This proposed work focuses on the development of the temporal logic based optimal control policies using reinforcement learning. Specifically, we use temporal logic to describe the learning of complex task specification. Combining with control Lyapunov functions to explore safety states and development process. The goal is to develop a safe reinforcement learning policy for the control in unknown environment. This research is supported by AFRL ML-RCP.