Safety of Autonomous Systems Using Reinforcement Learning: A Comprehensive Survey

The safety of autonomous systems has become an important concern in recent years. A new mechanism for developing safe and reliable autonomous systems is a very challenging task. It is also very complex to accomplish trustworthy and optimized outcomes in uncertain and dynamic environments. Safety and security issues of machines are frequently raised with the advancement of technology, for example, medical robots with smart technology can perform complex surgeries, self-driving cars with smart road assistance models can significantly reduce the number of traffic crashes, and smart autopilots can perform delicate docking makeovers. Therefore, in this paper, we studied various autonomous system safety challenges, Machine Learning (ML) methodology aspects, theory, and applications. An overview of ML categories and sub-categories that are majorly used for the safety and security of Autonomous Systems are also explored. This paper also presents numeral applications of Reinforcement Learning (RL) which helps in the safety of autonomous systems. Reinforcement Learning lacks a good literature review in this area that has been also focused.