Integrating the Simplex Architecture to Enhance Safety in Deep Learning Autonomous Systems

The rapid advancement of artificial intelligence (AI) has led to the development of deep neural networks (DNNs) that surpass human performance in various specialized tasks, including image classification [5], object detection [10], and control [8]. Numerous industries took notice of these findings and began implementing deep learning techniques in cyber-physical systems (CPSs), including automobiles, robots, and drones, to enhance their autonomy and carry out complex tasks. Those solutions are not free of drawbacks. The lack of trustworthiness of deep learning models in comparison to the stringent industrial-grade standards used in critical domains is one of the primary barriers to their deployment in CPSs. Generally speaking, trustworthiness refers to a collection of desirable attributes that a learning-based CPS ought to possess, such as safety, security, and temporal predictability.