Dual-Process System: An Architectural Pattern for Assurable Autonomous Robots Inspired by Dual-Process Theory

This paper proposes the ‘dual-process system,’ an architectural pattern for assurable autonomous robots, inspired by the dual-process theory of the human mind. It seamlessly integrates end-to-end neural architectures with symbolic AI methods, mitigating critical challenges in assurability, such as interpretability, robustness, and the handling of rare inputs, thereby enhancing both performance and safety.