Model-Driven Development of Service Robot Applications Dealing With Uncertain Human Behavior

Abstract

In the future, robots will interact with humans in highly variable and unpredictable settings, such as healthcare and home assistance. Frameworks are, therefore, fundamental to analyze and develop interactive robotic applications that can deal by design with the uncertainty of human behavioral and physiological features. Our framework is built upon formal modeling, verification, and learning techniques providing sound mathematical guarantees of human wellbeing preservation throughout the interaction and timely mission completion despite the uncertainties at play. The framework's workflow makes it accessible to professional figures without expertise in formal methods, and the high degree of automation minimizes the manual effort required throughout the toolchain. The development of toolchain has been tested on use cases from the healthcare setting, in which human behavior is unconstrained, and subjects are often in critical mental or physical conditions.