Opinion-Driven Robot Navigation: Human-Robot Corridor Passing

We propose, analyze, and experimentally verify a new approach for robot social navigation in human-robot corridor passing that is driven by nonlinear opinion dynamics. The robot forms an opinion over time in response to its observations, and its opinion drives its motion control. The algorithm inherits a key feature of the opinion dynamics: deadlock, also known as the "freezing robot" problem, is guaranteed to be broken even if the robot has no bias or evidence for whether it is better off passing on the right or the left. The robot can also overcome a bias that is in conflict with the passage choice the human makes. The approach enables rapid and reliable opinion formation, which makes for rapid and reliable navigation. We verify our analytical results on deadlock breaking and rapid and reliable passage with human-robot experiments. We further verify through experiments that a single design parameter can tune the trade-off between efficiency and reliability in human-robot corridor passing. The new approach has the additional advantage that it does not rely on a predictive model of human behavior.