Hybrid Co-Learning for Proximate Human-Robot Teaming

Proximate human-robot teaming, where robots and humans share a common physical space to interact, imposes new challenges to capacitate effective co-learning between humans and robots. This paper investigates a hybrid co-learning problem arising from collaborative assembly scenarios, where the robot's action is operated in the continuous space and the human's intention transits in the discrete space. The human and the robot are distinguished as heterogeneous online learning systems, and their system dynamics, i.e., their learning strategies, are modeled with different adaptation rules according to their distinct properties. The evolution of the hybrid co-learning system has been studied by performing simulations under different conditions, where some unique interaction patterns, such as chatter, have been discovered and analyzed.