A combined learning and optimization framework to transfer human whole-body loco-manipulation skills to mobile manipulators

Humans' ability to smoothly switch between locomotion and manipulation is a remarkable feature of sensorimotor coordination. Learning and replicating such human-like strategies can lead to the development of more sophisticated robots capable of performing complex whole-body tasks in real-world environments. To this end, this paper proposes a combined learning and optimization framework for transferring human loco-manipulation soft-switching skills to mobile manipulators. The methodology starts with data collection of human demonstrations for locomotion-integrated manipulation tasks through a vision system. Next, the wrist and pelvis motions are mapped to the mobile manipulators' End-Effector (EE) and mobile base. A kernelized movement primitive algorithm learns the wrist and pelvis trajectories and generalizes them to new desired points according to task requirements. Then, the reference trajectories are sent to a hierarchical quadratic programming controller, where the EE and the mobile base reference trajectories are provided as the first and second priority tasks, respectively, generating the feasible and optimal joint level commands. Locomotion- integrated pick-and-place and door opening tasks have been chosen to validate the proposed approach. After a human demonstrates the two tasks, a mobile manipulator executes them with the same and new settings. The results showed that the proposed approach successfully transfers and generalizes the human loco-manipulation skills to mobile manipulators, even with different geometry.