An Adaptive Multi-objective Motion Distribution Framework For Wheeled Mobile Manipulators Via Null-space Exploration

Wheeled mobile manipulators (WMMs) are primarily composed of a manipulator and a mobile base, which lead to their agility, maneuverability, and mobility. Despite impressive progress in recent years, there remains some substantial work in improving WMMs' motion precision, owing to the current limitations in unpredictable wheel slippage, high system redundancy, and absence of external perception information. One intuition is to distribute more of a given set of end-effector motion requirements to the manipulator adaptively -- to reduce the motion error introduced by wheel locomotion. With the ultimate goal of improving a WMM's motion accuracy without external perception information, here we present a novel adaptive multi-objective motion distribution framework (AMoMDiF) for a redundant WMM, which is drawn inspiration from null-space exploration. The framework adopts a hierarchical structure to explore the WMM's null space iteratively to achieve three tasks, which are end-effector motion achievement (primary objective), adaptive motion distribution (secondary objective), and manipulability enhancement to avoid singularity (tertiary objective). The secondary objective strives to assign more of an end-effector motion to the manipulator when possible. The tertiary goal will enhance the manipulator's capability to stay away from singularities via the WMM's remaining redundancy after the primary and secondary tasks are completed. Experiment results on a physical platform show that, compared with the traditional motion planning method, the proposed AMoMDiF significantly improves the WMM's motion accuracy through the achievement of the three objectives.