Journal
AUTOMATION IN CONSTRUCTION
Volume 137, Issue -, Pages -Publisher
ELSEVIER
DOI: 10.1016/j.autcon.2022.104211
Keywords
Delta parallel robot; Pick and place; Dynamic model selection; Robust back-stepping control; Reinforcement learning; Extended external model; Insufficient excitation; Construction management
Ask authors/readers for more resources
This article introduces a novel smart online dynamic model selection method and a back-stepping sliding mode controller for robotic manipulators in the construction field, specifically a 3-DOF Delta Parallel Robot used in pick-and-place operations. By identifying reduced-order extended models based on external loads in an online manner and utilizing an off-policy reinforcement learning approach for smart dynamic model selection, a robust evolving controller is developed to handle pick-and-place tasks under various configurations of external loads, resulting in better tracking properties compared to a single external model.
Applications of robotic manipulators in construction fields is notorious; however, changes in system dynamics in the presence of heavy external loads and disturbances in pick-and-place operations are inevitable. To elude this, a novel smart online dynamic model selection is introduced and accompanied by a back-stepping sliding mode controller which is implemented on a 3-Degrees-Of-Freedom (DOF) Delta Parallel Robot. In order to fit the dominant behavior of the disturbances, reduced-order extended models, based on external loads, are identified in an online manner; thereafter, an off-policy reinforcement learning approach is exploited for smart dynamic model selection. Consequently, a robust evolving controller emerges able to perform pick-and-place tasks under any configuration of external loads, resulting in better tracking properties in comparison to fitting a single external model. Data-driven methods have potential for further improving the external loads' dominant behavior identification using the derived models' kernels opening up new avenues as future works.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available