The optimal control methodology based on normalized Gaussian functions for vibration-driven locomotion robots
Binbin Diao,Xiaoxu Zhang,Jian Xu
DOI: https://doi.org/10.1007/s11044-024-09994-0
IF: 3.4
2024-06-08
Multibody System Dynamics
Abstract:The earthworm-like robot excels in navigating confined spaces for tasks like pipeline inspection and post-disaster rescue, owing to its wheel- and leg-free design. As its simplified dynamical model, the vibration-driven robot serves as the foundation for exploring locomotion mechanisms. Building upon this model, this paper centers on enhancing the mobility of the robot by maximizing velocity in scenarios where the actuation is unknown, which is denoted as the optimal control problem. Specifically, the normalized Gaussian functions, which enable a smooth approximation to arbitrary functions, are adopted to characterize the unknown actuation. As a result, the optimal control problem is transformed into an optimization challenge involving normalized Gaussian functions, which is addressed using the MOEA/D algorithm. Firstly, the numerical results of the one-module robot under isotropic friction show a close alignment with the theoretical results, which underscores the competitiveness of this new methodology. Furthermore, investigations into both one-module and five-module robots with anisotropic friction have confirmed a consistent three-phase control mode under the same friction ratios. Notably, this mode regresses to the two-phase control mode when the friction ratio equals zero. In addition, the precision of the actuation values is higher based on normalized Gaussian functions, which contributes significantly to advancing our understanding of the fundamentals of optimal control problems for vibration-driven locomotion robots, especially for multi-module robots.
mechanics