Autonomous action generation through easiness-based GBP solution

This article is concerned with the autonomous action generation algorithm by solving the Generalized Bernstein Problem (GBP) with the easiness criterion. The GBP is a class of optimization problem with redundancy resolution involved. Future autonomous robots should determine so many control parameters to accomplish various large-scale tasks, where there is a large arbitrariness in the control if only the task objective is given. The “easiness” is a universal criterion introduced instead of additional criterion task by task, and can be utilized not only for the redundancy resolution but also for action generation without any a priori control knowledge. The easiness-based action generation was applied to a simple muscle-based joint control problem, and the effectivity of the algorithm is examined and its evolvability is discussed.