In this chapter, we discuss optimization as a way to generate whole-body motions for humanoid robots. Optimization helps to solve many difficulties related to humanoid motion generation: redundancy, feasibility, exploitation of physical capabilities, maintaining stability, as well as handling of underactuation and changing contacts with the environment. We present the formulation and numerical solution of optimal control problems for whole-body humanoid optimization with multiple phases discussing in detail different choices for objective functions and constraints to be considered. We summarize example applications of optimization for humanoid motion synthesis and motion imitation. Potential promising combinations of optimization with learning methods and movement primitives are discussed. In addition, we describe the inverse optimal control problem that helps to determine the optimality criterion underlying a recorded human motion, which then can serve as input for humanoid motion optimization. The formulation and numerical solution of these problems for locomotion examples are discussed, and example results of inverse optimal control for human locomotion based on whole-body models are shown. Further research directions in humanoid motion optimization are discussed. To give the full picture, we also mention some results for optimization in locomotion path generation and for template models which are not the focus of this chapter.
CITATION STYLE
Mombaur, K. (2018). Humanoid Motion Optimization. In Humanoid Robotics: A Reference (pp. 1805–1842). Springer Netherlands. https://doi.org/10.1007/978-94-007-6046-2_33
Mendeley helps you to discover research relevant for your work.