Integration of linguistic and numerical information for biped control

Bipedal locomotion is an important hallmark of human evolution. Despite of complex control systems, human locomotion is characterized by smooth, regular, and repeating movements. Therefore, there is the potential for applying human locomotion strategies and any knowledge available to the biped control. In order to make the most use of the information available, a linguistic-numerical integration-based biped control method is proposed in this paper. The numerical data from biped measuring instruments, and the linguistic rules obtained from intuitive walking knowledge and biomechanics study have been classified into four categories: direct rules, direct data, indirect rules, and indirect data. Based on inverse learning and data fusion theory, two simple and intuitive integration schemes are proposed to integrate linguistic and numerical information with various forms, such as direct and indirect. One is neurofuzzy-based integration, and another is fuzzy rules extraction-based integration. The simulation results show that the biped gait and joint control performance can be significantly improved by the prescribed synergy method-based neurofuzzy gait synthesis and fuzzy rules extraction-based joint control strategies using linguistic and numerical integrated information.