Biologically Inspired Locomotion Generation and Control of Humanoid Robots Employing a Network of Neural Oscillator

In fact, many researchers have dealt with these fields individually as inherent research approaches. However although these strategies have performed independently in robotics fields, these approaches are key matters connected with our objective. Although recently many humanoid robots have made their successful debut, they are still very difficult to control. Thus they may not behave as expected in certain environments. Behavior (or motion) can be learned through imitation which lets the user specify entire behaviors by demonstration without programming. By this, imitation has been considered one of the effective and powerful forms of learning, but its application to humanoid robot programming has not been formalized yet. In this research, we will develop a biologically-inspired framework for human-like autonomous motion generation and control of humanoid robots. Specifically, the new motion control architecture will be designed to make the robots fault-tolerant, easy-to-learn, and behave autonomously. Control approach based on neural oscillator has two essential merits distinguished as compared with traditional control approach in view of dynamical behavior such as robustness over parameters under same environment condition and adaptation over variety of environment. But it is hard to appropriately tune parameters of the neural oscillator for entrainment under a certain condition. And also, to generate the desired motion for objective task is very difficult because of nonlinearity of the neural oscillator. These are major drawbacks that control algorithm based on neural oscillators cannot be employed to various robotic files. In motion adaptor of new framework developed in this research for motion learning by imitation between dissimilar bodies and in precision control based on biologically inspired system for desire task, reciprocal action for compensating their demerits is employed and proposed in this research. Imitator is not able to obtain information related with internal force, interaction and effect of demonstrator’s actions, dynamic properties, etc through vision capture system for imitation learning. Thus, in biologically inspired system based on neural oscillators, the output can be used as reference data for the desired torque or angle output to perform a given task, and it is not necessary to compensate accurate dynamic effect of demonstrator in imitator implemented with biologically inspired system. Therefore, we introduce new policy to successfully achieve our goal, “Biologically inspired locomotion generation and control of humanoid robots employing a network of neural oscillator”, as well as to cope with problems as previousmentioned.