Adaptive motion pattern generation on balancing of humanoid robot movement

This paper discusses about adaptive trajectory control applied in motion pattern trajectory of humanoid robot movement. The aim of this research is to increase the stabilization of robot during walking and running. In this research, the control system produced the next step of the trajectory based on the current condition and analyzed the center of gravity point from the body of the robot. According to this, robot posed the foot step depend on the location of center of gravity point and stop the swing of its foot when the foot has reached the ground. In order to reduce the vibration effect arised by the swing of robot steps, this system is supported by vibration control. Robot is also supported by hand reaction learning system based on recurrent neural network. The trajectory pattern of robot movement has 2 trajectory equations: ankle trajectory formed by circle function in Cartesian coordinate space and pelvis trajectory formed by the third order polynomial equation. Both of them are influenced by inclination of the body of robot. We used the inverted pendulum approach combined with dynamic step trajectory. By using this system, robot can walk in the different surface and uneven surface. This system is applied on humanoid robot EROS (EEPIS Robot Soccer).