A basic concept of super rabbit

Abstract

We have performed a basic research on a jumping robot — Super Rabbit — like a rabbit or a kangaroo aiming a dynamical walking robot. Normally, linear quadratic [LQ] control method has been used to the linear or linearized system. We propose the control method with the artificial neural network [ANN] and the modern control theory to realize more effective and stable control for the nonlinear system such as a one-legged locomotion robot. We have got a posture controller [LQ + ANN] for this super rabbit to take a dynamic stepping or jumping through the simulation after solving the equation of this system dynamics, Riccati equation and the neural network dynamics. Furthermore, we have proposed to incorporate a rhythm pattern generator using neural network corresponding to the animal (or biological) nervous system into the robot system, i.e., "Rabbit robot". Many animals including human learn the rhythmical motion pattern and generate these motion pattern unconsciously such as walking or jumping. This rhythm/central pattern generator [CPG] exists in the spinal cord or higher level central nervous system. In this paper, we also propose to incorporate CPG constructed with the ANN into the robot system to realize a self-organized learning mechanism. This super rabbit consists of a rigid body and foot. Each part is connected with a rotatable joint corresponding to the ankle. The torque around the ankle, which is given by the motor system, makes the super rabbit jump forward or backward. The motion of this robot consists of four modes. Mode_1 is the neutral state. Mode_2 is the beginning of jumping or stepping, when the center of gravity moves back or forth corresponding to the body angle. The velocity vector to take a step is generated by the torque around the ankle. Mode_3 is the jumping state in the air. Mode_4 is the posture control state at landing. The posture in the air (Mode_3) is detected by the optical sensor incorporated to the foot.