UC San Diego

This paper presents a method for optimizing the throwing distance of a two-wheeled, self-balancing, remote controlled robot. The robot is maneuverable and comparable in size to a remote controlled toy car, but it moves around in an upright configuration using feedback control. In addition, it is capable of automatically picking up and throwing ping-pong balls. When throwing a ball, the body attached throwing arm is allowed to rotate quickly from a lay-down position to an upright position utilizing the principle of conservation of angular momentum. The equations of motions of a representative model of the dynamic system are derived with Lagrange equation and Rayleigh dissipation functions. The optimization consists of two parts. The shape optimization is based on the ball's exit condition from the arm, which wraps around an adjoint based motor input optimization. In simulation, this optimization scheme results in a significant increase in throwing distance while keeping the magnitude of the motor input within comparable range