This paper proposes a population based adaptive tuning for dynamic position control of robot manipulators. The dynamic behavior of a robot manipulator is highly nonlinear, and the positional control is conventionally achieved by inverse dynamics feedforward and PID feedback controllers. The proposed method tunes the PID controller parameters using cross-entropy optimization to minimize the error in tracking a repeated desired trajectory in real-time. The stability of the system is granted by switching the inappropriate settings to a stable default using a real-time cost evaluation function. The proposed tuning method is tested on a two-joint planar manipulator, and on a planar inverted pendulum. The test results indicated that the proposed method improves the settling time and reduces the position error over the repeated paths.
Download Full PDF Version (Non-Commercial Use)