Nonlinear Model Predictive Control of Robots, Cranes and Vehicles

The main goal of the paper is to show that nonlinear model based predictive control is an effective alternative method for controlling uncertain nonlinear underactuated systems satisfying real time expectations where uncertainty is caused by friction and disturbance. The paper presents control algorithms for a two degree of freedom robot and for two underactuated systems, the small size model of a planar crane and a weeled mobile robot, which are based on nonlinear model predictive control. The continuous time dynamic model has been discretized and the finite dimensional optimization problem is solved by conjugate gradient technique in every horizon. Extended Kalman filter is used for state and disturbance estimation. For the crane the initial approximation of the control sequence within the actual horizon is determined by using the flatness properties. For development purposes a multiprocessor system has been elaborated where the control algorithm is running under QNX real-time operating system. The paper also presents experimental results that demonstrate the applicability of the proposed algorithms under real time conditions.