查看文章

It is well known that robotic systems present physical limitations in the maximum torque that the motors can apply, and its variation rate. Also, safety regulations give strict limitations in the maximum link speed and acceleration. To include these limitations in the robotic control systems this paper proposes the implementation of a Multivariable Constrained Predictive Controller, that solves at each sampling time online the problem of calculating an optimal control sequence in the presence of constraints, using a model of the process. It is proposed to use for prediction a linearized model of the robot at each working point, which makes possible to meet the strict computational limitations of a real-time implementation, without degrading the closed-loop performance. The performance of this approach in the presence of model uncertainties, constraints in the maximum torque, its variation and in the generalized coordinates …