Control allocation applied to robots subject to input constraints

Abstract

[EN] In control systems with multiple inputs and multiple outputs (MIMO), the control signal of a given degree of freedom can be distributed among several actuators by means of a strategy called control allocation. Nonetheless, the control allocation problem may not be trivially solved, since in the case that the system actuators are subject to constraints (e.g. saturation), the desired behavior may not be completely satisfied. In this scenario, the control allocation consists of an optimization problem which searches the best possible control distribution while it satisfies these constraints. The control allocation techniques may have as a primary objective to maintain the control direction or to minimize the error, or may even satisfy a secondary objective. To encompass all of these situations, the techniques Direct Control Allocation, Linear Programming with Simplex, Weighted Least Squares with Active Set and Primal-dual Interior Point are employed in an experiment with a differential drive mobile robot and in simulations of an Unmanned Aerial Vehicle (UAV), such as a quadrotor, and also of a Remotely Operated Underwater Vehicle (ROV). A second approach of the control allocation aims to distribute the load among the agents in a cooperative task, which in this dissertation is represented by means of the simulation of a system composed by two manipulators and an object. All these systems are subject to trajectories with different requirements, their dynamics are analyzed and their results compared.