Controle cooperativo ótimo dependente de grafo e sistema anticolisão voltados para formação e liderança de drones do tipo quadrimotor

Abstract

This thesis is based on the theory of multi-agent systems through the use of coopera- tive control, aiming at the centralization, formation and leadership of multiple dynamic

systems, specifically, four-engine drones. It presents the multi-agent concepts, given the perspective of guide systems, when a virtual agent serves as a reference for the real agent, and direct control, when protocols are applied directly to dynamic systems. Different ways of representing these systems are presented, from the simplest model

(based on Newton’s equations) to the most generalized ones, such as nonlinear sys- tems. In view of these models, the four-engine engine and its dynamic equations are

addressed, including its internal controllers and how the multi-agent system (protocol) operates in this type of application, enabling its use in agriculture, mobile sensing and security. Having defined the protocols, laws that determine the global behavior of the agents, an approach of cooperative optimal control is presented for the determination of a linear quadratic regulator based on the global behavior of the agents through the

existence of a virtual leader, including the concepts of formation and group centrali- zation. For this, a cost function was developed based on the variable of tracking and

control action, allowing the optimal parameters to be determined as a function of the

representative structure of the information flow between the group of agents, the La- placian matrix, considering the coupling in relation to the leader. These formulations

make it possible to calculate the optimal cooperative control for both homogeneous and heterogeneous systems, when the dynamics of the agents that make up the group differ from each other. As a main contribution, an anti-collision system is proposed based on the effect of divergence and convergence through the effects of centralization and decentralization of the flock, which resulted in an anti-collision parameter controlled according to predefined criticality regions. At first, this system is applied to protocols as a guide system, that is, when they serve as a trajectory reference for the four-engine. Finally, the 3D platform has been developed, which aims to give a closer perspective of what happens to the agents and generate reports of the results, both simulated and experimental. In the experiments, mini-drones are used that are controlled via radio and their positions measured based on infrared sensors. The control laws used are constituted by the protocols and anti-collision laws developed that act by controlling the trajectory of these minidrones, obtaining results similar to those simulated.