Visual control for trajectory tracking of Quadrotors and real-time analysis on an emulated environment

摘要

Quadrotor are non-inertial systems with inherent nonlinear, very fast and very unstable dynamics where measurements of their state variables in an absolute frame of reference are very complex in GPS denied environments, such as indoors. In order to design control strategies for Quadrotors, we solve problems of imprecisely known models and noisily, or unavailable, position and speed measurements, and requirements of a safety test-bench for the first stages of design. In this paper, we face the problem of control of a Quadrotor using visual information and of a safe virtual Real-Time test of the controlled system. The contribution is twofold; first we present a model-free control law for tracking non-inertial robots, using as a case of study a Quadrotor, and we propose a Real-Time analysis of the dynamical behavior of the closed-loop system based on experimental data obtained using an emulation-based implementation architecture where the environment is a numerical emulation of the real one. The control of attitude and position are designed based on a Second Order Sliding Mode and PD like approaches, respectively, and the feedback state is obtained visually using an easy to program and portable Particle Filter for positions and a Second Order Differentiator for time derivatives. We describe the numerical simulation test bench and include some results obtained in Real- Time on an emulated environment.