Control PID basado en odometría visual monocular a un vehículo de superficie marino

Abstract

The aim of this work is to study the behavior of a marine vehicle applying vision-based control. We use a DMVO algorithm (Direct Monocular Visual Odometry) that is based at the feature extraction of two image frames taken consecutively at two continuous tame frames in order to estimate a marine vehicle’s position; vehicle’s dynamics are modeled without external perturbations and therefore apply a vision based Proportional Integral Derivative (PID) control for the study of its behavior through simulations using computational tools to simulate a 3D scenario and get the matrix that contains the image from the inertial frame of reference view from the vehicle. First we need a static scenario that contains reference points such that we are able to apply the proposed vision based PID control; the virtual scenario was designed at the MATLAB’s virtual world editor that let us add sensors such as cameras, and set their parameters for make the study and simulations such as path following.