Visual Homing Navigation With Two Landmarks: The Balanced Proportional Triangulation Method

Abstract

Current approaches to the visual homing for mobile robot navigation are generally inspired in insects' behavior and based on the observed angular information of fixed points of the environment (landmarks). They suppose the capacity for absolute orientation of the observations (compasses) and/or require the detection and identification of at least three landmarks; furthermore, they tend to generate homing paths with circumvolutions which reduce the efficiency of navigation. This document proposes a new approach to the visual homing navigation based on the observed angular information of only two landmarks, without any previous knowledge of their locations and real sizes or any compasses or range sensors. The route calculation towards the home point is accomplished systematically by means of triangulation/trilateration, considering the apparent size observed for each landmark and obtaining rectilinear navigation paths. The proposed visual homing navigation method has been simulated extensively in an obstacle-free and uncertainty-exempt environment, verifying its strength and efficiency in the navigation of short and long range with different scene configurations; it can be easily integrated with other navigation systems (e.g.: obstacle avoidance), and even be used to calculate the robot's relative location