Performance Analysis of a Continuous Vision-based Control System for the Navigation of a Mobile Robot

In the last year, we have proposed a new continuous visualnavigation approach defining the navigation as a global visualcontrol task. It implies continuity problems produced by thevisibility changes of the image features during the navigation. Toavoid the discontinuities in the control law, a smooth taskfunction based on weighted features was proposed and a continuouscontrol law was obtained and tested. In this paper, a performanceanalysis of this visual navigation approach is presented. First ofall, the influence of the parameters selection in the weightfunctions is analyzed. The visual navigation approach presented inthis paper is used to control the robot to repeat the walk learnedbefore. The departure position of the robot must be near theinitial position memorized before. The word "near" is in someway quantified in this paper.