High gain observer-based visual servoing for depth estimation in robotics

Abstract

In the engineering field, we are always facing with the system or process with nonlinear type which is not trivial to control its behaviour, and to response this problem we just equipped different type of sensors to be able to analyse the output measurement. Unfortunately, we can not use lot of sensor in a system due to its cost and also the limited in terms of technology examples: finding the temperature in a heater exchange or to know the impact of the wind on a car, etc. That is why we use to create the virtual sensor called observer to replace or make a redundancy with the physical sensor, by applying this type of sensor we can improve the performance of the system in terms of cost and especially in the case when we can not use the physical sensor. This paper proposes a mono-vision depth estimation method for mobile robots, based on high gain observers. The objective of this paper is to estimate the distance between the robot and target object by using only a camera fixed on-board. The proposed method switches between multiple observers in order to gauge the distance between the camera and the target object. Simulations and experimental results illustrate the practical applicability and benefits of this method for visual servoing.