A novel line of sight control system for a robot vision tracking system, using vision feedback and motion-disturbance feedforward compensation

Abstract

SUMMARYThis paper presents a novel line of sight control system for a robot vision tracking system, which uses a position feedforward controller to preposition a camera, and a vision feedback controller to compensate for the positioning error. Continuous target tracking is an important function for service robots, surveillance robots, and cooperating robot systems. However, it is difficult to track a specific target using only vision information, while a robot is in motion. This is especially true when a robot is moving fast or rotating fast. The proposed system controls the camera line of sight, using a feedforward controller based on estimated robot position and motion information. Specifically, the camera is rotated in the direction opposite to the motion of the robot. To implement the system, a disturbance compensator is developed to determine the current position of the robot, even when the robot wheels slip. The disturbance compensator is comprised of two extended Kalman filters (EKFs) and a slip detector. The inputs of the disturbance compensator are data from an accelerometer, a gyroscope, and two wheel-encoders. The vision feedback information, which is the targeting error, is used as the measurement update for the two EKFs. Using output of the disturbance compensator, an actuation module pans the camera to locate a target at the center of an image plane. This line of sight control methodology improves the recognition performance of the vision tracking system, by keeping a target image at the center of an image frame. The proposed system is implemented on a two-wheeled robot. Experiments are performed for various robot motion scenarios in dynamic situations to evaluate the tracking and recognition performance. Experimental results showed the proposed system achieves high tracking and recognition performances with a small targeting error.

A prey-predator model for efficient robot tracking

“…During this time period, the distance it covers is smaller than d− r, thus keeping distant from the aiming location at least by r. Eq. 7and (6)…”

“…In the tracking strategy for an intelligent wheelchair robot in indoor environments [5], an Extended Kalman filter is proposed to estimate robot pose based on Haar-like features, followed by obstacle avoidance navigation approach based on spline trajectory planning and optimization. An Extended Kalman filter based vision feedback controller to compensate for the positioning error is implemented for visual object tracking by autonomous vehicles in [6]. However, Kalman filters works only for linear systems with Gaussian noise as stated in [7].…”

Vision System for the Navigation of a Mobile Robot

“…Furthermore, when a system of multiple robots is considered, control and coordination are complex and demanding tasks. According to [16], vision systems for mobile agents, can be classified in six types, sensing and perception [31,23,9,17,7], mapping and self-localization [22,2,4], recognition and location [5,28,29,30,14], navigation and planning [13,20], tracking [21,19,1] and visual servoing [25,11,27].…”

Diseño de un robot móvil controlado por un agente reactivo en tiempo real

“…De acuerdo con [3] los sistemas de visión para agentes móviles se pueden clasificar en seis tipos, sensado y precepción [4][5][6][7][8]; mapeo y auto-localización [9,10,11]; reconocimiento y localización [12,13,14,15,16]; navegación y planeación [17,18]; seguimiento [19,20,21]; y control por visión (servoing) [22,23,24].…”