Visual servo control for a human-following robot

[摘要] ENGLISH ABSTRACT: This thesis presents work completed on the design of control and vision componentsfor use in a monocular vision-based human-following robot. The useof vision in a controller feedback loop is referred to as vision-based or visualservo control. Typically, visual servo techniques can be categorised into imagebasedvisual servoing and position-based visual servoing. This thesis discusseseach of these approaches, and argues that a position-based visual servo controlapproach is more suited to human following.A position-based visual servo strategy consists of three distinct phases:target recognition, target pose estimation and controller calculations. Thethesis discusses approaches to each of these phases in detail, and presents acomplete, functioning system combining these approaches for the purposes ofhuman following.Traditional approaches to human following typically involve a controllerthat causes platforms to navigate directly towards targets, but this work arguesthat better following performance can be obtained through the use of acontroller that incorporates target orientation information. Although a purelydirection-based controller, aiming to minimise both orientation and translationerrors, suffers from various limitations, this thesis shows that a hybrid,gain-scheduling combination of two traditional controllers offers better targetfollowingperformance than its components.In the case of human following the inclusion of target orientation informationrequires that a definition and means of estimating a human's orientationbe available. This work presents a human orientation measure and experimentalresults to show that it is suitable for the purposes of wheeled platformcontrol. Results of human following using the proposed hybrid, gain-schedulingcontroller incorporating this measure are presented to confirm this.

[发布机构] Stellenbosch University