Content area
Full Text
1.
Introduction
Interaction control is vitally important for robot to accomplish various tasks in industrial applications, such as spot welding, polishing and assembly (Tsai, 1999; Namvar and Aghili, 2005; Stolt et al. , 2015). When the end-effector contacts with the environment, the end-effector position is constrained by the environment, and a proper compliant behavior of the robot is necessary. In this case, pure position control, which entirely depends on the accuracy of task planning and control algorithms based on the accurate model of the manipulator and environment, is not possible (Deng et al. , 2005). To solve this problem, the three basic approaches to robot interaction control are, namely, impedance control, hybrid position/force control and parallel force/position control, as surveyed by Chiaverini et al. (1999). Many extended and improved approaches have also been proposed (Petkovic et al. , 2012; Jafari and Ryu, 2014; Mendes and Neto, 2015).
As a robot is expected to be more autonomous and flexible to accomplish complicated tasks in unknown environments, the need to integrate a number of different sensors into the robot system becomes increasingly important. Consequently, multisensor-based control scheme, which can compensate for changes in the environment and uncertainties in the dynamic models without explicit human intervention or reprogramming, has been proposed and developed (Xiao et al. , 2000; Lippiello et al. , 2008; Long et al. , 2014). In various sensors, the force and vision sensors are very widely used. Vision sensor can provide the non-contact global environmental information by real-time image measurements. The requirement for the exact specification of the target pose and exact positioning of the robot end-effector can be relaxed. On the other hand, force sensor can provide localized and accurate contact information to increase the dexterity and safety of interactive robotic control tasks (Cheah et al. , 2010).
Combing the advantages of vision and force sensing, many hybrid vision/force controller have been proposed. Three vision/force serving strategies were compared by Nelson et al. (1995): traded control, hybrid control and shared control. The benefits of combining vision and force sensing within the feedback loop of a robot manipulator were presented. Deng et al. (2005) combined three different visual control strategies: position-based, image-based and hybrid control with an impedance-based force controller and made comparisons among...