Giedrius Rakauskas

Movement Replicating Robotic Arm Control Interface

Lancaster University, MSc by Research, 2013

Supervisor: C.J. Taylor


This thesis presents a robotic manipulator control interface based on real–time marker–less operator hand position and orientation tracking with active collision detection. The proposed system presents users with an intuitive robotic arm control approach that does not constrain hand movements. In contrast to some other approaches, for example, the system is not limited to a set of pre–defined control input pairs in which hand movements are matched to fixed robotic commands. In this manner, the new interface has the potential to reduce the overall staff training time and to minimize human error when controlling robotic systems, which is especially important when working in environments that are potentially dangerous to humans or manipulators. Although similar robotic manipulator control systems are starting to appear in the literature, they lack direct and natural end–effector orientation control and/or do not operate in combination with active collision detection.

The experimental results are all based on two off–the–shelf products, namely the Microsoft Kinect depth sensor and a Robai Cyton Gamma 300 7–DoF humanoid manipulator. In this context, the thesis presents the author’s research into various depth sensor–based hand position and orientation tracking libraries, in addition to a comparison between two integrated development environments, namely Microsoft Visual Studio and the freely available Processing programming language. The software communication approach, coordinate data manipulation, smoothing algorithms and system integration software developed by the present author are all discussed in detail.

Preliminary design work relating to an active collision detection and avoidance algorithm is also considered in the thesis, although not yet integrated into the test system. Nonetheless, the present laboratory version of the proposed system is already capable of providing the operator with an accurate and intuitive control approach for manipulating robotic arms. For example, the system has been successfully used for pick and place operations. For these illustrative experiments, a pinching movement of the fingers is used to close the gripper (a pre–defined control input pair as alluded to above) whilst the new system is employed for the more demanding task of positioning and orientation.