Haptic Telemanipulation with Dissimilar Kinematics Angelika Peer, Bartlomiej Stanczyk, and Martin Buss Institute of Automatic Control Engineering Technische Universit¨ at M¨ unchen D-80290 M¨ unchen, Germany {Angelika.Peer, stanczyk, mb}@tum.de Abstract— This work addresses some practical issues re- garding development of a telerobotic system for 6 degrees of freedom (DoF) tasks. The system consists of a hyper redundant haptic input device ViSHaRD10, a redundant 7DoF manipulator and a stereo vision system. The redundancy of the devices is exploited to assure large convex workspace and singularity-free operation. The anthropomorphic construction of the telemanip- ulator enables intuitive manipulation and increases the ”user- friendliness” of the overall system. As a practical benchmark an assembly experiment in 6DoF for a case of a negligible time delay was successfully performed. Issues regarding inverse kinematics, spatial interaction control, transparency, and intuitiveness of teleoperation are discussed. Index Terms— Teleoperation, Haptic Exploration, Redun- dancy, Compliant Control I. I NTRODUCTION In teleoperation (TO) scenarios tasks are performed by a mechanical manipulator (slave) controlled remotely by a hu- man operator provided with a force reflecting interface (haptic interface or master), which enables the interaction with the remote environment, see Fig.1. In general, the kinematic Feedback Feedback Visual Feedback Auditory Stereo Camera Head e.g. Internet Barrier Command Signal Information Sensor Haptic/ 7DoF Arm Tactile Fig. 1. Teleoperation robotic system structure of both manipulators are different. That is why both devices need to be equipped with an universal kinematic interface, independent from the device structure. There are relatively few works addressing this subject directly. Existing teleoperation systems still exhibit limitations of several kinds. Even without a deep analysis, one can see that in most cases, the functionality is limited to a few DoFs [1] and/or their workspace is relatively small, so that the full spatial immersion is not achieved. Moreover, due to the unmatched workspace of the two coupled manipulators, indexing or shift- ing techniques are used [1], which is unnatural and fatiguing to the operator. The resulting system is neither transparent nor intuitive to the user, and is not acceptable in TO scenarios. As a principal criterium, the ability of reproducing the human motion is considered. A large, convex workspace and the ability to operate in unmodeled environments are crucial requirements for the transparent teleoperation. In the following, we present a teleoperation system of superior performance, with six degrees of freedom, and human sized, singularity free workspace. The main focus of the paper is the development of compliant control strategies and singularity robust kinematic transformations. Sections II and III present the control of a 10 DoF haptic input device and a 7 DoF telemanipulator respectively. The teleoperation experiments are described in section IV. Section V concludes with final remarks and comments on the future research. II. CONTROL OF THE HAPTIC INTERFACE A. General control scheme The haptic simulation of a human’s bilateral interaction with a remote environment requires the control of the motion- force 1 relation between operator and robot. This can be achieved by either controlling the interaction force of the device with the operator (impedance display mode) or the device motion (admittance display mode). Admittance control is particularly well suited for robots with hard non-linearities and large dynamic properties com- pared to the environment being emulated. In this display mode forces are measured and motion is commanded, i. e. the robot acts as admittance and the human as impedance. Accordingly, a force sensor is required for admittance control. Contrary to haptic displays driven in the impedance mode all admittance control implementations aim at forming the closed-loop inertia. The high gain inner control loop closed on motion allows for an effective elimination of nonlinear device dynamics as for instance friction. It is thus possible to render an isotropic closed-loop dynamic behavior in order to provide the operator a more ”natural feeling”. The drawbacks are a reduced capability for the display of low impedances and a decreased closed-loop bandwidth of the force feedback. A more detailed analysis of haptic control schemes is given in [2]. In order to provide an effective compensation of distur- bances due to friction and to be able to render inertia and mass admittance control is implemented as illustrated in Fig. 2. 1 Throughout the paper force stands for both, linear force and torque, while motion in terms of a generalization of position, velocity, and acceleration refers to both, translational and angular motion quantities. 2483 0-7803-8912-3/05/$20.00 ©2005 IEEE 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems