ISBN: 978-93-86256-27-0 824 Proceedings of 6 th International & 27 th All India M anufacturing Technology, Design and Research Conference (AIM TDR-2016) College of Engineering, Pune, M aharashtra, INDIA December 16-18, 2016 Position and Impedance Control of a Multi-Finger Tendon-Driven Robotic Hand I.A. Sainul 1 , A.K. Deb 2 and Sankha Deb 3 1 Advanced Technology Development Centre, IIT Kharagpur, Kharagpur–721302 2 Electrical Engineering Department, IIT Kharagpur, Kharagpur–721302 3 FMS and CIM Laboratory, Mechanical Engineering Department, IIT Kharagpur, Kharagpur–721302 E-mail: 1 sainul@iitkgp.ac.in, 2 alokkanti@ee.iitkgp.ernet.in, 3 sankha.deb@mech.iitkgp.ernet.in ARTICLE INFO ABSTRACT Keywords: Robot Hand Underactuated Tendon Driven Robot Hand Impedance Control PD Feedback Control This paper proposes a control method of an underactuated tendon-driven robot hand. The robot hand consists of a palm and three identical fingers actuated remotely by a tendon and pulley mechanism. One finger has two joints, while the other two fingers have three joints each and the motors are mounted in the palm. All the tendon wires run over the joint pulleys and some routing points inside the finger. The actuation mechanism is underactuated as the number of motors is fewer than the degrees of freedom of the hand. The objective of ourwork is to design a joint space controller. Joint space controller includes joint position control and joint torque control. Proportional Derivative (PD) feedback control law is used for position control and impedance control law is used for force control. One challenge with the position controller is to ensure positive tension in the tendons at all times. Tension force feedback along with position feedback in impedance controller loop are used to keep positive tension in the tendons. All the control laws are simulated with the three-fingered robot hand model. This work also compares the simulated performance of position controller and impedance controller. 1. Introduction There are numerous tendon driven robotic handsthat had been developed over the last few decades. There was no well- established design method for tendon driven mechanism. Ozawa el. al (2009) proposed a compact design method for tendon driven mechanism. They also had analysed kinematics of tendon driven mechanism and classified them into three different classes, namely, full actuated, underactuated, and hybrid-actuated TDM's based on the type actuation. Ulrich et al. (1988) proposed a medium-complexity compliant end effector. The name suggests that the gripper does not support very complex dexterous manipulation task but at the same time it is not like a simple gripper with one or two degrees of freedom. It has medium capability with some extent of the versatility of a complex hand in term of dexterity. The gripper is very effective in term of performance and the control is simple. Ulrich et al. had not discussed about the actuation mechanism used in the fingers. Sainul et al. (2016) proposed a robotic hand actuation mechanism based on previous work by Ulrich et al. (1988). In their work, a tendon driven mechanism had been used and kinematics of hand also discussed. Palli (2006) developed a five finger tendon driven robotic hand with four identical fingers and one opposable thumb. Compliant mechanisms were used at the joint to provide external compliance at the joint and kept overall dimensions similar to the human hand. There are, however, few challenges with the underactuated systems. Joint configurations are not uniquely determined with the tendon displacement, which makes the system uncontrollable within the full work space and can generate only limited sets of joint torque as the joints are coupled with each other. Arai et al. (1991) discussed a method of controlling the position of an underactuated manipulator consisting of active and passive joints. Ozawa el. al (2014) developed a prototype of three finger tendon driven robotic hand and experimentally demonstrated how passivity-impedance and force controllers can be combined to have fast and secure grasps. They had used active and spring loaded passive tendon to incorporate compliant behaviour in the finger. They also showed the advantage of using active and passive tendon which makes the joint configuration uniquely determined by the tendon displacement. Abdallah et al. (2012) implemented a position control scheme along with a force control in two-tier architecture to improve the overall performances and compensate the inability of force controller alone. The two-tier architecture gave better performance with regards of speed, accuracy, and transient response. Use of independent tendon tension controllers in tendon space exhibits transient coupling in the response, Abdallah et al. (2013) proposed joints torque control in joint space instead of tendon tension in tendon space which results a decoupled response. Impedance control law gives great flexibility while interacting with environment, it allows motion control in the presence of contact force due to interaction with the environment [4]. Diftler et al. (2011) implemented dual-priority based control architecture on Robonaut hand which is impedance based control architecture. Wimbock et al. (2008) investigated the nonlinear elasticity properties of tendons and developed a control law based on exponential tendon stiffness.