EMG Based Approach for Wearer-centered Control of a Knee Joint Actuated Orthosis Walid Hassani, Samer Mohammed, Hala Rifa¨ ı and Yacine Amirat Abstract— This paper presents a new human-exoskeleton interaction approach to provide torque assistance of the lower limb movements upon wearer’s intention. The exoskeleton interacts with the wearer; the shank-foot orthosis system behaves as a second order dynamic system with gravity and elastic torque balance. The intention of the wearer is estimated by using a realistic musculoskeletal model of the muscles actuating the knee joint. The identification process concerns the inertial parameters of the shank-foot, the exoskeleton and the musculotendon parameters. Real-time experiments, conducted on a healthy subject during flexion and extension movements of the knee joint, have shown satisfactory results in terms of tracking error, intention detection and assistance torque generation. This approach guarantees asymptotic stability of the shank-foot-exoskeleton and adaptation to human-exoskeleton interaction. Moreover, the proposed control law is robust with respect to external disturbances. I. INTRODUCTION Recently, functional assistance-based robots have gained a particular attention as they are seen as an alternative to physiological (endurance) and psychological (repetitive tasks) limitations [1]. They are widely used in the medical domain, mainly for the functional rehabilitation; and in the military field to improve the endurance and to augment the physical forces [2]. They are often called exoskeleton or motorized orthoses in the medical domain. Exoskeletons are mechanical structures that embody the human limbs to ensure functional assistance during rehabilitation or during movement restoration of everyday activities. Exoskeletons should have suitable design able to ensure smooth cogni- tive and physical interaction with the wearer. Ideally, the exoskeleton should respond with precision to the intention of the wearer. Most of the current works focus mainly on the mechanical aspects of exoskeletons in order to optimize the effort transfer and to improve the wearer’s comfort and ergonomy. The human-exoskeleton interface consists of two parts: the first one is physical and concerns measurement of the force interaction between the exoskeleton and the wearer. The second one is cognitive and concerns the control, based on the sensor information measured from the wearer such as kinematics (electrogoniometers), muscular activities (EMG), torque, ground interaction (baropodometric soles), etc. The force sensors are placed between the exoskeleton and the wearer to measure the interaction force. Impedance-based control laws are mainly used to ensure interaction between W. Hassani, S. Mohammed, H. Rifa, and Y. Amirat are with the Laboratory of Images, Signals and Intelligent Systems (LISSI), Uni- versity of Paris-Est Crteil, 122 Rue Paul Armangot, 94400 Vitry-Sur- Seine, France. E-mails: {walid.hassani, samer.mohammed, hala.rifai, amirat}@u-pec.fr the wearer and the exoskeleton [3]. Effort-based strategies are often adopted because force sensors are easy to use and calibrate. Their main drawback concerns the time delay introduced in the closed loop control [5]. Moreover, it is not obvious to differentiate voluntary human efforts from external disturbances that may occur during the movement restoration process. On the other hand, EMG-based control laws are also used to control the wearable robots based on the measurements of the wearer’s muscular activities. In [1], [4], a fixed upper limb exoskeleton is used to assist the flexion/extention movements of the elbow joint. The wearer’s force is estimated from EMG signals and a simplified Hill- type muscle model is used along with neural networks. In [5], a knee joint exoskeleton controlled through the wearer’s intention estimation is proposed. A musculoskeletal knee joint model is introduced to improve the estimation of the wearer’s torque. In [6], the authors proposed a one DOF powered ankle foot orthosis with artificial pneumatic muscle for gait rehabilitation. The control schema is based on the assisting torque estimation by using a proportional controller as a function of the EMG activation level. In [7], an EMG interface is proposed to control two degrees of freedom exoskeleton. This exoskeleton assists flexion/extention and pronation/spination shoulder joint to perform predefined movements. The controller output is only adapted using raw EMG patterns previously learned using neuro-fuzzy model. In [8], raw EMG pattern recognition is used to control a two DOF wrist exoskeleton. A support vector machine is used to classify the EMG signals and to estimate the wearer’s intention. The Hybrid Assistive Limb (HAL) relies on the detection of motion intention and the achievement of the movement task. The so-called voluntary control system estimates the wearer’s intention through the detection of the nerve signals (EMG) [9], [10]. This paper lies in the continuity of previous works [11], [12]. However, in this study, instead of having a predefined position trajectory, the generated movement of the exoskeleton is based on the intention of the wearer and the dynamics of the interaction imposed through a stable control law. The rest of the paper is organized as follows: Section II presents the modeling of the shank-foot exoskeleton interaction dynamics as well as its parametric identification. Section III presents the proposed control law used to generate an appropriate exoskeleton torque with respect to the voluntary wearer muscular ac- tivities. Section IV shows the experiments as well as the robustness tests conducted on a healthy subject and section V presents the conclusion of this study. 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) November 3-7, 2013. Tokyo, Japan 978-1-4673-6357-0/13/$31.00 ©2013 IEEE 990