ARMin – Design of a Novel Arm Rehabilitation Robot Tobias Nef , Robert Riener, Member, IEEE

Abstract—Task-oriented repetitive movement can improve movement performance in patients with neurological or orthopaedic lesions. The application of robotics can serve to assist, enhance, evaluate, and document neurological and orthopaedic rehabilitation of movements. ARMin is a new robot for arm therapy applicable to the training of activities of daily living in clinics. ARMin has a semi-exoskeleton structure with six degrees of freedom, and is equipped with position and force sensors. The mechanical structure, the actuators and the sensors of the robot are optimized for patient-cooperative control strategies based on impedance and admittance architecture. This paper describes the new robot, the mechanical structure, the control circuit, the sensors and actuators and some safety aspects.

I. INTRODUCTION A. Rationale for Movement Therapy Task-oriented repetitive movements can improve muscular strength and movement coordination in patients with impairments due to neurological or orthopaedic problems. Arm therapy is used for patients with paralysed upper extremities after spinal cord injury (SCI) or stroke. Several studies prove that arm therapy has positive effects on the rehabilitation progress of stroke patients (see [1] and [2], for review). Besides recovering of motor function and improving movement coordination, arm therapy serves also to learn new motion strategies, so called “trick movements”, to cope with activities of daily living (ADL). Movement therapy serves also to prevent secondary complications such as muscle atrophy, osteoporosis, and spasticity. It was observed that longer training sessions per week and longer total training periods have a positive effect on the motor function. In a meta-analysis comprising nine controlled studies with 1051 stroke patients Kwakkel et al. [3] showed that increased training intensity yields positive effects on neuromuscular function and ADL. This study did not distinguish between upper and lower extremities. The finding that the rehabilitation progress depends on Manuscript received February 4, 2005. This work was supported in part by the Schweizer Nationalfond Project NCCR on “Neural Plasticity and Repair” and by the Gottfried and Julia Bangerter-Rhyner Foundation. Tobias Nef and Robert Riener are with the Automatic Control Lab, Swiss Federal Institute of Technology (ETH) Zurich, and the Spinal Cord Injury Center, University Hospital Balgrist, Zurich. E-mail: [email protected] and [email protected]

the training intensity motivates the application of robotaided arm therapy. B. Rationale for Robot-Aided Arm Therapy Manually assisted movement training has several major limitations. The training is labour-intensive, and, therefore, training duration is usually limited by personnel shortage and fatigue of the therapist, not by that of the patient. The disadvantageous consequence is that the training sessions are shorter than required to gain an optimal therapeutic outcome. Finally, manually-assisted movement training lacks repeatability and objective measures of patient performance and progress. In contrast, with automated, i.e. robot-assisted, arm training the duration and number of training sessions can be increased, while reducing the number of therapists required per patient. Long-term automated therapy appears to be the only way to make intensive arm training affordable for clinical use. One therapist may be able to train two or more patients in the future. Thus, personnel costs can be significantly reduced. Furthermore, the robot provides quantitative measures, thus, allowing the observation and evaluation of the rehabilitation progress. C. Requirements for a Rehabilitation Robot It is crucial that the robot is adapted or adaptable to the human limb in terms of segment lengths, range of motion, and the number of degrees of freedom (DoF). A high number of DoF’s allows a broad variety of movements, with many anatomical joint axes involved. There is evidence that a therapy focusing on activities of daily living (ADL) not only increases patient motivation but also yields an improved therapeutic outcome compared to single joint movements [4]. This kind of therapy is also called “Motor Relearning Programme”. To allow the training of ADL’s, the robot must be able to move the patient’s arm in all its degrees of freedom and to position the human’s hand at any given point in the three dimensional space. This can be achieved by an end-effector based robot or by an exoskeleton. End-effector based robots are ground mounted and are connected with the patient’s hand or forearm at one connection point. Having more than three degrees of freedom it is possible to control the position and the orienttation in space. The structure of exoskeleton robots resemble to the human arm anatomy. Therefore the arm is connected with the exoskeleton at several points. Several elements must be of variable length in order to adapt the robot to different body sizes.

II. METHODS A. Specification of ARMin To be capable to perform training of ADL’s the robot must be able to move the shoulder (approximated with a three DoF ball joint) and the elbow (approximated with a one DoF hinge joint). The range of motion (ROM) of the robot must correspond to the ROM of the human arm. For good performance of model-based patient-cooperative control strategies, the robot must have low inertia, low friction and no backlash. Furthermore, the motor/gear combination needs to be backdrivable. The dynamics of the robot must be capable to move the patient’s hand with the velocity 1m/s and the acceleration 10 m/s2. Safety must always be guaranteed for both patient and therapist.

module via two hinge bearings. Elbow flexion/extension is realized by a harmonic drive rotary module. All mechanical components are optimized with respect to maximal stiffness and minimal weight. This optimization has been done intuitionally and with simple FEM validation in the CAD environment. a)

b)

c)

d)

B. Kinematics Fig. 2 Custom made mechanical components

A semi-exoskeleton structure was selected for the mechanical structure of the robot (Fig. 1). The robot is fixed via an aluminum frame at the wall with the patient sitting beneath. The patient’s torso is fixed to the wheelchair with straps and bands. The distal part is characterized by an exoskeleton structure, with the patient’s arm placed inside an orthotic shell.

Internal/external shoulder rotation is achieved by a special custom-made upper arm rotary module that is connected to the upper arm via an orthotic shell. For easy access of the patient’s arm, the module is made out of a half cylinder. An inner half cylinder (Fig. 2d) is guided by 32 ball bearings fixed to the exterior wall (Fig. 2c). It is actuated by three steel cables fixed at the two ends of the cylinder and rolled around the extension of the motor shaft (Fig. 3). This guidance allows transfer of static loads in several DoF while remaining backlash-free and enabling low friction circular motion.

Fig. 1 Mechanical structure, sensors and motors

The current version comprises four active and two passive DoF in order to allow elbow flexion/extension and spatial shoulder movements. C. Mechanics A vertically oriented, linear motion module performs shoulder abduction/adduction (axis 1). Shoulder rotation in the horizontal plane is realized by a backlash free and backdrivable harmonic drive module attached to the slide of the linear motion module (Fig. 2a). The interconnection module (Fig. 2b) connects the horizontal arm rotation drive with the upper arm rotary

Fig. 3 Functional principle of the upper arm rotary module

D. Sensors and Control Circuit All four brushed DC motors are equipped with optical incremental sensors for position and velocity measurement. A six DoF force and torque sensor beneath the horizontal arm rotation module measures forces and torques of the shoulder actuation (axis 1-3). The torque of the elbow actuation is measured by a separate torque sensor (Fig. 1). Two custom-made motor modules control two DC motors. Equipped with a digital signal processor (DSP), two

MOSFET full bridges and analog electronics for the sensor interface, the motor module communicate via serial bus (CAN) with the real time target computer system (XPC Target) (Fig. 4).

Fig. 4 Control circuit unit for 2 channels

The user interface runs on a windows machine (ARMin Host) and is connected with the real time target via Ethernet. This so called distributed control structure reduces the length of the analog cables between sensors and the analog to digital converters and, therefore, improves the signal to noise ratio. Furthermore, it allows to run a lower level control loop at very high update rates (>4 kHz). E. Passive and Active Safety Safety was a main issue for the design of the robot. Passive safety features, for example to avoid sharp edges in mechanical construction and mechanical end stops to guarantee that no joint can exceed the anatomical range of motion of the human limbs are combined with active safety features. Thus, four additional redundant absolute position sensing potentiometers allow detecting a malfunction of a position sensor or a controller. Several surveillance routines are included in the software. This comprises current and speed surveillance, a collision detection algorithm and several watchdog systems. Whenever an abnormal event has been detected, the safety circuit will immediately cut the power of the motor drives. As the robot is equipped of passive weight compensation, the robot doesn’t fall down after power loss. Because all drives are backdrivable, the robot can be moved manually by a therapist in order to release the patient from a eventual uncomfortable position. Last but not least, an experienced physiotherapist always observes the training, holding a dead-man’s button in his hands. Releasing the button directly interrupts the motor power and stops the robot immediately. This can also be achieved by pressing one of the three emergency stop buttons. Beside patient’s safety, also safety of the therapist needs to be considered. As the robot cannot know the position of the therapist, it is important that the therapist does know about the danger of collisions with the robot. Nevertheless, the probability of a sever accident is low because of the fact

that the maximal speed of the robot is limited by the surveillance circuit. Furthermore, the therapist can always interrupt the motion by releasing the dead-man’s button. A detailed risk analysis shows that the risk for a patient using the robot is acceptable compared to the expected benefit for the rehabilitation process [5]. F. Control Modes Four different control modes are provided. With the prerecorded trajectory mode the therapist can first guide the human arm and the robot while position data is recorded. Once recorded, the trajectories are repeated with different speeds by the robot. In the predefined motion therapy mode, the therapist can choose between several preprogrammed standard therapy exercises. When the point and reach mode is selected, a graphical display shows desired positions of the human’s hand in space and the patient is asked to move his arm from the basic position towards the desired point. In case that the patient is not able to perform the desired movement, the robot supports and guides as much as necessary. This control mode has been proposed by Hogan et al. [6] and has been tested in two dimensions with the MIT-Manus [7]. For this application, the strategy will be expanded to three dimensions. In the patient guided force supporting mode [8], the trajectory is routed by the patient. The robot measures positions and speeds and predicts required forces and torques using a mechanical model of the patient and the robot. According to an adjustable supporting factor, the robot delivers a part of the required forces and toques. The first two modes are purely position controlled and forces and torques are measured only for monitoring. While these modes can be realized with simple control strategies, model based control strategies based on impedance and admittance control must be developed to realize the other two control modes.

III. RESULTS AND DISCUSSION A first functional model of ARMin has been realized and tested with healthy subjects at the University Hospital Balgrist in Zurich. The robot is fixed to an aluminum frame in order to position the linear motion module 0.6m in front of the wall. This allows the use of almost all available passive and active wheelchairs. A simple experimentally tuned PID controller fulfills the requirements of the prerecorded trajectory mode and the predefined motion therapy mode. Fig. 5 shows three typical examples of reference and measured position trajectories of the upper arm rotary module (Fig. 3), recorded during tests with a healthy person who was asked to relax. As a healthy person is stronger than the robot, the maximal difference between the reference and the measured

position depends on the person’s activity and is for a relaxed person smaller then 2°.

misalignments do not cause stress in the human arm. Shoulder misalignments can result from the anatomical complex structure of the shoulder (5 DoF) or from misalignments of the patient’s body. IV. CONCLUSION AND OUTLOOK

Fig. 5 Three reference (dashed line) and measured (continuous line) position trajectories of the upper arm rotation module recorded during tests with a relaxed healthy person.

First tests showed that the mechanics, the electrical circuit, the control and the software work well. The main focus of these tests was to check whether the movement of the robot and the orthotic cuffs are comfortable for the proband. One important finding is that a passive weight support of the rotary module is needed to release the patient’s shoulder from the robot’s mass. In combination with this weight support, the shoulder actuation mechanism guides the patient’s arm precisely and in accordance to natural shoulder movements.

Fig. 6 ARMin with a healthy subject

Compared to pure exoskeleton structures, the semiexoskeleton structure has the advantage that small shoulder

A new semi-exoskeleton robot called ARMin for arm rehabilitation has been developed. ARMin is fit to perform several therapy modes on patients with impairments due to neurological or orthopaedic problems. Providing movements in four DoF, ARMin can support arm therapy related to activities of daily living. The functionality has been proven with the first realization of the robot. Complex mechanical parts like the shoulder rotation module are working fine. The prerecorded trajectory mode and the predefined motion therapy mode has been validated on healthy subjects. Ongoing work is to improve the mechanical interface between the robot and the patient, to test the robot on patients and to validate the model based control structures.

V. ACKNOWLEDGEMENT We thank the occupational therapists and Prof. Dr. V. Dietz of the Balgrist University Hospital Zürich and Gery Colombo from Hocoma AG for their contribution to this work. REFERENCES [1]

T. Platz, „Evidenzbasierte Armrehabilitation: Eine systematische Literaturübersicht“, Nervenarzt, 74, pp. 841-849, 2003.

[2]

R. Riener, T. Nef, G. Colombo „Robot-aided Neurorehabilitation for the Upper Extremities“, Medical & Biological Engineering & Computing, Jan. 2005

[3]

G. Kwakkel, R.C. Wagenaar, T.W. Koelman, G.J. Lankhorst, and J.C. Koetsier “Effects of intensity of rehabilitation after stroke. A research synthesis,” Stroke, 28, pp. 1550-1556, 1997.

[4]

Langhammer, B., and Stanghelle, J. K. (2000): Bobarth or motor relearning programme? A comparison of two different approaches of physiotherapy in stroke rehabilitation: A randomised controlled study, Clin. Rehabil., 14, pp. 361-369

[5]

T. Nef, R. Riener, „Risikomanagementakte ARMin“, internal report, Automatic Control Lab, ETH Zürich, 2004

[6]

Hogan, N., Krebs, H. I., Sharon, A., and Charnnarong, J. (1995): Interactive robotic therapist. US Patent 5466213

[7]

B. T. Volpe, H. I. Krebs, N. Hogan, L. Edelstein, C. Diels, M. Aisen „A novel approach to stroke rehabilitation“, Neurology, 54, pp. 19381944, 2002

[8]

Riener, R., and Fuhr, T. (1998): Patient-driven control of FESsupported standing-up: A simulation study. IEEE Transactions on Rehabilitation Engineering, 6, pp. 113-124

[9]

T. Nef, R. Riener, H. Wegmann, G. Colombo, “System und Verfahren für die kooperative Armtherapie sowie Rotationmodul dafür”, European patent deposed at the Swiss Federal Institut of Intelectual Property, Dec. 2004