Design and validation of a novel mechatronic transmission system for a wearable tremor suppression device

Robotics and Autonomous Systems 91 (2017) 38–48

Contents lists available at ScienceDirect

Robotics and Autonomous Systems journal homepage: www.elsevier.com/locate/robot

Design and validation of a novel mechatronic transmission system for a wearable tremor suppression device Yue Zhou a , Michael D. Naish a,b,c,d , Mary E. Jenkins e , Ana Luisa Trejos a,b,d, * a

Biomedical Engineering Program, Western University, London, Ontario, Canada Department of Electrical and Computer Engineering, Western University, London, Ontario, Canada c Department Mechanical and Materials Engineering, Western University, London, Ontario, Canada d Canadian Surgical Technologies and Advanced Robotics, London Health Sciences Centre, London, Ontario, Canada e Department of Clinical Neurological Sciences, Western University, London, Ontario, Canada b

highlights • • • • •

A mono-input-multiple-output mechatronic mechanism for tremor suppression is presented. The power transferred from the power source to the output is controlled by a 2 W DC motor. A single drive motor can be used to support multiple independent outputs. The mechanism was validated using tremor data from patients with Parkinson’s disease. An average of 12.4% RMS error was achieved on a dynamic tremor suppression test.

article

info

Article history: Available online 5 January 2017 Keywords: Multi-channel mechatronic splitter Wearable tremor suppression glove Tremor suppression Single-input-multiple-output mechanism Exoskeleton device

a b s t r a c t Traditional treatments, medication and surgery, for tremor management in Parkinson’s disease have shown varying effectiveness carry a risk of significant side effects. Recent research and development of wearable tremor suppression technology have shown a promising third solution for tremor management. This paper presents the design of a novel multi-channel mechatronic splitter (MMS) for use in wearable tremor suppression devices. This mechatronic system allows a single drive motor to support multiple independent outputs. The operation (speed and direction) of the MMS is controlled by a 2 W DC motor. This low power characteristic may provide a promising approach to achieving a prolonged operating life for wearable devices. Furthermore, the size of the MMS can be scaled proportionally according to different applications for optimal performance. This paper describes the design, modeling, implementation and characterization of the MMS. The weight of the MMS prototype is 129 g, the maximum output speed is 120 rpm, and the maximum continuous torque is 0.15 Nm. In addition, recorded tremor motion along with voluntary movement from 7 individuals with Parkinson’s disease was used to validate the performance of the MMS. The MMS was controlled to suppress tremor motion while following the voluntary movement of the subject. An average of 12.4% RMS error in voluntary motion tracking was achieved on a dynamic tremor suppression test. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Tremor, the most common movement disorder, is described as an unintentional, rhythmic, oscillatory movement with a particular amplitude and frequency [1,2]. Its manifestation is primarily generated by the stimulations of reciprocally innervated muscles [3]. Clinically, tremors are categorized as resting tremor and action tremor [4]. Resting tremor is often associated with Parkinson’s

*

Correspondene to: 1151 Richmond Street, London, ON, Canada N6A 5B9. E-mail address: [email protected] (A.L. Trejos).

http://dx.doi.org/10.1016/j.robot.2016.12.009 0921-8890/© 2017 Elsevier B.V. All rights reserved.

disease (PD) and drug-induced Parkinsonism. Action tremor, which is further divided into postural tremor, isometric tremor and kinetic tremor, is commonly associated with PD, cerebellar lesions, essential tremor, enhanced physiological tremor, metabolic disturbance, and drug or alcohol withdrawal. The reported frequencies of tremors are typically in the range of 3–12 Hz [3,5]. Some recent studies have shown that Parkinsonian tremor and essential tremor incorporate multiple harmonics with frequencies that range from 3.5 Hz to 17.3 Hz [6,7]. This high-frequency tremorous movement significantly affects the activities of daily life for patients and may cause social embarrassment leading to avoidance of social activities.

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

39

Fig. 1. Schematic of a hand tremor suppression glove. The dashed line separates the actuation box section and the sensing glove section. The actuation box controls the motion of the joint of interest through a number of nonstretchable cables. The cables are located on both sides of the hand. The ones on the dorsal side control the extension of the joint, and the ones on the palmar side control the flexion of the joint. Sensors embedded in the glove provide position information for each joint.

Tremor is often treated with medication. However, the efficacy of medication is unfortunately low, and carries significant side effects [8,9]. In addition to medication, many researchers have investigated the effects of surgical interventions on tremor, including deep brain stimulation (DBS) and stereotactic thalamotomy [10,11]. These procedures may ease certain tremors, but they carry a potential risk of permanent complications, paresthesia, dysarthria, speech impediments and even stroke and hemiparesis [12,13]. Given the complications and adverse effects of the traditional treatments, a less invasive approach with fewer side effects is needed. Rosen, et al., have demonstrated experimentally that people with pathological tremor are unable to accomplish the activities of daily living (ADL) because the magnitudes of the tremors approach the magnitudes of the ADL [14]. The development of mechanical tremor suppression orthoses was proposed as a possible solution to this problem. Several research groups have developed tremor suppression devices for wrist and elbow tremors [15–26]. An actuator with a controllable viscous rate was designed by Loureiro, et al. [15]. This actuator employed magnetorheological fluids (MRF), the viscosity of which can be controlled in the presence of a magnetic field. The controllability of the MRF is the main advantage over other similar designs [16,17]. However, it is not possible to suppress tremor without impeding voluntary motion using this type of passive actuator. Also, the suppression power of this device is relatively low due to the limited strength of the actuator. Another application of MRF in tremor suppression was investigated by Case, et al. [18]. This actuator was designed using a damper configuration. The performance of the actuator was greatly improved, but the issue of suppressing both voluntary and tremorous motion was not solved. Although passive technology is safer when applied to orthoses, active technology can achieve better performance. A recent study has shown the success of using pneumatic actuators for suppressing tremor [19]. The force generated by the actuator is sufficient to suppress tremor to a great extent. The proposed device is compact and efficient. However, as a wearable device, it is not realistic for the user to carry the required air compressor. In addition, the noise from either the cylinder or the air source may cause embarrassment to the user. More importantly, air leak and valve failure of the pneumatic actuator may result in serious injury to the user. A 3 degree-of-freedom (DOF) Wearable Orthosis for Tremor Assessment and Suppression (WOTAS) was the final product of a five-year European project called Dynamically Responsive Intervention for Tremor Suppression (DRIFT) [20–22]. This device was

developed to monitor and suppress tremor with minimal restriction to voluntary movement and employs both active and passive strategies. The study results showed superior performance of the active strategy (up to 90% reduction rate in tremor amplitude) over the passive approach. However, this device is very bulky and cosmetically limited. These disadvantages make it infeasible for constant wear. A successful demonstration of the application of a pneumatic cylinder for suppressing wrist tremor [19] was extended to elbow tremor [7]. Two more DOF were added to the device to control the flexion–extension of the elbow and pronation–supination of the forearm. The fundamental frequency of the tremor was suppressed to a large degree (96.8–98.8%), but the reduction in the second harmonic was only 52.7%–82%. This may be caused by the slow response of the pneumatic cylinder. Although the reduction in tremor magnitude is good, the issues related to a pneumatic actuator make this device unsuitable as a wearable device. With most of the aforementioned devices, the tremor reduction rate was the primary goal, with the suitability for patient use often being neglected. This has led to many of the devices being bulky and heavy to use. Furthermore, although finger tremor is often significant, existing devices do not compensate for it. Under these perspectives, a wearable tremor suppression glove (WTSG) was conceived in order to fill this gap in the area of tremor suppression. The main challenge for the design of a WTSG and its application in patient use is the size and weight of the actuators, as well as battery life. Thus, it is necessary to develop a novel mechatronic transmission system that meets the requirements for hand tremor suppression with a compact configuration and sufficiently lightweight. 2. Mechanical design 2.1. Conceptual model and prototyping The goal of this work was to design and develop a mechatronic transmission system that can be used in a WTSG. A schematic of the concept of the WTSG is shown in Fig. 1. It consists of a sensing glove and an actuation box. These two sections are connected through nonstretchable cables. The sensing glove contains sensors, tubes that guide the cables and insertion points for all cables to transfer force from the actuation box to the hand. The core component of the actuation box is the mechatronic transmission system. Within the mechatronic transmission system are several modules each, referred to as a multi-channel mechatronic splitter

40

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

Denoting the input speed and torque as ωin and τin respectively, the output speed (ωout ) and torque (τout ) are given as follows:

ωout = sgn (ϕ) · η · µ · ωin τout = sgn (ϕ) · ξ · τin /µ,

(1)

where η and ξ are the efficiency coefficients of the continuous variable transmission (CVT) for velocity and torque transmission. Since the control signal of the MMS is independent from the Power Source (input), the MMS can be used as a unit in a parallel network (Fig. 3) for a WTSG. In the parallel configuration, each MMS unit shares the same power source, while each is controlled by different control signals. Assuming that the impact of the internal power consumption of all MMSs on the power source can be neglected and the power source can provide sufficient power for all units, then the output speed and torque of the ith unit can be given as follows: Fig. 2. Function block of the MMS. Each MMS consists of a speed control block and a direction control block. The control signal manages the direction of the output with respect to the power source and the output speed.

Fig. 3. Signal flow of n MMS units in a parallel network. All MMSs share with one power source. The outputs are independent of each other.

(MMS). The purpose of the MMS is to couple a single input power source to multiple output applications (e.g., one for the wrist and additional ones for the fingers) with the overall goal of reducing the size and weight of the WTSG. It allows each output to operate with an independent direction of motion, velocity and torque. An MMS consists of two functional blocks: the speed control block and the direction control block (Fig. 2). The power source supports the unit with a stable input speed and torque, while the speed control block and the direction control block change the transmission ratio (γ ) and direction of motion (σ ) according to the control signal

ωi = ωin · sgn (ϕi ) · ηi · µi τi = τin · sgn (ϕi ) · ξi /µi .

(2)

With this configuration, all outputs are independent of each other. This feature allows the use of one power source to supply multiple different applications. Moreover, each MMS unit can be customized according to specific application requirements. Since the voluntary motion of the human hand is considered to be a lowfrequency movement and the motion of each joint is independent of the others, the adoption of this parallel configuration is viable for a WTSG. In this case, each application in Fig. 3 would represent a single joint to be controlled. The conceptual model of the MMS and a schematic of its operation are shown in Fig. 4(a). It incorporates a hemispherical disc (input) and a cylinder (output). The disc rotates with constant speed, the cylinder is held against the disc so that the power can be transmitted to the cylinder via friction. The pivot angle of the disc is a control variable that determines the transmission ratio. The transmission ratio, which depends on the operating radius of the disc, is given as follows: rout rout u=η· = sgn (ϕ) · η · rin R · sin (ϕ) rmax = R · sin (ϕmax ) { (3) 1 ϕ>0 0 ϕ=0 sgn (ϕ) = −1 ϕ < 0 where rin is the distance between the contact point and the rotation axis of the disc, rout is the radius of the output shaft, R is the radius of the hemispherical disc, η is the efficiency coefficient, rmax is the maximum operating radius for rin , ϕmax is the maximum pivot angle, and sgn(ϕ ) is a sign function of ϕ . The practical meaning of the sgn(ϕ ) is shown in Fig. 4(b).

Fig. 4. (a) Conceptual model of the MMS. R represents the radius of the hemispherical disc. rin represents the distance between the contact point and the rotation axis of the hemispherical disc. rout is the radius of the output shaft. ωin and ωout represent the input speed and output speed respectively. ϕ represents the pivot angle. (b) Relationship of the hemispherical disc pivot angle with the output shaft speed.

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

41

Fig. 5. Configuration of the MMS. 1. Machine screw, 2. pulley, 3. timing belt, 4. linear bearings, 5. shaft collar, 6. stainless steel shafts, 7. frictional cylinder, 8. grooved shaft collar, 9. cylinder mount, 10. hemispherical disc, 11. ball bearings, 12. disc support, 13. steel shaft, 14. gear set, 15. arc guide rail, 16. shaft carriage, 17. nonstretchable cable, 18. power motor, 19 . steering motor.

Fig. 6. (a) Configuration of the gear–spool–cable mechanism. 1. Circular groove, 2. arc guide rail, 3. shaft carriage, 4. nonstretchable cable. (b) Prototype of the gear–spool–cable mechanism with installation sequence. (c) Cable winding method for the shaft carriage.

Based on the conceptual model introduced above, a CAD model was created (Fig. 5). The hemispherical disc (#10), a ball bearing (#11) and a disc support (#12) are connected through a steel shaft (#13). There are two ball bearings placed on each side of the disc support, allowing the hemispherical disc to pivot along the major axis of the disc support. The cylinder incorporates four linear ball bearings (#4) on each side of the central shaft. The central shaft is supported by two holders (#9). The use of four linear bearings guarantees that the contact force between the disc and cylinder is distributed evenly on the supporting shafts, furthermore, it reduces the torque requirement for the steering motor. A gear–spool–cable mechanism is used to control the pivot angle of the disc (Fig. 6(a)). There are two circular grooves that function as a spool on either side of the larger gear. A nonstretchable cable (#4) is wound on the spools with both ends fixed to the

gear (#14). The midpoint of the cable is fixed inside of the carriage (#3) to transmit the motion to the hemispherical disc. The angle range of the arc guide rail (#2) determines the pivot angle range of the hemispherical disc. The use of a nonstretchable cable provides good power transmission. In order to minimize the effect of backlash, appropriate tension was applied to the cable [27]. To attach the cable (0.2 mm R diameter Spectra⃝ fiber with a 4.5 kg breaking strength) to the gear, two 0.8-mm diameter bores were drilled into the aluminum gear (Fig. 6(b)). The cable was first glued onto the bore ① on the back of the gear, then wound through the shaft carriage ② in order to provide enough force to the disc shaft, and finally glued to the bore ③. A schematic of cable winding on the shaft carriage is shown in Fig. 6(c).

42

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

Fig. 7. Prototype of the MMS. Side views of the MMS actuation system are shown at the top. Side view of the passive cylinder is shown at the bottom. The cylinder is fixed onto a base that can be connected to a torque sensor for experimental evaluation.

Based on the CAD model, a prototype was created as shown in Fig. 7. As a proof of concept, the power motor was connected to only one MMS. The frictional cylinder was installed on a cylindrical base that can be connected to a torque sensor for experimental evaluation. The dimensions of the current MMS prototype are given in Fig. 7; the miniaturization of the MMS will be implemented in the future study that incorporates it into a WTSG prototype. The total weight of the current MMS prototype is 129 g. This weight does not include the weight of the power motor support. The majority of the weight is from the steel support that constrains the motion of the disc support to only 1 DOF. The weight can be reduced to 112 g if the disc support is machined using aluminum. 2.2. Frictional material selection for the MMS Friction plays an important role in the MMS. In order to achieve good transmission efficiency, high friction must exist between the disc and the cylinder. Considering the small size of the MMS, especially with further miniaturization, and the materials used to manufacture its components, it is important to choose a proper coating material for the disc and the cylinder. An experiment was designed to select suitable coating materials. The experimental setup includes a weight hanger (99.8 g) and weight set which includes 50 and 100 gram weights. The weight hanger was connected to the grooved shaft collar (10 mm radius) through a nonstretchable cable. Four different materials were selected and tested on the disc and two materials were tested on the cylinder. For consistency, the contact force between the disc and cylinder was held constant for all trials. The four materials tested on the disc were Ecoflex 00-30 silicone, natural rubber, a commercial mounting tape and 180 grit silicon carbide. The two materials tested on the cylinder were neoprene rubber and mounting tape. A total of 20 trials were performed for each material combination. The disc shaft was first fixed at the midpoint of the arc guide rail to guarantee the cylinder’s speed remaining at zero. Weights were then mounted

Fig. 8. Friction comparison between different materials for the disc and the cylinder. Data points represent the mean and bars the standard deviation for the 20 trials conducted with each material combination. The red curve shows the results for the rubber, and the blue curve shows the results of the mounting tape. A clear increase in maximum weight is observed from the silicon to the silicon carbide. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

on the hanger until the cylinder started to rotate. The mean and standard deviation (SD) of the maximal applied weights for the different materials are shown in Fig. 8. The results showed that the silicon carbide coated disc generates the highest friction on the mounting tape covered cylinder. However, it was found during the evaluation process that the high compliance of the mounting tape introduces large nonlinearities to the system, and the thickness of the mounting tape has a large

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

43

Fig. 9. Hemispherical disc coated with 180 grit silicon carbide. The top view is shown on the left and the side view is shown on the right. R = 12.5 mm, rmax = 10 mm.

impact on the effective radius of the disc. Therefore, the silicon carbide coated disc (Fig. 9) and neoprene rubber cylinder combination was adopted for the MMS.

Fig. 10. (a) Configuration of the actuation system of the MMS. The rotation axes are labeled in red. ωs is the shaft pivot speed and ωs is the steering motor rotary speed. 1. Timing belt pulley, 2. timing belt, 3. power motor, 4. steering motor, 5. gear set, 6. shaft carriage. (b) Diagram of the gear–spool–cable mechanism. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3. The actuation system The actuation system of the MMS incorporates a single power motor (Fig. 10(a), #3) and a steering motor (#4) for each desired output. The power motor is a brushless DC motor (EC 16 brushless R DC motor, Maxon Motors⃝ , Switzerland) with a gearbox that provides a reduction ratio of 84:1. It provides constant power to the disc through a timing belt (#2). The steering motor is a miniature R brushless DC motor (EC 6 brushless DC motor, Maxon Motors⃝ , Switzerland) with a gearbox that provides a reduction ratio of 3.9:1. The continuous torque is 1.68 mNm. The use of four linear bearings in the cylinder reduces the torque requirement on the steering motor as the static friction between the cylinder and the disc changes to rolling friction. Furthermore, similar to the reduced torque required to steer a vehicle when it is moving, the MMS requires less torque to steer when the disc is rotating. Once power is transmitted from the power motor to the disc, only a small amount of torque is required to pivot the disc. The designs of the disc, elongated disc shaft, the gear set and the linear bearings contribute to reducing the power requirement of the steering motor. A diagram of the power reduction mechanism is shown in Fig. 10(b), and the reduction ratio is given below,

µ=

ωs rm · rs = , ωm rg · ls

(4)

where ωs and ωm are the shaft pivot speed and steering motor rotary speed, respectively, rm is the radius of the gear on the steering motor, rs is the radius of the spool on the bigger gear, rg is the radius of the bigger gear, and ls is the length of the disc shaft from the pivot center to the point where the nonstretchable cable and the shaft intersect. The relationship between the pivot angle (ϕ ) and the steering motor rotation (Nm ) can be defined as follows:

ϕ=

2 · π · rs · rm rg · ls

· Nm .

(5)

The lateral pushing force generated on the disc can be defined as follows: Fd =

τm µ · ld

(6)

where τm is the steering motor nominal torque, and ld is the length of the disc shaft from the pivot center to the surface of the disc. The calculated maximal lateral pushing force is 8.1 N. Here, rm = 2.5 mm, rs = 5.25 mm, rg = 15.5 mm, ls = 53 mm, ld = 13 mm, and τm = 1.68 mNm (with a gearbox efficiency of 0.88). This

pushing force is sufficient to move the cylinder with a maximal load of 162 N (0.025 rolling friction coefficient and a 0.5 efficiency). Accordingly, the maximal speed of the cylinder is 229 mm/s (steering motor nominal speed is 41,000 rpm). Normally, the voluntary human motion has frequencies of less than 3 Hz. In order to keep the output of the MMS following the voluntary motion while suppressing tremor motion, the frequency of movement for the cylinder should be greater than 3 Hz. The travel distance of the cylinder is 26 mm, therefore, the minimum speed for the cylinder is 78 mm/s, which is the product of the travel distance of the cylinder (26 mm) and the frequency of movement (3 Hz). Thus, this configuration is sufficient to be used in a tremor suppression application. Lastly, an example application of the MMS in a WTSG is shown in Fig. 11. There are two MMSs assembled in the actuation box, allowing two joints (e.g., thumb and forefinger flexion/extension) to be controlled. Both MMSs share the same power source, following the network configuration shown in Fig. 3. The actuation box is attached to the dorsal side of a forearm. The output of each MMS is connected to two nonstretchable cables (indicated in green in Fig. 11) that operate in an antagonistic manner. When the output of the MMS turns counterclockwise (as shown in Fig. 11), the upper cable is in tension (and lower cable is slack), resulting in an extension of the associated joint. Similarly, a clockwise output of the MMS places the lower cable in tension thereby flexing the associated joint. 4. Experimental platform 4.1. Experimental setup and data processing In order to evaluate the performance of the MMS, an evaluation platform was designed (Fig. 12). It contains three motion controllers (EPOS 24/1, EPOS2 24/2 and ESCON 36/3 EC, Maxon R R Motors⃝ ), an incremental rotary encoder (E6C3-CWZ, OMRON⃝ ), R ⃝ a rotary torque sensor (FSH02056, FUTEK ), a data acquisition R (DAQ) card (USB-6002, NI⃝ ), one microcontroller unit (MCU) R R ⃝ (STC89C52RC, STC MCU ), and a power supply (BK PRECISION⃝ ). A USB–Serial adapter was used to connect the EPOS 24/1 motion controller to a PC. The ESCON motion controller and the DAQ card communicate with the PC through USB directly. The motion controllers are supplied with 24 V, the torque sensor is supplied with 12 V, and the DAQ card and the MCU are supplied with 5 V.

44

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

Fig. 11. Example application of the MMS in a WTSG. Two MMS are attached to an arm CAD model. The green cables are the nonstretchable cables that transfer torque from the MMS to the controlled joints. Cable 1 and 3 control the extension of the associated joints; cable 2 and 4 control the flexion of the associated joints. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. Experimental setup for MMS evaluation. Left: front view of the evaluation platform, which consists of a power supply, an encoder, a torque sensor, two motors, an MMS and two motor controllers. Right: bottom view of the evaluation platform, which consists of a DAQ card and an MCU.

The purpose of the DAQ card is to capture the torque signal, the velocity of the steering motor and the velocity of the cylinder. Since the DAQ card has only one counter, an additional MCU is introduced to calculate the cylinder’s direction of rotation from the output of the encoder. The direction of rotation is signaled to the DAQ card with digital 0 and 1. The DAQ’s Analog Input One is connected to the torque sensor’s output, and Analog Input Two is connected to the ESCON’s analog output one, which exports the speed of the steering motor. All signals were recorded directly by a PC using LabView R (Version 2014, NI⃝ ). The sampling frequency was configured to 1 kHz. Data processing and analysis were performed offline using

MATLAB (Version R2013a, The Mathworks, Inc.). Prior to data analysis, each signal was low-pass filtered with a cutoff frequency of 30 Hz. 4.2. Control system According to the operating principle of the MMS, the cylinder speed is controlled by the pivot angle of the disc, while the pivot angle is proportional to the rotation angle of the small brushless DC (BLDC) steering motor. Therefore, in order to control the position of the cylinder, a double position–velocity control algorithm is applied (Fig. 13). In the outer position–velocity control loop, the position error is fed to a speed PI controller with the position coefficient (K1 ). The cylinder’s actual position (pc ) and speed (ωout )

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

Fig. 13. Block diagram of the double position–velocity control algorithm. Pe represents the expected position of the cylinder, Pc represents the measured cylinder position, ωc represents the measured cylinder speed, ωs represents the measured steering motor speed, and ωp represents the measured power motor speed. Table 1 Parameters of the double-closed loop. Parameter

Value

K1 K2 Proportional gain, kc Integral time, Ti Derivative time, Td Sampling frequency

6 20 0.006 0.08 0 1 kHz

come from the encoder. In the position–velocity control loop of the MMS, the function G converts the cylinder speed to the disc’s pivot angle using the current power motor’s velocity (ωp ), G = sin

−1

(

ωout · η · rout . ωp · R

) (7)

The parameters used in this control system are given in Table 1. 5. Results and discussion The experimental evaluation was divided into two sections: (1) characterization of the MMS specifically, the speed range and step response; and (2) evaluation of the MMS performance for tremor suppression using patient data. The patient data were obtained as part of a previous study [6]. All patients were diagnosed with PD and recruited by a neurologist. The hand that presented the greatest amount of tremor was selected for data collection. During the trial, each patient’s forearm was strapped to a table, and then he/she was asked to reach a lightweight pencil with his/her index finger and thumb, which is placed over his/her hand. This task was repeated 5 times. Since tremor also presents when a patient performs daily activities, such as eating and writing, the aim of this task is to record patient’s voluntary motion and tremor motion together, in order to validate the performance of the MMS for suppressing tremor while allowing voluntary motion to occur. During the task, the position information of the patients’ fingers and wrist were recorded using four 5 DOF electromagnetic motion R trackers (Aurora, NDI⃝ ). The experimental procedures were approved by the Western University’s Human Research Ethics Board. 5.1. Characterization of the MMS First the speed range of the MMS was tested by controlling the pivot angle of the disc. It was moved through its positive and negative extremes, with three repetitions. As shown in Fig. 8, the maximal angular range of the arc guide rail is 68o . Since each side of the shaft carriage is confined by the end of the arc guide rail, the actual operating range of motion for the disc shaft along the arc guide rail was measured as 64o . The power motor was controlled to a constant speed of 20,000 rpm for this test. The disc shaft was initially set to the neutral position, i.e., ϕ = 0◦ . The measured speed range of the cylinder is given in Fig. 14.

45

The bottom of Fig. 14 shows a time series of the rotation of the steering motor (three cycles are shown). The maximum rotation is 6.5 revolutions, which corresponds to 32o of the disc shaft. With respect to the peak value of the rotation, the cylinder speed range was calculated as [−115 ± 8, 120 ± 7] rpm with a sample size of 10. Since the cylinder speed is proportional to the speed of the power motor at a given position, the cylinder speed of the current configuration could reach up to 236 rpm when the power motor operates at its nominal speed, which is 39,400 rpm. It is worth noting that the cylinder speed should reach up to 238 rpm when the edge of the disc contacts the cylinder (with the power motor operating at 20,000 rpm). However, the current design only reaches 120 rpm. This is because the maximal pivot angle of the disc shaft is limited by the arc guide rail, ϕmax = 32◦ . Therefore the effective contact radius of the disc is only 61.3% of the cylinder radius, rmax = 6.6 mm. Furthermore, the precision of the 3D printed disc, the thickness of the coating material, and the surface regularity of the coated disc also affect the effective contact radius. Therefore, the effective transmission ratio is restricted within the range of 0–0.5 for the current design. The advantage of using a smaller angular range for the arc guide rail is the reduced size of the MMS. The response time of the MMS is also shorter. However, the maximum transmission ratio is restricted because the effective radius of the disc (rin ) is limited to a smaller value. In addition to the speed range, the response time of the MMS was also tested. The step response of the cylinder speed and steering motor rotation are shown in Fig. 15. The response time is measured as the time span before 95% of the actual value is reached after the input is applied. Using the current parameters, the response time of the steering motor and cylinder are 160 ms and 200 ms respectively. Ideally, the cylinder speed should reach its maximum value at the same time that the miniature motor rotation reaches its maximum value. However, comparing the top figure with the bottom figure in Fig. 15, it was found that there is a 40 ms time delay in the transmission, i.e., the cylinder reaches the target value 40 ms after the steering motor reaches its target value. This delay may be caused by the inertia of the cylinder or excessive contact surface area. It is important to consider this delay into the MMS control system design, so that the output motion can be predicted ahead of time. This will allow the precision of the motion tracking of the MMS to be improved. Ideally, the contact area between the cylinder and the disc is merely a line. However, different materials have different compliance, with different resultant contact areas. For the current configuration, a neoprene rubber rod was adopted as the cylinder. The compliance of this material is relatively high compared to the other solid materials. This generates an enlarged contact area, which results in a slowness in changing of the effective rin . Therefore, the existence of the time delay may be caused by the material compliance. Finally, the dynamic output torque of the MMS was tested. The disc was positioned at its maximal pivot angle. A weight set hanger R was connected to the grooved shaft collar through a Spectra⃝ fiber. A 700 g weight was first added to the hanger followed by 100 g and 50 g weights until the cylinder started to slip. A total of 20 trials were performed. The maximal lifting force of the MMS was found to be 16.2 ± 1.7 N. The corresponding torque was 0.15 ± 0.02 Nm. According to the specifications of the power motor, the maximal continuous torque is 0.4 Nm. With the current configuration, the output of the MMS only reaches 37.5% of the dynamic torque of the power motor. Feasible solutions to improve the torque transmission efficiency include increasing the contact force and finding an alternative high friction material.

46

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

Fig. 14. Cylinder speed with respect to the rotation of the steering motor. The steering motor was controlled under a dynamic input within the range of -5 to 5 revolutions.

Fig. 15. Step response of the MMS. A step input was given at the time 0.1s. ∆tc and ∆ts represent the response time of the cylinder and the steering motor, respectively.

6. Validation of the MMS on real tremor signals In this section, the performance of the MMS on tremor suppression was evaluated. Since tremor is present, not only when the patient is at rest, but also when they perform daily activities, an experimental procedure was designed in [6] to simulate these conditions. Data recorded from patients while performing a task consisted of both tremor motion and voluntary motion. To simulate these motions in this study, Motor 2 in the experimental platform (Fig. 12) was used to recreate the recorded motion. A cascade position–speed–current control was applied to Motor 2, with a controlled the output torque of 0.05 Nm. The previous study [6] showed that finger tremor acceleration is around 2.1 m/s2 . Considering the weight of a finger and the distance between the center of mass and MCP joint to be 150 g and 5 cm, the estimated tremor torque was calculated as 0.016 Nm. Therefore, the generated torque was sufficient to simulate the actual tremor and the voluntary motion. The filtered voluntary motions of a total of 7 Parkinsonian patients were used as input to the MMS control system. The filtered voluntary motions were acquired using the tremor estimator developed in a previous study [28]. The power transmission between the outputs of the MMS and Motor 2 was achieved using a bevel gear set. The graphical results of a total of 5 trials from one out of seven patients are shown as an example in Fig. 16. The dashed blue

lines represent the angular position (degrees) of the suppressed motion, and the solid red lines show the filtered voluntary movement (degrees) of the patient. For each patient, the Pearson linear correlation coefficient (γ ) and percentage root mean square error (RMSE) were assessed over the five repetitions (Table 2) The mean values of the RMSE and γ of the seven patients are 12.37 ± 3.10 % and 0.971 ± 0.038. This section verified the concept of the MMS in the application of tremor suppression. Comparing the suppressed trajectories and the filtered patient voluntary motions showed that MMS is able to suppress tremor motion while allowing the voluntary motion. As demonstrated in Table 2, an overall error of less than 20% was obtained. Comparing with other studies [19,22] which have had achieved more than 90% tremor reduction, the difference in tremor suppression could be the result of a variety of issues, such as an under-tuned MMS control system, insufficient power transmission from the power motor to the disc, and most likely, insufficient precision in the manufacturing and assembly of the MMS. 7. Conclusion The mechatronic design of the MMS and its evaluation were presented in this paper. The experimental results verified the concept of the MMS for use in tremor suppression. The core components of the MMS consist of a friction drive and a gear–spool–cable

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

47

Table 2 Means and standard deviations of the RMSE and the Pearson linear correlation coefficient for each patient.

RMSE (%)

γ

Patient #1

Patient #2

Patient #3

Patient #4

Patient #5

Patient #6

Patient #7

10.73 ± 1.89 0.986 ± 0.008

11.92 ± 3.46 0.932 ± 0.072

16.60 ± 4.20 0.937 ± 0.039

12.40 ± 1.99 0.982 ± 0.011

12.67 ± 1.89 0.979 ± 0.015

12.08 ± 1.61 0.981 ± 0.019

10.22 ± 2.37 0.997 ± 0.003

Fig. 16. Motion tracking performance, with suppression of tremor, of the MMS using sample data. The dashed blue lines represent the angular position of the suppressed motion, and the solid red lines represent the filtered voluntary motion of a patient. A total of five trials are shown. y axis unit: degrees. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

mechanism. The main feature of the MMS is that its size can be scaled up or down with little change to the transmission ratio. This feature allows the size and weight of the MMS to be customized depending on the application. In addition to the size and weight, another benefit of the MMS is the power consumption. The MMS is controlled by a 2 W miniature motor and the power source is a 60 W motor. The adoption of two MMS’s would significantly reduce the total power consumption compared to an application that uses two 60 W motors. This feature is especially important for wearable devices because battery life is always a significant constraint. Longer working times typically require a larger battery, resulting in a direct increase in the weight of the device. The MMS may provide a promising approach to address this issue. There are currently two issues that affect the performance of the MMS. First, the surface regularity of the coated disc, the thickness of the coating material and the compliance of the cylinder impact the motion transmission. An irregular disc surface increases the nonlinearity of the output, which affects the robustness of the control system. A thick coating material reduces the transmission ratio, and high material compliance prolongs the response time of the MMS. Second, the torque transmission efficiency is limited by the friction and contact force between the disc and cylinder materials. Future development of this work will include further reduction of the device size through configuration optimization, incorporating advanced high friction materials, and integrating the MMS in the WTSG. Acknowledgment The authors would like to thank Abelardo Escoto for his help in printing and machining the components of the MMS Zeel Shah for her help in testing coating materials for the MMS and all of the subjects who participated in the experiments. This work was supported by the Academic Development Fund, Western University,

Proposal ID 32405; the Natural Sciences and Engineering Research Council (NSERC) of Canada under grant RGPIN-2014-03815; the Ontario Ministry of Economic Development, Trade and Employment; the Ontario Ministry of Research and Innovation through the Early Researcher Award, File number ER14-10-159; and by the Peter C. Maurice Research Fellowship in Biomedical Engineering. These funding sources had no involvement in the work performed here in or in the preparation of the manuscript.

References [1] R.J. Elble, W.C. Koller, Tremor, Johns Hopkins University Press, 1990, pp. 54–89. [2] T.A. Zesiewicz, R.A. Hauser, Phenomenology and treatment of tremor disorders, Neurol. Clin. 19 (2001) 651–680. [3] G. Deuschl, P. Bain, M. Brin, Consensus statement of the movement disorder society on tremor, Mov. Disorders 13 (1998) 2–23. [4] S. Smaga, Tremor-problem-oriented diagnosis, Am. Fam. Physician 68 (2003) 1545–1552. [5] P.D. Charles, G.J. Esper, T.L. Davis, R.J. Maciunas, D. Robertson, Classification of tremor and update on treatment, Am. Fam. Physician 59 (1999) 1565–1572. [6] Y. Zhou, M.E. Jenkins, M.D. Naish, A.L. Trejos, The measurement and analysis of Parkinsonian hand tremor, IEEE International Conference on Biomedical and Health Informatics, Las Vegas, NV., February 24–27, 2016, pp. 414–417. [7] B. Taheri, Real-time pathological tremor identification and suppression in human arm via active orthotic devices (Ph.D. Dissertation), Southern Methodist University, Dallas, TX, USA, 2013. [8] O. Rascol, D.J. Brooks, A.D. Korczyn, D.P. De Deyn, C.E. Clarke, A.E. Lang, A fiveyear study of the incidence of dyskinesia in patients with early Parkinson’s disease who were treated with ropinirole or levodopa, N. Engl. J. Med. 342 (2000) 1484–1491. [9] K.E. Lyons, R. Pahwa, Pharmacotherapy of essential tremor, CNS Drugs 22 (2008) 1037–1045. [10] K.E. Lyons, R. Pahwa, Deep brain stimulation and tremor, Neurotherapeutics 5 (2008) 331–338. [11] H. Boecker, A.J. Wills, A. Ceballos-Baumann, M. Samuel, D.G. Thomas, C.D. Marsden, D.J. Brooks, Stereotactic thalamotomy in tremordominant Parkinson’s disease: An H215O PET motor activation study, Ann. Neurol. 41 (1997) 108–111.

48

Y. Zhou et al. / Robotics and Autonomous Systems 91 (2017) 38–48

[12] Y. Katayama, T. Kano, K. Kobayashi, H. Oshima, C. Fukaya, T. Yamamoto, Difference in surgical strategies between thalamotomy and thalamic deep brain stimulation for tremor control, J. Neurol. 252 (2005) 17–22. [13] E.D. Flora, C.L. Perera, A.L. Cameron, G.J. Maddern, Deep brain stimulation for essential tremor: a systematic review, Mov. Disorders 25 (2010) 1550–1559. [14] M. Rosen, B. Aldenstein, The effect of mechanical impedance on abnormal intention tremor, in: Proceedings of the 9th Annual Northeast Bioengineering Conference, Elmsford, NY, March 19–20, 1981, pp. 187–192. [15] R.C. Loureiro, J.M. Belda-Lois, E.R. Lima, J.L. Pons, J.J. Sanchez-Lacuesta, W.S. Harwin, Upper limb tremor suppression in ADL via an orthosis incorporating a controllable double viscous beam actuator, in: IEEE 9th International Conference on Rehabilitation Robotics, Chicago, Illinois, USA, June 28–July 1, 2005, pp. 119–122. [16] M.J. Rosen, A.S. Arnold, I.J. Baiges, M.L. Aisen, S.R. Eglowstein, Design of a controlled-energy-dissipation orthosis (CEDO) for functional suppression of intention tremors, J. Rehabil. Res. Dev. 32 (1995) 1–16. [17] J. Kotovsky, M.J. Rosen, A wearable tremor-suppression orthosis, J. Rehabil. Res. Dev. 35 (1998) 373–387. [18] D. Case, B. Taheri, E. Richer, Design and characterization of a small-scale magnetorheological damper for tremor suppression, IEEE/ASME Trans. Mechatronics 18 (2013) 96–103. [19] B. Taheri, D. Case, E. Richer, Robust controller for tremor suppression at musculoskeletal level in human wrist, IEEE Trans. Neural Syst. Rehab. Eng. 22 (2014) 379–388. [20] E. Rocon, A.F. Ruiz, J.L. Pons, J.M. Belda-Lois, J.J. Sánchez-Lacuesta, Rehabilitation robotics: a wearable exo-skeleton for tremor assessment and suppression, in: Proceedings of the IEEE International Conference on Robotics and Automation, Barcelona, Spain, April 18–22, 2005, pp. 2271–2276. [21] E. Rocon, J.A. Gallego, J.C. Moreno, J.L. Pons, Wearable robots in rehabilitation engineering tremor suppression, in: Tan Yen Kheng (Ed.), Rehabilitation Engineering, INTECH Open Access Publisher, 2009 Available: http://cdn.intechweb. org/pdfs/9316.pdf. [22] E. Rocon, J.M. Belda-Lois, A.F. Ruiz, M. Manto, J.C. Moreno, J.L. Pons, Design and validation of a rehabilitation robotic exoskeleton for tremor assessment and suppression, IEEE Trans. Neural Syst. Rehab. Eng. 15 (2007) pp. 367–378. [23] S.S.C. Lavu, A. Gupta, Active vibration control of essential tremor, in: Proceedings of the 14th National Conference on Machines and Mechanisms, Durgapur, India, December 17–18, 2009, pp. 446–449. [24] S. Kazi, A. As’Arry, M.M. Zain, M. Mailah, M. Hussein, Experimental implementation of smart glove incorporating piezoelectric actuator for hand tremor control, WSEAS Trans. Syst. Control 5 (2010) 443–453. [25] T. Ando, M. Watanabe, K. Nishimoto, Y. Matsumoto, M. Seki, M.G. Fujie, Myoelectric-controlled exoskeletal elbow robot to suppress essential tremor: extraction of elbow flexion movement using STFTs and TDNN, J. Robot. Mechatronics 24 (2012) 141–149. [26] G. Herrnstadt, C. Menon, On-off tremor suppression orthosis with electromagnetic brake, Internat. J. Mech. Eng. Mechatronics vol. 1 (2013) 7–24. [27] J.W. Hurst, The Role and Implementation of Compliance in Legged Locomotion (Ph.D. Dissertation), Carnegie Mellon University, Pittsburgh, Pennsylvania, USA, 2008. [28] Y. Zhou, M.E. Jenkins, M.D. Naish, A.L. Trejos, Design and Validation of a HighOrder Weighted-Frequency Fourier Linear Combiner-based Kalman Filter for Parkinsonian Tremor Estimation, in: 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Orlando, FL, August 1–20, 2016, pp. 5893–5896.

Yue Zhou is currently a doctoral student in the program of Biomedical Engineering. He received the bachelor degree in Automation from the University of Tianjin Polytechnic University in 2011, the MESc degree in biomedical system from the University of Western Ontario in 2015. His research interests include pathological tremor suppression, mechatronic system and sensing system.

Michael D. Naish received the BSc degree in computer science and the BESc degree in mechanical engineering from Western University, both in 1996, the MASc degree in mechanical engineering from the University of British Columbia in 1999, and the Ph.D. degree in mechanical and industrial engineering from the University of Toronto in 2004. Since 2003, he has been with the Department of Mechanical and Materials Engineering at Western University, where he is currently an Associate Professor, cross appointed with the Department of Electrical and Computer Engineering, and is the Director of the Mechatronic Systems Engineering program. He is also a Scientist at Canadian Surgical Technologies and Advanced Robotics, Lawson Health Research Institute in London, ON, Canada, and a licensed Professional Engineer in the Province of Ontario. His research interests include mechatronic systems, device design, sensing systems, surgical training, minimally invasive surgery and therapy, and medical robotics.

Mary E. Jenkins is a Movement Disorders Neurologist and an Associate Professor in the Department of Clinical Neurological Sciences at Western University. She completed her Neurology residency at Western University, followed 2 years of Fellowship Training in Movement Disorders; the first year at Western University and the second year at the University of Rochester, New York. Her research is in Movement Disorders, in particular Parkinson’s disease, focusing on motor control and development of treatments and strategies to improve motor function.

Ana Luisa Trejos is an Assistant Professor in the Department of Electrical and Computer Engineering at Western and an Associate Scientist at Canadian Surgical Technologies and Advanced Robotics, Lawson Health Research Institute. She has expertise in the design, development and testing of medical mechatronic systems. Her research is focused towards evaluating how novel mechatronic devices can improve patient care during surgery, therapy and rehabilitation. This includes the design of smart wearable mechatronic braces that can provide improved treatment options for musculoskeletal disorders. She has 22 journal publications and 35 international conference papers and has won several awards for her contributions.