Abstract
The preliminary experimental study toward the implementation of an arm rehabilitation device based on a twisted string actuation module is presented. The actuation module is characterized by an integrated force sensor based on optoelectronic components. The adopted actuation system can be used for a wide set of robotic applications and is particularly suited for very compact, light-weight, and wearable robotic devices, such as wearable rehabilitation systems and exoskeletons. Thorough presentation and description of the proposed actuation module as well as the basic force sensor working principle are illustrated and discussed. A conceptual design of a wearable arm assistive system based on the proposed actuation module is presented. Moreover, the actuation module has been used in a simple assistive application, in which surface-electromyography signals are used to detect muscle activity of the user wearing the system and to regulate the support action provided to the user to reduce his effort, showing in this way the effectiveness of the approach.
1. Introduction
Research laboratories worldwide are concurrently designing a novel era of robots much more advanced than their predecessors in terms of cognitive capabilities and are able to adapt to constantly changing environments. These new robots have been intentionally designed for physical interaction with unstructured environments and humans such as servoactuated prostheses and exoskeletons. An exoskeleton is a wearable robotic system usually composed of an external structural mechanism with joints and links corresponding to those of the human body and provided with a suitable actuation system to support the user’s movements. In these devices, the force and torque measurements are used to support and fulfill the human goals. Exoskeletons can be applied to a number of fields, and in particular in rehabilitation and haptic applications, ultimately benefiting all populations, both disabled and healthy ones.
In this paper, the design of a novel twisted string actuation (TSA) module is presented. The main purpose of this device is the development of wearable assistive and rehabilitation systems. Due to its particular structure, TSA is well suited in case the force is directly applied to the user’s limbs, without external rigid structures supporting the limb and/or the actuation, as schematized in Figure 1. The design of the proposed wearable assistive and rehabilitation system aims to (i) remove all rigid joints to have a lighter and more comfortable system, which can be easily adapted to any user; (ii) decrease the weight, size, and mechanical complexity of the exoskeleton, avoiding complex regulation mechanisms, to reduce the costs and improve its reliability and affordability; (iii) design a modular actuation system that can be reused for implementing different assistive movements. The proposed TSA module is characterized by an integrated force sensor and embedded acquisition and control electronics, and its main components and structure are conceptually depicted in Figure 2.
The paper presents a complete overview of the actuation module structure and the integrated force sensor based on optoelectronic components and the sensor working principle. The motor module structure is manufactured in ABS plastic by 3D printing, and a commercial optoelectronic component, called light fork because of its forked structure, has been adopted for the force sensor implementation. The light fork hosts in its package both the Light Emitting Diode (LED) and the Photodetector (PD) required for measuring the compliant frame deformation. This solution presents several advantages with respect to alternative strain-gauge based or optoelectronic based solutions, such as a simpler assembly, because of the well-defined package, high sensitivity, and low cost. The module is driven by a low-cost DC motor, and a combined axial-radial bearing is adopted to support the output shaft at the location of the twisted string connection. Moreover, the actuation module is also provided with an optical encoder for output shaft position measurement. The Finite Element Analysis (FEA) together with experimental measures has been executed to validate the actuation module design, and the sensor calibration process has been performed. To validate the properties of the proposed TSA module, two experiments have been performed in dynamic conditions, and the results have been validated by means of a reference strain-gauge based force sensor. Finally, as an application example, the TSA module has been used together with a surface-ElectroMyoGraphic (sEMG) signal acquisition system to implement an assistive device able to support the user during a lifting task involving his/her forearm, and preliminary results are reported.
The paper is organized as follows. Section 2 reviews some related works, whereas Section 3 describes the mechanical design of the TSA modules and the working principle of the integrated force sensor. In Section 4 the force sensor calibration and the experimental evaluation of the TSA module are presented, while in Section 5 the results about the use of the proposed device for the assistive application is reported. Finally, Section 6 draws the conclusions and outlines future research activities.
2. Related Works
In more recent years, the use of exoskeletons for the hand and lower/upper-limb support applications has drastically increased [1–3]. In [4], an upper-limb power-assist exoskeleton by pneumatic muscle actuation with two metal joints was developed. An integrated cable-driven, low-cost, and light-weight wearable upper body orthotics system that can be worn over the upper body to generate effective torques to move the arm through a set of assistive motions was introduced in [5]. In [6] a wearable upper body exoskeleton system incorporating a model-based compensation control framework to assist robot-aided shoulder-elbow rehabilitation and power assistance functions is presented. An anthropometric seven-degree-of-freedom powered exoskeleton for the upper limb based on a database defining the kinematics and dynamics of the upper limb during daily living activities is presented in [7]. In [8], a cable-driven arm exoskeleton (CAREX) was developed to achieve desired forces on the hand, that is, both pull and push, in any direction as required in neural training while a 6-DOF upper-limb exoskeleton robot [9] and a muscle suit with providing human physical support was developed in [10].
The TSA concept implements a light-weight, low-cost, and compact linear transmission system [11]. With an appropriate choice of string parameters (in particular the string radius and length) and a rotative electric motor, it is possible to fulfill the generally stringent requirements for the implementation of miniaturized and highly integrated mechatronic devices. Moreover, the slender structure of TSA makes it particularly suitable for wearable devices. The measurement of the actuation force plays a crucial role in the control of this actuation system, as reported in [12]. Previous studies on integrated actuation module for TSA are reported in [13–15].
Many different uniaxial force gauges are available on the market, with the majority of them based on strain gauges that provide exceptional linearity at the expense of intricate electronics for signal acquisition and sensitivity to electromagnetic noise. However, several advantages can be seen through alternative solutions, such as piezoelectric-based force sensors [16] or force sensors based on optoelectronic components [17, 18], mainly because of easier integration and simpler electronics. Optoelectronic sensors exploit the modulation and/or reflection of a light beam emitted by a source and received by suitable detectors to directly measure the deformation of a compliant structure or the relative displacement between elastically coupled elements caused by the external forces. An optical micrometric force sensor based on the differential measures of the light intensity is reported in [19]. Moreover, a sensor based on changing the coupling of optical power between a photodiode and a vertical-cavity surface-emitting laser is presented in [20]. In [21] the movements induced on a compliant structure by external forces and torques along 6 axes are detected by exploiting multiple Photodetectors and a single LED light source. Sensors based on the use of discrete optoelectronic components for the measurements of the tendon force have been implemented in [22, 23].
3. System Description
The designed system is conceived for rehabilitation and assistive applications, as schematically reported in Figure 1, where a twisted string actuator with an integrated force sensor is mounted on the back of the user and connected to a forearm through a shoulder path. The designed TSA module fits very well with this kind of application due to its light and compact structure and the ability of acting similarly to human muscles, as will be shown in the experiments described in Section 5.
3.1. Sensor Working Principle
The optoelectronic sensor is described in detail in [24], and therefore here only a brief summary of its main features is reported. The basic idea is to have LED illuminating a photodiode (PD), where the current flowing through the PD can be modulated by means of a mechanical component that partially intercepts the light emitted by the LED. The position of the mechanical component depends on the deformations occurring on the sensor’s body as a consequence of the application of an external force, Figure 3.
The OMRON EE_SX1108, called a light fork because of its forked structure, is the sensitive element used for the implementation of the proposed sensor. This component is embedded with both LED and PD (a phototransistor) facing each other and is provided with a window restricting the light cone coming from the LED to the PD to a thin and well-focused region.
Several advantages accompany the use of this component. A compact sensor implementation is permitted because both the LED and the PD are embedded together in a compressed structure with precise relative position. The light cone is well shaped, providing a very steppy and linear transition region between the fully covered and the fully free light conditions and allowing good sensitivity and linearity to be obtained.
3.2. TSA Module Design
A schematic view of the designed TSA module is reported in Figure 2. With reference to this picture, the TSA module is composed, from left to right, by (i) a connection element to connect the module to the supporting frame; (ii) the force sensor to measure the actuation load; (iii) a frame hosting the DC motor, the output shaft where the twisted strings are connected, and all the electronics; (iv) the twisted string itself connecting the motor module with the load, and the load itself represented in Figure 2 as a translating mass. The basic TSA concept and modelling can be found in [12]. A couple of strings are attached to a rotative electrical motor and twisted on one end, whereas on the other end the strings are connected to a linear moving element, that is, the load. The overall string length is reduced by the rotation produced by the electrical motor. Therefore, the rotative motion of the electric motor is converted into a linear motion on the other side of the strings.
Figure 4 reports a detailed 3D view of the TSA module design. Its mechanical structure is manufactured by rapid prototyping in ABS plastic. A pair of axial-symmetric compliant beams has been integrated in the TSA module frame on the opposite side with respect to the twisted strings. These beams behave as a linear spring, granting a properly designed compliance to the structure required for converting the force exerted by the TSA module into a proper frame deformation and thus to the implementation of the force sensor. The integration of a force sensor into the TSA module is of paramount importance to successfully measure the force the actuator applies to the load. In the proposed TSA module, the force sensor is located in between the frame connection point on the robot structure and the frame hosting the DC motor, that is, on the opposite side of the twisted strings with respect to the rotative motor, as can be seen in Figure 4. A picture of the TSA module prototype developed in this work is reported in Figure 5. In particular, Figure 5(a) shows also the TSA module embedded controller based on an Arduino NANO board. In the TSA module, an optoelectronic device is then used to detect the frame deformation and convert it back into the applied force causing the deformation. As shown in Figure 5, a DC motor is hosted in the module frame along with an optical encoder for motor angular position sensing, while the output shaft is supported by a combined axial-radial bearing at the point of the twisted string connection to both reduce the friction and prevent the transmission force to damage the motor. A silicon tube is used to connect the DC motor and the module output shafts in order to solve problems related to misalignment of the rotational axes of the motor and the module output shaft. The designed TSA module structure allows the transmission force to be completely supported by the output shaft through the combined bearing, whereas the motor is only used to transmit the necessary torque for driving the twisted string actuation to the output shaft. In this setup, all of the components of the actuation system are integrated in a single element, thus optimizing the system encumbrance. A digital interface to input and output signals required for the actuation system control is provided by the actuator electronics, including both the motor power electronics, the conditioning system for both the optical encoder and the force sensor, and a microcontroller board for controlling the TSA module (see Figure 5). The communication of the TSA controller with external systems can be implemented either with UART, SPI, or I2C interfaces.
(a) Upper view of the TSA module prototype and control electronics
(b) Lateral view of the TSA module prototype
4. Evaluation of the TSA Module
In this section, the experimental evaluation of the proposed actuation module is investigated.
4.1. Experimental Setup
Figure 6 depicts an overview of the experimental setup used for the validation of the proposed TSA module design. In this setup, the TSA module rotation axis is aligned with the translation axis of a linear motor (LinMot-37 × 160), and the twisted string is connected on one side to the TSA module and on the other side to the linear motor slider. The linear motor is able to apply a maximum continuum force up to 160 N and is equipped with a load cell able to measure a maximum force up to 100 N and an integrated encoder with resolution of 1 μm to measure the load position. The linear motor is therefore used to apply variable and controllable loads to the TSA module structure. A shielded cable for noise rejection connects the linear motor load cell to the amplifier. The linear motor is driven by a low-level control system able to generate different force profiles compensating for the friction acting on the motor slider [25]. Data collection and control of the system during the experiment are managed by a PC-104 running RTAI-Linux real-time operating system and a sensoray 526 data acquisition board. The same PC-104 is used to communicate commands and data with the TSA module embedded controller through a high-speed SPI serial connection.
4.2. TSA Design Validation
The TSA module structure is designed to exhibit a symmetric deformation along the direction of the measured force (i.e., the -axis). Moreover, to fully exploit the characteristics of the optoelectronic component used for detecting the frame deformation, the mechanical structure has been designed to attain a proper linear displacement within the TSA module force range. Both experimental measurements and FEA have been used to validate the proposed actuation module design. In particular, Figure 7 reports the FEA of the TSA module at the maximum load condition, that is, with a load of 80 N. This limit is not given by the TSA capabilities, but it is limited by the control system to preserve a suitable lifetime of the twisted string. It is also important to say that even if the verification of the frame deformation is not significant in that case, the TSA module structure has been designed to sustain a maximum load 4 times higher than the previously mentioned force range to prevent damage to the mechanical structure in case of overload. Figure 8 portrays the comparison between the measured deformation and FEA. The result evidences that the measured deformation is slightly smaller than the FEA, which can be influenced by the manufacturing process. The maximum deformation with the maximum load of 80 N is m indicating the compliant frame deformation is within the goal working region < 0.12 mm and can be considered linear.
4.3. Input-Output Characteristic
To evaluate the input-output characteristic of the system in dynamic conditions, suitable experiments have been performed. A force feedback loop based on the data gathered from the load cell mounted on the linear motor slider controls the force applied to the TSA module.
In the first experiment set, the linear motor is programmed to apply a constant load force of 5, 20, and 40 N, respectively, to the TSA module, and the rotative motor is then commanded to move with a trapezoidal trajectory from the initial zero position (untwisted string) to a maximum value of 540 rad, and the displacement caused on the load is then measured by means of the linear motor integrated encoder. Figure 9(a) presents the input-output position characteristic of the proposed TSA module with a constant load.
(a) Experimental evaluation of the constant load
(b) Experimental evaluation of the spring-like load
(c) Input-output position characteristic with constant load (right) and with spring-like load (left)
Different load conditions have been adopted during the second experiment set. In this case, the linear motor is programmed to act as a mass-spring-damper system with the load stiffness of 1000, 2500, and 5000 N/m, respectively. Figure 9(b) presents the input-output position characteristic of the proposed TSA module with spring-like load. It can be noted that the maximum position reached by the load visibly decreases as the input load increases.
For a better evaluation of the input-output characteristic of the TSA module over multiple working cycles, the relation between rotative motor position and load (i.e., linear motor) position has been reported in Figure 9(c) for all the performed experiments. These plots highlight that a certain variation of the input-output characteristic is present, due to settlements in the twisted string coil arrangement during the initial motion phases. As can be also seen in Figures 9(a) and 9(b), this phenomenon tends to vanish as the number of cycles increases.
5. Experimental Evaluation of the TSA Module as an Assistive Device
To show the effectiveness of the proposed system for rehabilitation and assistive applications, the TSA module has been tested on a user as a support to the biceps muscle activity during a load lifting task. In particular, in this test application, sEMG signal has been used to drive the assistive device. As a preliminary step toward the implementation of an elbow flexion/extension assistive system, in which we considered the installation of the device on the back of the user; see Figure 1, the TSA module has been fixed to a rigid structure (see Figure 10) to simplify the experimental test, in order to use the necessary string length without considering friction and curvature related phenomena introduced by the twisted string path across the shoulder, as can be seen in Figure 1. To solve this issue, dedicated experiments to study the effect of friction on the TSA transmission characteristics are under development [26]. In this preliminary experimental evaluation, the TSA module operates on the user’s forearm and is regulated using a variable stiffness control approach. In this relation, the biceps sEMG signal is used within a specific PI (Proportional-Integral) control scheme in order to suitably adjust the device stiffness to obtain a load partial compensation, limiting the sEMG activity under a suitable threshold value.
5.1. Hardware and Setup Overview
The experimental setup used during this experiment is shown in Figure 10. With reference to this picture, a rigid structure has been implemented to allow the TSA module to be firmly fixed in a point above the user’s shoulder, such that the string can be in tension with respect to its full length. In particular, the length of the string for this experiment has been chosen equal to 1 m, in order to exploit the mechanism contraction range related to the best module working condition, that is, the 20% of the full string length [12]. The string is therefore connected to an armband fixed on the forearm of the user.
The sEMG signal has been acquired by means of low-cost disposable surface electrodes connected to the acquisition board Cerebro [27]. The sEMG signals are acquired by a high-performance Analog Front End (AFE) [28] that is connected with an ARM Cortex M4 Microcontroller. In this application, the data are sampled at 1 kHz and streamed to a nearby PC using an onboard Bluetooth interface. Before using the sEMG signals in the TSA variable stiffness control scheme, a processing step is necessary. The following filtering procedure is applied to the signals [29]: (i) a 50 Hz notch filter for powerline interference cancellation, (ii) a 20 Hz high-pass filter, (iii) the rectification (absolute value of the signal), and finally (iv) the Root Mean Square (RMS) value is computed on a 200 ms window.
5.2. Experiment Configuration, Control Scheme, and Results
A healthy male subject is involved in the experiment, in particular, the user put on the TSA module, which is connected to the rigid structure and, on the other string end, to the armband fixed to his forearm. During the experiment, the subject is sat in a normal and comfortable position, holding his upper arm parallel to the trunk. A couple of differential sEMG electrodes have been placed in proximity of the biceps brachii muscle in the upper arm, referring to methods and best practices outlined in [30]. In this validation test, the goal is to assist the biceps during the flexion of the elbow for lifting tasks of a load of 2 Kg applied on the wrist. For this reason, only the biceps sEMG activity is acquired, considering the triceps activation negligible. Specifically, the subject is required to perform five consecutive forearm motions, each one consisting in a flexion followed by an extension, covering the elbow angle range between approximately 10° and 90°. The user is also requested to perform the movements smoothly and with a reasonably low velocity. Once the 2 Kg load is applied to the forearm by means of a wristband, in the first part of the experiment the user executes the flexion/extension motions freely, without the help of the TSA module. The behaviour of the biceps activity during this task can be observed in Figure 11 looking at the acquired sEMG signal and where the elbow angle is computed from the module’s DC motor encoder signal and the setup geometry. In the second phase of the experiment, the subject connects his forearm to the TSA module in order to be assisted during the lifting task of the 2 Kg load. To achieve this goal, the TSA module is controlled through a specific sEMG-driven control scheme illustrated in Figure 12(a). With respect to this figure, a stiffness control is implemented for the TSA module, which means that the device is regulated in such a way to behave like a spring with stiffness and contraction given by the difference between the actual string length (spring length) and an initial string length less than or equal to the length in the case of an elbow flexion of 90° (spring rest length). This is achievable because of the possibility of calculating the string length from the DC motor’s encoder signal and thanks to the presence of a force control loop embedded on the TSA module. In particular, such lower-layer controller uses the force sensor measurement within a standard Proportional-Integral-Derivative (PID) control loop in order to track the requested reference force value. Therefore, in this way, the assistive device is equivalent to a spring attached on the user’s forearm and the way it interacts with the human is determined by the value of . Based on this observation, in our control scheme the value of is determined by a sEMG control loop, which makes use of a PI controller modified according to a double threshold logic whose functioning is observable in Figure 12(b).
(a) Control scheme
(b) Double threshold logic behaviour
In detail, the controller is based on a sEMG error between the measured biceps sEMG signal and lower sEMG threshold, the latter used as reference value. The proportional action of the PI controller (identified by a properly chosen proportional gain ) continuously takes as input the sEMG error. On the other hand, the integral action (characterized by a properly chosen integral gain ) is activated only when the sEMG signal is continuously over a higher threshold for a period of 1 s, until the increasing of the stiffness of the TSA module makes the sEMG signal touch the lower threshold value. Such double threshold logic has been implemented because of two considerable advantages: (i) a more reliable regulation of the sEMG signal (which presents a substantial inherent variability), with respect to a single reference value approach, and (ii) the capability of filtering sEMG peaks that surpass the higher threshold for less than 1 s, which are usually present in case of slightly faster motions or unwanted impulsive contractions, avoiding an increasing of the integral action not oriented to compensate the applied load. Furthermore, the threshold values are determined for the subject by means of a simple calibration phase. In particular, the sEMG signal of the user with the elbow flexed at 90° and with a load of 0.5 Kg is recorded. Then, denoting the thresholds as and , they are computed aswhere is the mean value of the sEMG calibration recording and is the standard deviation computed over the same recording. The effects of this control strategy during the lifting task of the 2 Kg load by the user with the TSA module support are shown in the graphs of Figure 13. In particular it is possible to observe how the sEMG signal initially goes over the higher threshold (red), causing the activation of the sEMG controller integral action after one second. This makes the stiffness of the TSA module rapidly increase in such a way that the biceps sEMG activity decreases until it reaches the lower threshold (green). Then, for the subsequent flexions of the lifting task, the sEMG signals remain in the neighborhood of the lower threshold without surpassing the higher threshold except for peaks that are successfully filtered by the implemented double threshold logic, which means that the integral action is not activated is these cases. In other words, this means that the biceps exerts a muscular activity that never surpasses the sEMG value corresponding to the 0.5 Kg calibration recording, saving a quantity greater than or equal to 1.5 Kg-related muscle effort. On the light of this and, especially, comparing the sEMG values during the lifting of the 2 Kg load in the case with no support (Figure 11) and with support (Figure 13) the effectiveness of the TSA module as an elbow assistive device is proved.
6. Conclusions
This paper presents an innovative wearable TSA module system with integrated force sensor which can be used for rehabilitation and assistive robotic applications. The proposed TSA module structure, the integrated force sensor, and the basic sensor working principle are discussed in this paper. To validate the TSA module compliant frame, the FEA has been compared with experimental measures. Different experiments have also been carried out to verify the properties of the proposed TSA module.
Furthermore, to show the capabilities of the proposed TSA module in real assistive applications, an experiment involving a male healthy subject has been carried out, where the TSA module has been used to provide load compensation to the user’s elbow joint during lifting tasks of a 2 Kg load. The experimental results show that the device, together with a suitable control scheme, can be successfully used to help and limit the user’s biceps muscle activity during dynamic motions when a load is applied in proximity of the wrist.
Future work will be devoted to the installation of the module on the back of the user, in order to obtain a wearable elbow assistive device and to the implementation of a sEMG-driven control scheme that makes use of both biceps and triceps, exploiting antagonistic muscle activation concepts.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.