Lately, exoskeletons are designed to provide strength in gait and heavy transport loads. There are also designs for assisting people with disorders in motion or older adults. Gait rehabilitation is one of the most significant challenges for society in the coming years due to population ageing and the increase of diseases affecting motion. Partial or total paralysis of one side of the body due to injuries in the motor centres of the brain is called Hemiplegia. Hemiplegia is a disorder that causes one-half of the human body to fail to perform its functions. This disorder is caused mainly due to stroke, and in many cases, it is hereditary. Recovery from a stroke is complex, and the treatment is prolonged. Wearable robotics is an area that provides solutions for such problems. A wearable robot extends, complements, or empowers the human limb where it is worn. These kinds of robots are classified according to the function they perform:
ARCSEC DRIVE contains other products and information you need, so please check it out.
Empowering robotic exoskeletons: These kinds of robots are known as extenders since they extend the strength of the human hand beyond its natural ability while maintaining human control of the robot.
Orthotic robots: An orthosis maps the anatomy of a limb to restore lost functions. The robotic counterpart of orthosis is robotic exoskeletons that complement the ability of the limbs. Exoskeletons are also capable of restoring handicapped functions.
Prosthetic robots: These robots are devices that fully substitute lost limbs [1].
Figure 1 shows two examples of wearable robots. The scientific community differentiates exoskeletons from orthosis by defining the former as the devices that enhance the physical capabilities of wholesome users and the latter as the devices that assist persons with limb impairments [2]. Specifically, in Figure 1, the lower extremity of an orthotic exoskeleton for mobility problems is presented as developed by the authors, and the lower extremity of a prosthetic robot, according to work in [1], is presented in Figure 1b. Despite their differences, both devices act in parallel with the limb. In the medical field, in combination with rehabilitation therapies, exoskeletons can help patients with spinal cord injuries, strokes, and lower limb paralysis caused by hemiplegia [1].
The studies of the calculation of torque equations in each lower extremity exoskeleton joint were based on the kinematic analysis. Specifically, in [3], forward kinematics was applied to find the foot's position when values were given for the corners of the joint. The torque required on each joint is determined using free-body diagrams of different joints. The work found in [4] proposed the lower limb robotic exoskeletons (LLRE) model. The free-body diagram of force on the knees and hips was constructed. Dynamic hip and knee models were obtained, considering the hips and knees as support points. The torque equations of the lower limb joints were calculated according to the parameters of the specific model. Another approach to calculating the joint torques was also based on kinematics. The kinematic analysis is applied through forward and inverse kinematics as proposed in [1]. The Euler'Lagrange method is used to obtain the dynamic equations of the exoskeleton. The literature review was performed by querying the Google Scholar database. To identify papers on robotic lower limb exoskeletons, we mainly focused on electric actuation technologies. The results were filtered based on the officially used torque calculation models to determine the percentages. Nearly 600 scientific articles have been published in the last three years on robotic exoskeletons for the lower extremities based on kinematics (thus, excluding the upper extremities cases).
Nearly 32% are surveyed on the topic, and 35% mainly present simulations of proposed models based on the formally used torque calculation models. The rest include works on reducing cost or power dissipation and applications of ML in the control of the exoskeleton. As the works above show, the researchers of robotic exoskeletons calculate the torques of the lower limb exoskeleton joints based on their model's kinematic analysis.
It was noted by the authors that there is a lack in the international literature on the measurement of lower extremity joint torques embedding the physical part of the implementation, which differentiates significantly by both the building components of the exoskeleton and the user's physical characteristics.
Thus, the following are issues that motivated this work, and the proposed approach to offer a combined solution is presented in the rest of this work.
Lack of a well-defined framework for calculating the torques of the joints based on multiple factors.
Creation of a set of parameters for simulating the operation and calculating the torques.
Inclusion of the user's physical data in the calculation of the parameters for calculating the torques.
Use of the motor characteristics to assess their suitability or not for the targeted solution (exoskeleton).
Most authors seem to agree that clinical gait analysis (CGA) data sources are a good start for the initial design of the actuation to be used in their prototypes [5]. However, Beyl, in his work [6], remarks on the large variability observed in gait data and cautions designers of actuated exoskeletons to be careful in interpreting CGA data and formulating design recommendations based on those data. Joint torque data determine the required characteristics for the actuation to be applied at each assisted joint. The intensity of the joint torques fluctuates within the gait cycle [5,6,7,8], and therefore, in most cases, designers use maximum values (peaks) as requirements for the sizing of their actuators [5,6,7]. However, in [9], the authors used optimisation methods and models of human motion to estimate the required torques for their passive, assistive systems. The aforementioned shows the need for a method to optimise the torque calculation based on the characteristics of the human, the exoskeleton, and the motors selected to decrease the time-to-production and achieve smooth motion.
According to the above literature references, the mathematical model in this article differs from other models in terms of its variability. First, the user can configure multiple parameters that affect the robotic exoskeleton, taking into account the characteristics of the human body (weight, height, etc.) and the characteristics of the robotic exoskeleton (exoskeleton weight, actuator weight, etc.). Second, the proposed solution considers the self-correction of the model by allowing its dynamic modification.
Figure 2 shows the robotic exoskeleton drawing showing the joints of the actuators. We are developing a hybrid rehabilitation system (FesRobex) combining Functional Electrical Stimulation (FES) and an exoskeleton to control patients' gait with lower limb mobility problems. This study aims to use the mathematical model to calculate the torque at each joint and analyse the individual characteristics that influence the mathematical model for the appropriate selection of motors in exoskeleton joints.
First, in Section 1.1, an introduction to servo motors is offered, and Section 2 presents the proposed approach. Then, in Section 3, the evaluation of the proposed model and the calculations are offered, as well as its effectiveness in selecting the appropriate motors for embedding on a targeted exoskeleton. Finally, the paper concludes with Section 4 featuring the discussion and Section 5 outlining the conclusions of the mathematical model.
In this section, an introduction to servo motors that are used in the design and implementation of exoskeletons is provided. This part of the work is considered significant to highlight the variety of the available servo motors and emphasise the parameters that need to be considered when embedded in an exoskeleton.
Servo motors have been used in automatic control systems for many years, especially in applications that require speed, position and torque control of the motor shaft. Classic examples include robotic arms, automatic machine tools, remote-controlled models, and automatic navigation systems for ships and aircraft [10].
The essential characteristics of the servo motor
The motor torque is proportional to the applied control voltage that the amplifier develops due to the error at its input.
The direction of the torque is determined by the control voltage's polarity (instantaneous value).
Servo motors are structurally very similar to ordinary motors but are not identical. They differ because they contain measurement devices and a feedback system that is used in conjunction with a servo-drive mechanism to control torque, speed, or position [11].
The criteria for selecting servo motors are response speed, accuracy, and errors due to external distortions combined with the cost, availability, and reliability of the motor. Another important selection criterion is that the performance should cover both the power of the load (due) and the friction (losses) of the device. In addition, the servo motor must operate at the desired speeds and provide the required acceleration for the rotor and the load [12].
(a) Mechanical and geometric
The size, weight and inertia of the motor.
The placement of the motor and the way it is connected to the moving mechanism.
(b) Electromechanical
Required power and concentration of motor power (power to mass ratio).
Torque requirements and characteristics.
Speed range and response to changes.
Sensitivity to changes in servo motor parameters [13].
DC Servo motors
There are various types of electric servo motors, such as permanent magnet servo motors, rotor-controlled servo motors, stator-controlled servo motors, and stator and rotor servo motors in series.
R/C Servo motors
R/C servo motors are special compact servo motors that include a complete servo motor system consisting of a motor, a gearbox, a feedback device, and a drive and control circuit (Figure 3). The main parts constituting it are the following:
DC electric motor;
An electronic circuit controls the end drive shaft and gearbox position.
The final drive shaft does not perform complete rotations but rotates between two extreme positions. The servo operation requires (Figure 3) the provision of the appropriate electrical voltage and a signal that determines the position of rotation of the final shaft. The control of the servo requires a specialised controller, and the open-loop control method is used. The main disadvantage of RC servo motors is the inability to perform complete and continuous rotation (Figure 4).
Nevertheless, these servomechanisms have essential advantages such as:
Low cost.
Small dimensions and easy-to-use shape, which surrounds all parts.
Produce high torque values.
The use of sensors and feedback circuits is not required to determine the position of the driveshaft.
Servo motors are essential in robotics as they facilitate intelligent and natural movement. They are used in robotic systems of all kinds and can transmit information about the rotation of the motor on its axis so that the robot can 'know' the movements of its various parts. The realisation of the desired movement of a robotic mechanism requires the combined movement of its joints. This is achieved by using a servo motor, which drives a mechanical system as a whole. In recent years, with the development of biomedical technology, servo motors have had broad applications in medicine and specifically in robotic medical systems [12].
DC motors are used in applications where DC power sources are available, such as aircraft, automobiles or robotic systems. However, this type of small motor has some drawbacks. The main disadvantage is the excessive scintillation and wears on the brushes. Small and fast DC motors are too small to use compensation winding and auxiliary poles. The reinforcement reaction and the effects L didt tend to create sparks in the transducer brushes. In addition, the high rotation speed of these motors increases the wear of the brushes, thus requiring shorter maintenance. In some applications, the maintenance required by the brushes of these DC motors may not be acceptable. A typical example is the DC motor in an artificial heart, where an incision must be made in the patient's chest for maintenance. In addition, sparks in the brushes are dangerous, as they can cause an explosion and excessive RF noise. This developed a fast, small and reliable direct current motor with low noise and long life. This is a combination of a small motor that is very similar to a permanent stepper motor and has:
A cursor position sensor; and
An electronic circuit breaker.
These motors are called brushless DC motors (S.R.) because they are powered by a direct current source without switches and brushes. They are also called Modern Permanent Magnet Motors [15].
The S.R. motors without brushes have many advantages such as:
Size and low weight.
Relatively high performance.
Long service life and reliability.
Minimal or no maintenance.
Very low RF noise level compared to S.R. motors with brushes.
High-speed capability (50,000 rpm).
High torque.
Their main disadvantage is the high purchase cost.
Finally, in the introduction, we note that most studies use clinical gait analysis (CGA) data and human motion models to estimate the required torques for their passive support systems. As a result, there is no possibility of parameter variability (mass, length of body parts, height and weight of a person, the mass of actuators, etc.). Most studies use ready-made data from libraries and data they extract from their model with specific characteristics, as we mentioned in the introduction to the bibliographies. Therefore, they do not have the option to vary the characteristics of the robotic exoskeleton or the characteristics of the human body over a wide range to calculate the joint torque. The aim of the article is to fill the gap that has been created, i.e., the computation of torque with the possibility of changing parameters of the exoskeleton and human characteristics. The purpose of this work is to model the best choice of the appropriate motor (in the physical implementation phase) for each lower limb joint of the hybrid exoskeleton.
In this section, the materials of interest are initially presented along with their parameters. Then, the theoretical approach, according to the evidence from the literature, is explored, aiming at the comparative data for servo motors. Collecting the servo motors' data is critical since introducing any servo motor to a robotic component dynamically affects the system's characteristics, especially in the case of the inclusion of heterogeneous servo motors to various joints of the exoskeleton. The collected data are organised in a database to allow the selection of the appropriate motor for the targeted joint. The following two subsections refer to the parameters of the exoskeleton that should be considered for calculating the joints' torques and the characteristics of the actuators that may be used. This concludes the extraction of the parameters and values of interest, developing the proposed mathematical model to calculate the torques of heterogeneous motors installed on an exoskeleton's joints.
After the presentation of the materials and their characteristics, the methodology to select the appropriate motor for each joint follows. Next, the mathematical model is analysed, and the steps to be followed are presented. The proposed methodology has two steps. During initialisation, the parameters of the user are set based on the user's characteristics (e.g., weight, height, etc.) and the coefficients derived from known models. In the first step (A), the masses for the exoskeleton are calculated, and the motors' characteristics are considered as a penalty on masses. In the second step (B), the joint torque is calculated and verified with the desired one. If the achieved torque is sufficient, then the motor is selected. Otherwise, the process is repeated until an appropriate configuration is found.
In many automation control applications, the main competitors of servo motors are stepper motors. Both types of motors have their advantages and disadvantages. Their differences relate mainly to their performance because they are differently designed. For example, a rotor motor's poles are much larger than the poles of a servo motor, so a rotation requires a larger current to flow through its windings. In addition, the stepper motor at high speeds degrades its torque, which is a phenomenon that can be reduced using a higher supply voltage. In contrast, the large number of poles of a stepper motor has a beneficial effect at lower speeds, thus giving it a torque advantage over a servo motor of the same size. Another difference is the way each type of motor is controlled. The open-loop method is used to control the stepper motors. This reduces the cost, as no feedback device is required (e.g., encoder for most positioning applications). However, in stepper motor systems, the excess power is converted into heat, thus generating a significant amount of heat in the motor and drive, which must be considered in various applications, especially those in the medical and healthcare field. Servo control solves this problem by supplying the motor with the current needed to move or hold the load. It can also provide maximum acceleration torque, which is often more significant than the maximum continuous torque of the motor. However, an encoder can also control a stepper motor in a complete closed-loop servo system. In terms of equipment, stepper motors are more superficial than servo motors.
Therefore, they are much easier to maintain and cost less, especially in small motor applications. When they operate within the design parameters, they do not lose their steps and do not require encoders. At the same time, when they are at rest, they remain stable, holding their position without any fluctuations, especially in dynamic loads. Specifically, a Brushless Direct Current (BLDC) generally operates better for speeds below RPM, lower acceleration values, and high retention torque. Servo motors are best in applications that require speeds above RPM and high torque at high speeds or where a high dynamic response is required. In conclusion, servo motor control systems respond better to high speeds and high torque applications involving dynamic load changes. BLDC motor control systems are less expensive than their respective servo motors and are ideal mainly for applications that require relatively low acceleration values, high holding torque, and flexibility to operate in an open or closed-loop system. For a complete picture of the differences between a Servo Motor SR (Brushed) and Step Motor (Hybrid) or BLDC motor, the table below shows the characteristics of a DC servo motor with a collector'brush system and a BLDC brushless motor (Table 1) [16].
We consider that the two motors are of the same quality and have the same rated power. There are many designs of robotic exoskeletons proposed for different purposes in the literature. Table 2 summarises some of the primary active projects in the scientific community with the types of motors they use, which have been reported in journals and conferences [1].
Exoskeletons are anthropomorphic mechanical devices worn by an operator that closely match the body's anatomy and work in coordination with the user's movements. Among the main requirements of an exoskeleton to be taken into account when designing are the following:
The design must be anthropomorphic: Current designs have an abnormal shape; another limitation of exoskeletons is the lack of direct exchange of information between the human nervous system and the wearable robotic part.
The design must be flexible: The length of the thigh, stem, and waist must be adjustable, and the variation in length and the stem is approximately 6 cm for average people, from 1.60 to 1.80 m. The length of the torso is approximately 0.246 times the height, and the length of the thigh is about 0.245 times the height.
Increase joint strength: Exoskeletons do not transfer the substantial load to the ground but augment joint torque. This consideration might be used to reduce joint pain or increase joint strength in paralysed or weak joints.
Selection of Degrees of Freedom (DoF): The exoskeleton must comply with the free movement of the joints. Table 3 shows the DoF of a lower extremity exoskeleton.
The exoskeleton robot actuator: It must have a high output-to-weight power ratio and features such as low inertia, fast response, high accuracy, etc. [1].
The joints in the lower limb of the human body are the hips, knees, and ankles. Each joint has different abilities to move or DoF, as shown in more detail in Table 3. The types of lower limb exoskeletons based on joint motions are differentiated into several types based on how the actuators drive the exoskeleton. The actuators can drive just the hips, the knees, or the ankles.
In a small number of studies, exoskeletons have multiple actuators to drive a combination of joints. These combinations of actuators are hips and knees, knees and ankles, and all three joints (hips, knees, and ankles) [17].
Most types of actuators used in robotics cannot be used in exoskeletons since this application requires high speeds during operation at higher speeds than most actuators can provide. Electric, pneumatic, hydraulic, and Series Elastic Actuators (SEA) are the leading candidates available for use as actuators in exoskeletons. The design and selection of actuators was based on the average torque and power of each subject during normal walking (not pathological) at average speed. In addition, a study of different potential candidates was evaluated. The most relevant criteria for selecting activation technology for driving human joints were specific strength (activator power ratio to actuator weight) and portability. In this respect, linear hydraulic and pneumatic actuators have a high power density but are usually massive and present internal leakage and friction problems.
They have been used in some recovery devices but still face a standard limitation on the constant spring of the tyre element that is stable. Harmonic coordination of strength and position between patient and exoskeleton is complex between different subjects. The literature shows that the use of electric motors provides a reduction in energy consumption during walking. DC motors meet the criteria of the necessary power with a compact and portable solution for portable devices. Based on this, brushless DC motors connected to a harmonic drive gearbox were selected. Torque calculation is necessary to construct robotic exoskeletons and specialised robotic devices using servo and DC brushless motors [18].
For more exoskeleton joint actuatorinformation, please contact us. We will provide professional answers.
As mentioned in this section, the proposed methodology comprises two steps. Initially, the characteristics of the user are set, and the parameters regarding the skeleton are set. Then, the coefficients are calculated based on known models, namely the Zatsiorsky BSP and the Dempster BSP (body segment parameters). In Step A, the masses of the exoskeleton's parts are calculated, considering the collateral effect of installing a specific motor at a joint. This is a novel approach, since this effect was included (or not) in the mass of the body parts. The approach allows a more precise exploration of the motor under consideration, its effect on movement, and the motor's verification (hence, selection) with the desired behaviour. In Step B, the torques are calculated using additional parameters set by the user or a group of sensors to calculate more precise data during kinematic analysis. At the end of this step, the torques are reported, and the motor is either characterised as appropriate, or the process is repeated from the start, rejecting the motor as unsuitable. Steps A and B have been implemented in the LabView for verification reasons and to facilitate researchers to implement these steps (see Figure 5, flow chart). The calculations include the extensive mathematical model and the database of DC motor characteristics. In this way, selecting the appropriate DC motor per articulation is automated. The implementation of the mathematical model for calculating lower extremity joint torques may be found on GitHub [19]. The fact guarantees the commonality of this methodology that it utilises the extended mathematical model, takes into account the characteristics of the available DC motors, takes into account their effect on the behaviour of the robotic exoskeleton and automates the selection of the appropriate DC motor at the joints' lower extremity.
The calculations are based on the proposed extended mathematical model, which has as its fundamental principle that each lower limb robotic exoskeleton is an interconnected part of the human body. According to this assumption, every variable of the human body, such as weight, and height, affects the calculation of the mathematical model. Accordingly, the variables of the robotic exoskeleton, such as its total weight and the weight of actuators in the joints, should also be considered. It is essential to point out that the mathematical model can be used by any lower limb robotic exoskeleton researcher who wants to calculate the joint torques to choose correctly between brush and brushless DC motors at each joint. It has the ability to configure the parameters of the human body and corresponding variables of the robotic exoskeleton. It can also use a database of data on electric motors DC for robotic exoskeletons.
In Section 2.4.1, the extended mathematical model is presented in detail regarding calculating the masses of the variables that affect each part's mass, as shown in Figure 6. Then, in Step B of the methodology, as depicted in Figure 7, the calculation of the length of parts and torques of each part of the lower limb to the total mass (kg), its height (cm) and the change of the angles of each joint, as it is analysed in Section 2.4.2, is performed. The methodology allows the user to interact with a database containing characteristics of brush or brushless, as analysed in Section 2.4.3.
The mathematical model in Step A calculates the masses and centres of mass (COM) of the lower limb of the human body, such as the hip centre of mass (COMhip), the knee centre of mass (COMknee), the foot centre of mass (COMfoot), and mass of the knee joint. The user can choose between two mass calculation models, the Zatsiorsky BSP and the Dempster BSP, and the gender (male or female), as seen in Figure 8, which is the Graphical User Interface (GUI) developed in LabVIEW.
Dempster's method is reflected in Table 4, which gives the coefficients affecting body parts' mass or centre mass. Cadaver data from Dempster () are applied to water displacement data obtained from 135 living subjects (35 men and 100 women) to obtain the weight, centre of gravity, and radius of gyration for the segmented extremities. Some subjects (33 in total, 15 men and 18 women) were examined to obtain the weight of the segments of the trunk using the water displacement method, and 16 of these subjects (7 men and 9 women) were examined to locate the centre of gravity of each trunk segment [20]. In , Zatsiorsky et al. determined the centre of mass for different human body segments. Each human body segment was divided according to the bony landmarks defined by Zatsiorsky. Data for this operation were collected via means of gamma-ray scanning, and the measurements were completed on 100 male and 15 female Caucasian subjects aged between 19 and 25 years old [21]. In , de Leva observed that data provided by Zatsiorsky lead to many errors in the body COM calculation of USA college athletes. The source of those errors was caused by the body segmentation method, specifically by setting the reference points at bony landmarks. To obtain more precise results, de Leva decided to change the reference points from bony landmarks to the axis of rotation of body segments. To simulate the kinematic and dynamic behaviour of the body in movement, the body should be simulated. Segmentation methods allow the body to be modelled as connected segments are reduced to their centre of mass [22]. Zatsiorsky's method is reflected in Table 5, where the coefficients affecting body parts' mass or centre mass are given.
In addition, the user sets the total body mass in kg (see Figure 9). In Figure 8, the code for calculating the masses in LabVIEW is shown by applying the above coefficients to each calculation of mass and centres of mass of lower body parts (see Figure 7). Masses are calculated according to the following general relationship:
m=tm+mex×Coef (1)tm = Total mass body
mex = Total mass robotic exoskeleton - mass actuator (mactuator)
Coef = Coefficients affecting the mass or centre of mass of body parts.
It is possible to adjust the mass of actuators at each lower limb joint (actuator mass selection at hip joint, actuator mass selection at knee joint, actuator mass selection at ankle joint) for each brushless motor or brushed motor (grams). The user in Figure 9 defines them according to the motor selection present in the database (see Figure 10).
Therefore, according to the above, the mass of the hip, the mass of the knee, the mass of the leg, the mass of the knee joint, the mass of the actuator at the hip joint, the mass of the actuator at the knee joint, and the mass of the actuator at the ankle joint are considered. They will then be used in the calculation of the torques that are analysed in the following section.
Figure 6 shows the Power Balance Chart of the lower leg, which shows the forces acting on the hip, knee, and ankle joints, that help determine the torque required on each joint. Using the robotic exoskeleton, the authors have designed and considered similar exoskeleton robots, such as robotic exoskeleton (LLRE) [4], and the following torque equations were calculated.
Hip torque calculation model equation:
Knee torque calculation model equation:
Ankle torque calculation model equation:
Parameters of calculations:
m1 = centre of mass of the hips;
m2 = centre of mass of the knees,
m3 = centre of mass of the foot;
m4 = mass of the knee joint;
mhip = mass of actuator at hip joint;
mknee = mass of actuator at knee joint;
mAnkle = mass of actuator at ankle joint;
g = 9.81m/s2 gravitational acceleration;
l1 = length of thigh;
l2 = length of shank;
l3 = length of foot;
T1 = torque required in the hip joint;
T2 = torque required in the knee joint;
T3 = torque required in the ankle joint.
In Step B of the mathematical model, the torques at each joint of the lower limb are calculated, as shown in Figure 7, which is the GUI. In this step, the lengths of the lower limb body segments are calculated, selecting the person's height. The user adjusts the angles (θhip, θknee, θankle) of the lower limb joints or with motion sensors. Since calculating all the above quantities, the mathematical model calculates the torques in each joint using the equations of motion mentioned above, which are integrated into the proposed mathematical model (see Figure 8).
The proposed methodology considers the characteristics of brushed and brushless motors from the database storing the motor characteristics. Figure 10 depicts the GUI of LabVIEW, in which the elements of the motors are entered in the panels brushless and brushed. The GUI allows collecting torque data (torque hip, torque knee, and torque ankle) and deleting and plotting the torques from the mathematical model, as shown in Figure 11. According to the data collected in the database, a researcher can choose the appropriate actuator needed in each joint of the robotic exoskeleton he will implement. It can also change every element in the mathematical model, as already mentioned in the flow chart of the model in Figure 5, and feedback of the robotic exoskeleton.
Below, we mention methods of selecting electric actuators and calculating joint torques. Calanca et al. in [25], presented a methodology based on a graphical tool that matches the actuator's capabilities with the task's requirements. The proposed approach obtains the operating torques and speeds through experimental tests. A motion capture system allows positions and velocities to be acquired, while joint torques are calculated via inverse dynamics in a multi-body human exoskeleton model. Similarly, Barjuei et al. in [26], proposed an approach for selecting a brushless BLDC motor and a gearbox transmission based on optimisation through an analytical human'robot dynamic interaction model and a mathematical relationship between the weight and technical characteristics of its components. Finally, Belogusev and Egorov in [27] proposed an automatic measurement procedure for determining the starting torque of an electric gear actuator for an exoskeleton. In this work, a methodology is proposed for selecting the appropriate motor during the design phase, hence, at a higher abstraction level, avoiding experimental tests of the exoskeleton. Additionally, this work considered not only the characteristics of the human'robot interaction model but also the effect of candidate motors at each joint.
In the results presented in Table 7, the maximum torques of brush and brush motors in mNm were obtained from the database, as shown in Figure 12. Furthermore, the graph in Figure 13 calculates torques at each joint at the lower end over time, according to the extracted parameters resulting from the mathematical model. Therefore, according to the previous data, a thigh motor is suitable if it offers a maximum torque of at least 150 Nm. Similarly, a knee motor is suitable if it offers a maximum torque of at least 35 Nm, and an ankle motor is suitable if it offers a maximum torque of at least 10 Nm.
Since the maximum torque range, as shown in Figure 13 (torques at each joint per time from the mathematical model), does not achieve the target torque, a gearbox (Harmonic Drive, CSD-20-160-2AGR [28]) is connected to the motor shaft to reduce speed and increase torque. So, at the peak torque from Figure 12, the mNm (1.56 Nm) brushless motors will be closer to the desired thigh torque values. A ratio of 160:1 gives each combination a continuous net torque of 71 Nm and peak torques of 180 Nm. Therefore, the average hip actuator torque of 71 Nm is considered sufficient for most patients.
The choice of brushless motors is an option because of the coverage of the maximum torque and other factors such as their weight. According to Table 7, there is less weight ( g) in brushless motors compared to brushed motors ( g). Conversely, choosing heavier motors will increase the overall weight of the robotic chassis. The weight of the robotic exoskeleton is critical to the system's human factor stability, which highlights this work's impact.
In Figure 14 and Table 8 (Total torque), the calculation of the joints of the lower extremities in the male and women groups is shown. The torques at the knee and ankle joints have negligible differences in their maximum value between males and females, Tknee = 34 Nm and Tankle = 6 Nm. Thus, the motor required for the knee joint should perform a maximum torque Tknee = (34 + std. Deviation = 10N m) = 44 Nm. Likewise, the motor required for the ankle joint should perform a maximum torque Tankle = (6 Nm + std. Deviation = 2 Nm) = 8 Nm. According to Table 7, the torques of the joints in nominal torque (max. Continuous torque) = mNm or Nm.
The above calculations of the torques at the joints of the lower limbs of the human body and the calculation of the masses of the lower limbs and the masses of the actuator joint of the robotic exoskeleton, combined with the qualitative characteristics, were compared between brushless and brushed motors, as reported in Section 2.1. The following motor choices result in each joint of the robotic lower limb exoskeleton.
Initially, a choice of a motor (EC 90 flat Ø90 mm, brushless, 600 W [24]) for the thigh and knee was made at 15,600 mNm with the addition of a gearbox (Harmonic Drive [28], CSD-20-160-2AGR) connected to the motor shaft to reduce speed and increase torque. Regarding the ankle, the motor choice was based on a brushless one (EC 60 flat Ø60 mm, brushless, 100 Watt [29]) with the addition of a gearbox (Harmonic Drive [28], CSD-20-160-2AGR), which is connected to the motor shaft to reduce the speed and increase the torque, i.e., Tankle = (nominal torque, max. continuous torque = 227 mNm); (transmission ratio 160: 1) = 36Nm. Regarding the choice of the specific motors, their masses have been taken into account ( gr, brushless < gr, brushed) and nominal speed rpm (EC90), rpm (EC60) due to the use of a gear unit.
As shown in Fig. 1, 85% of the reviewed articles (corresponding to 44 exoskeletons) used compliant actuators and a rigid structure. Soft exoskeletons represent 11% of the reviewed articles (6 exoskeletons). Two exoskeletons (4%) belong to the intersection of previous groups, this is, exoskeletons integrating both soft structure and compliant actuationFootnote 5. We refer to the latter as 'fully compliant exoskeletons'.
In this group we found three types of actuations systems: Series Elastic Actuators (SEAs), Variable Stiffness Actuators (VSAs), and pneumatic actuators. As shown in Fig. 2-a, 31 exoskeletons use SEAs, which makes this actuation the most popular choice. SEAs are characterized by having an elastic element with fixed stiffness placed in series with the motor or the motor train, and before the actuator load [10, 11]. The use of SEAs has shown improved performance in terms of human-robot interaction, safety, energy efficiency, shock tolerance and backdrivabilityFootnote 6, when compared to stiff actuators [8, 12,13,14,15]. The deformation of the elastic component can also be used to measure the joint torque, thus reducing the need of force sensors [16]. In addition, in spite of their reduced bandwidth [17], they demonstrated better torque tracking during walking in exoskeleton experiments [18].
Variable Stiffness Actuators (VSAs) are implemented in eight exoskeletons. These actuators are a variation of SEAs, in which the degree of compliance can be mechanically modulated to change the actuator's output characteristics (e.g. output stiffness) [19]. These actuators have the theoretical ability to reproduce the human-like joint stiffness profiles, adapt to environmental changes, and reduce energy expenditure [20, 21].
Figure 2-A shows a classification of these actuation solutions based on the type of elastic element. Most actuators (23) use linear springs, due to commercial availability, ease of implementation and low cost. In spite of their very approximate linear deformation characteristic, these springs present hysteresis [22], which should be compensated to reach fine control [23]. Dos Santos et al. [24] suggested that connecting the load of the actuator in a direct-drive configuration can reduce the hysteresis and residual deflection. Torsional springs are implemented in seven exoskeletons. Five exoskeletons use springs based on a monolithic disc-shaped design. These springs are compact, lightweight, able to withstand high torques with low intrinsic stiffness and are usually custom-developed [25]. There is a wide variety of manufacturing materials, such as maraging steel (martensitic steel with aging treatment) [26] or high-grade titanium [25]. The geometry of monolithic disc-shaped springs is usually defined through an iterative Finite Elements Analysis (FEA) simulation-process [27]. This process has to be carried out carefully to make sure that the spring is able to withstand the expected deformations [16]. However, results from simulations often do not match experimental results, for instance with respect to the actual stiffness [16, 26]. Bowden cables, in combination with linear springs, are used in eight of the reviewed works. These cables allow the motor to be placed away from the actuated joint [28]. The main drawback of this solution is friction, which can be managed through control [28]. Spiral springs are used in one device [29].
Figure 2-B shows the relationship between the peak torque and the weight of the actuators, classified by type of actuator. We observed that this ratio is not proportional. For instance, the actuator presented by Beil et al. in [30] delivers 120'Nm and weights 1.38'kg while the actuator presented by Wang et al. in [25] delivers less torque (100'Nm) and weights 2.90'kg. The weight values of the actuators include data related to the mechanical components (i.e. motors, springs, pneumatic muscles, transmissions, etc.) without considering power supplies or wiring.
The selection of the spring stiffness is critical when designing a SEA or a VSA. Multiple selection criteria have been used [18]. The most common criterion is to set the spring stiffness as the slope of the desired torque-angle profile [25]. Another common principle is based on maximising the energy stored and released throughout the gait cycle [31, 32]. Stiffness has also high implications on control. High stiffness increases impedance, whereas low stiffness decreases bandwidth [10]. In VSAs, stiffness can be changed manually, e.g. through a screw [33,34,35], or with motors [36,37,38,39,40], through either pretension of the elastic element or a lever arm mechanism with a variable position pivot.
Figure 3 compares the spring stiffness values with the resulting actuation bandwidth. The lack of information related to the actuator bandwidth is apparent. The actuators of the KNEXO exoskeleton and the exoskeleton developed by Yu et al. [41] present the highest bandwidth among all compliant exoskeletons [41, 42], as shown in Fig. 3.
Compliant or spring-like behaviour can also be achieved by pneumatic actuators, which depend on input air flow rate to contract and/or expand [19]. Eight of the reviewed exoskeletons use this actuation solution. In contrast to SEAs, pneumatic actuators generate forces through compressed air [43]. These actuators, also known as Pneumatic Artificial Muscles (PAMs), stand out because of their low weight (without considering the compressor for air supply), backdrivability, cost and the high specific force and power they can exert [44, 45]. Among them, the McKibben-type pneumatic muscles [45] are implemented in five of the reviewed exoskeletons (see Fig. 2-A). These actuators consist of an expandable rubber tube surrounded by a textile mesh for tension transmission. Antagonistic configuration of pneumatic muscles can be also used to obtain bidirectional rotational actuation [42]. This configuration also allows to change the effective compliance that is applied to the joint [43]. The main limitations of PAMs are the high hysteresis [46] and nonlinear force-contraction characteristics [43]. This results in complex mechanical design and control [47], particularly when large ranges of motion and high torques are required. Special controllers (i.e. torque controllers) have been proposed to deal with these non-linearities [42, 43]. A particular type of PAMs, i.e. the Pleated Pneumatic Muscles (PPAMs), showed reduced hysteresis [42, 46].
Exoskeletons with compliant actuators normally have rigid structures, composed of mechanical links and transmission mechanisms placed in parallel with the user's limbs. This rigid configuration hinders a full kinematic compatibility with human joints [48, 49]. In order to cope with this issue, Cempini et al. [50] proposed an analytical method based on a kineostatic analysis of the coupled mechanism of the robot and the human. Other exoskeletons use mechanisms to self-align the axes and reduce the time to dress the exoskeleton on. For example, Celebi et al. [51], implemented a Schmidt coupling actuated by a SEA in order to improve ergonomics and comfort. The AssistOn-Ankle exoskeleton [52] includes a self-aligning parallel mechanism, whereas the exoskeleton presented by Giovacchini et al. [53] has slots in order to modify the structure length and guarantee the alignment. Also, Saccares et al. [54] included five passive Degrees of Freedom (DoF) in a knee exoskeleton to automatically adapt itself to different users. The solution proposed by Junius et al. [55] improved joint alignment by using 3 passive DoF with sliders and hinges.
To adapt to the user-specific morphology, the foot length, pelvis width and inter-joint distance are the primary design parameters. Moltedo et al. [56] proposed a footplate that can be manually adjusted to match different foot sizes and align the ankle joint. Giovacchini et al. [53] presented an exoskeleton whose pelvis structure can be modified in width. Telescopic structures and sliders are most commonly used to adapt to a wide range of user's height [25, 57,58,59].
Figure 4 shows the exoskeleton weight as a function of the maximum accepted user's weight. When available, we also included the height range [25, 34, 40, 44, 60,61,62]. The average maximum adult wearer weight considered in the reviewed exoskeletons is 100'kg [25], whereas the maximum adult wearer height is 190'cm [34, 40]. Bilateral exoskeletons are heavier than unilateral ones, presenting an average weight of 18.56'kg and 2.52'kg respectively. The weight of bilateral exoskeletons ranges from 4.2'kg [53] to 35'kg [40], whereas unilateral exoskeletons range from 0.87'kg [63] to 4.5'kg [42] in weight. Exoskeleton with series-elastic actuation are heavier than those with pneumatic actuation (see Fig. 4) [25, 26, 40, 64, 65]. It is worth mentioning that the pneumatic actuation usually depends on off-board pressure supplies with a negative impact on portability while favouring lighter structures [42, 59, 66, 67]. Additional details are available in the Additional file 1.
The braces, cuffs, straps and orthopaedic components used in the reviewed exoskeletons are based on a broad variety of materials and configurations. The majority (30) of exoskeletons with compliant actuators have only one brace per leg segment (i.e. thigh or shank). Five exoskeletons have two braces per segment. Five other exoskeletons present a combined solution, i.e. two braces for thigh and one for shank. Some solutions are based on orthopaedic commercial braces in order to reduce costs [30, 34, 40], but most of them adopted custom-made designs. Rossi et al. [68] present a customized brace made with a 3D printer from a model obtained from a 3D scanner. Moltedo et al. [56] and Sawicki et al. [69] use braces made of carbon fiber. This material ensures power transfer in the sagittal plane of motion while allowing for passive motion in the other two planes. The shape of the braces influences comfort, which is an essential requirement for ergonomics, whereas the fastening mechanism affects the time of dressingFootnote 8 on and offFootnote 9. This process can be simplified with similar solutions such as the double-tier beleaguered structure design of the exoskeleton developed by Zhang et al. [32]. In the solution presented by Neuhaus et al. [70], leg braces were designed with an opening of approximately 180 degrees in order to improve the ease of dressing on and off. Rigid parts of these braces are designed to be attached to the leg where soft tissue's deformation is minimal (e.g. calf) in order to optimize force transmission between exoskeleton and user's limbs. The thigh and shank cuffs of the exoskeleton developed by Costa et al. [67] are moulded structures tailored to the user. Position and orientation of the thermoplastic shells of KNEXO exoskeleton are adjustable with a slider mechanisms [42]. Padding covering braces with adaptable positions to user's preference is sometimes used in order to improve comfort [30].
Exoskeletons structures composed by non-rigid components such as textiles take advantage of compliance for providing compatibility with human kinematics and dynamics, with potential to improve comfort, safety, efficiency and functionality [8, 71,72,73]. Exoskeletons with soft structures are commonly known as soft exoskeletons [74]. We found six robots belonging to this category (see Fig. 1), which represent the 11% of the exoskeletons here reviewed.
The actuation mechanism of soft exoskeletons is usually composed of motors with different gearbox mechanisms (i.e. pulleys [75, 76], gear [72]), which deliver certain torques to the user's joints though flexible transmissions and textiles worn by the user [77]. Actuators can be placed off-board [72, 75], at the waist level or in a backpack [71, 74, 78], (see Fig. 5-A). The most common flexible transmission used is Bowden cable. Other types of transmission have been proposed, e.g. inextensible cords [76]. The main disadvantages of using cables are slacks and hysteresis [78]. Both can be minimized through appropriate control strategies [79]. A recent approach uses linear compression springs in series with Bowden cables in order to decrease the overall hysteresis [78]. The XoSoft exoskeleton [80] proposed a new type of actuation principle based on custom-made soft clutches.
The structure of soft exoskeletons is mainly composed of textiles made of neoprene and/or others flexible materials [74]. Velcro-covered tabs have been proposed to improve adaptation of the textiles to the user [78]. As these exoskeletons transmit torques through biological joints by applying tensile forces, they do not constrain wearer's joints. This minimizes undesirable interferences with gait biomechanics, overcoming in this way the problem of misalignments [76, 81].
We present in Fig. 5-B the relationship between soft exoskeleton weights and delivered torques. Weights of exoskeletons with off-board actuators do not include the actuators weight. Among the on-board solutions, the exoskeleton designed by Mooney et al. [76] delivers 35.6'Nm during ankle plantar flexion with a total weight 8.96'kg, considering power supply and actuators weight. Within the off-board applications, the exoskeleton delivered by Quinlivan et al. [75] has a weight of 0.89'kg and delivers the highest torque value (48.35'Nm).
The average weight of unilateral and bilateral exoskeletons is 4.67'kg and 4.37'kg respectively. The maximum weight is 9.12'kg for bilateral exoskeletons [44] and 8.96'kg for unilateral exoskeletons [76]. The minimum weight is 0.95'kg [73] and 0.86'kg [72] for unilateral and bilateral devices respectively.
The weight of soft exoskeletons with off-board actuators fluctuates within a narrow range, i.e. 0.86'0.89'kg [63, 75], whereas solutions with on-board actuators spans from 0.95'kg [73] to 9.12'kg [44]. The maximum accepted user's weight and height is not reported in the majority of the publications reviewed.
In soft exoskeletons, the attachment components, such as braces and straps, are part of the textiles and compliant structure. Four soft exoskeletons have one brace per segment and two exoskeletons have two. The XoSoft exoskeleton [74] includes a custom garment designed to fit the user.
Two out of the 52 reviewed exoskeletons have both compliant actuation system and structure. The exoskeleton developed by Park et al. [73] integrates four McKibben artificial muscles on the ankle joint and uses textiles and carbon fiber to reinforce foot and shank structures. It can apply up to 110'Nm for ankle dorsiflexion and has a weight of 0.95'kg. The exoskeleton presented by Wehner et al. [44] uses 8 pneumatic actuators and a soft structure with multiple textile straps. Its soft structure was designed considering the virtual anchor concept. They designed the distribution and location of the textiles to maximise efficiency and comfort.
Exoskeletons for rehabilitation or assistance are characterized by a large variety of number and typology of DoF (see Fig. 6-a). Active DoFFootnote 10 are usually needed to substitute or compensate the joint torques necessary for body transport. Passive DoFFootnote 11 may also be included to cope with other biomechanical functions, such as shock absorption or weight bearing [40, 82].
In patients affected by Spinal Cord Injury (SCI), the type of support depends on the level of the lesion and its severity [83]. When the lesion is complete, an exoskeleton must substitute the entire lower limb motor function and support the whole body weight. Lower limb exoskeletons for SCI patients are usually bilateral and have two or more DoF per leg (see e.g. [22, 25, 36, 40, 64, 66, 67] in Fig. 6). Exoskeletons for post-stroke individuals should compensate for the incorrect/insufficient lower limb motion. Therefore, actuation can be unilateral or bilateral. Most of bilateral exoskeletons for post-stroke individuals include more than 2 DoF per leg [34, 36, 64, 66, 67, 84], whereas unilateral devices include only one or two DoF [41, 51, 52, 71, 73, 74, 85]. While most of the reviewed exoskeletons focus on stroke, SCI, or older adults, other solutions address other neurological or non'neurological pathologies, such as cerebral palsy [44, 68, 73], multiple sclerosis [34, 42, 73], spinal muscular atrophy [86] or are used for strength augmentation [30, 39, 54, 59, 64, 76, 78].
Gait disorders and lower limb impairments are also related to ageing [87]. In this scenario, lower limb exoskeletons are used to compensate or augment motor function, and tend to be bilateral [34, 44, 53, 58]. The majority of exoskeletons were designed for adults, whereas only three devices were specifically designed for children [38, 60, 68].
Solutions based on SEAs and VSAs are highly heterogeneous, resulting in a strong non-linear relationship between maximum torque and weight. This, together with the scarce of technical data reported, make it difficult to identify one best design option. Most of the actuators are the result of a trial-and-error design process, based on past experience and the specific needs of the application. The selection of the elastic component type and spring stiffness remains a major open challenge for SEA or VSA designs. From the control point of view, higher stiffness is preferred in order to increase the bandwidth of the system [15]. However, this can hinder the intrinsic adaptability offered by such systems. The choice of the appropriate stiffness has implications on safety. In rigid actuators, including a compliant element could have the advantage of improving the safety, e.g. in unexpected impacts between the user and the device [27]. Nevertheless, this does not always hold true as the energy stored in a spring can be suddenly released during impacts or misuse of the device, generating unexpected and unsafe reactions [88]. The selection of the optimal spring stiffness should involve a theoretical analysis and experimental validation for the specific application [16, 25, 33]. The compliant/rigid dichotomy has to be considered at the start of the project and the actuation type should be carefully selected when designing a compliant exoskeleton. For instance, when testing actuation bandwidth, well-defined and standardised experiments (e.g. with fixed loads) should be performed in order to contrast with other actuators' results.
VSAs have been proposed as a solution to make robots more adaptable to user's residual abilities and biomechanical restrictions [19]. Nevertheless, this actuation concept requires the inclusion of extra mechanisms or motors for on-line stiffness modulation, thus resulting in considerable increases in weight and complex designs. The benefits of VSAs are still to be shown in view of these disadvantages.
We found several compliant exoskeletons with pneumatic actuators in the literature. Their dependence on off-board air supply restricts the ambulatory applications of these exoskeletons. Most of the publications did not include sufficient technical details with respect to, for instance, air flow rate, pressure level for contraction and expansion and other technical characteristics of the artificial muscle (i.e. diameter, length). This information, in our opinion, would have been beneficial to compare the different solutions and therefore help the convergence on successful design strategies.
Modularity, defined as the application of the same actuator to different active DoF, is a common practice in exoskeleton design as a strategy to reduce costs as well as effort in manufacturing, development and tuning [25]. While simplifying the mechanical design, this approach often results in oversized actuators. A promising approach to improve the power-to-weight ratio is to apply modularity only on the actuation principle, while optimizing the mechanical design of the actuator to the specific torque requirements for a given joint [22, 36].
The structure's total weight and weight distribution have considerable impact on the functional performance and metabolic consumption [89]. Simulation-based optimization demonstrated to be a practical tool to reach lower weights while maintaining high mechanical properties [90]. Misalignments are more likely to happen with rigid structures, with negative effects on functionality, comfort and user's safety [91]. Different solutions to solve these problems have been reviewed in a number of previous works [50, 51, 53, 54, 56]. The introduction of multiple passive DoF is still the most effective option. Nevertheless, this solution adds extra weight to the exoskeleton and tends to complicate the structure and its control due to increased inertia and friction [48, 92]. Some successful exoskeletons with rigid structure improve user-exoskeleton interaction by reducing metabolic cost and not considering extra passive DoF addition [69, 85]. Further research in this line is needed to find optimal solutions.
Inappropriate physical contact between the user and the robot is an issue potentially affecting pain and discomfort [93, 94], and inefficient or inappropriate, e.g. delayed, transmission of forces [95]. Attachment design has to consider the inherent non-linear viscoelastic properties of human soft tissues, such as tendons, ligaments and skin [96, 97]. Compliant actuation adds complexity to these interaction dynamics [33]. In this regard, models for predicting interaction forces are a promising approach [98]. Humidity and temperature changes occurring in the surfaces of the skin in contact with the interface lead to risk of skin damage [49], and should be carefully evaluated, in particular when a prolonged use of the device is envisioned. The anatomical fit of the robot is another challenge. The user-specific approach used in the upper limb exoskeleton developed by Chiri et al. [99] is a good example on how to personalize the interface.
Soft exoskeletons present three main advantages with respect to compliant exoskeletons with rigid structures. First, the cable transmission allows optimizing the number and location of the actuators, with direct effects on the weight and inertia of the device. Second, the soft structure strongly reduce the misalignments and kinematic incompatibility between user and device [81]. Third, the soft structures are usually thin and suitable to be worn under user's clothes [71], which is appropriate for usability. However, such design solutions entail inevitable drawbacks. Cable transmission requires actuators to be placed either off-board, preventing ambulatory use, or on-board, compelling the user to carry a backpack. Cables and textiles routed between the actuator and the targeted joint generate undesirable loads to the joints along the path [100]. The non-linear viscoelastic properties of soft element result in control bandwidth reduction and inefficient power transmission [78]. The absence of rigid structure through soft element is usually associated to shear forces and soft tissues deformation, which contribute to reduce user comfort [101]. Additionally, the inability to passively support the user weight limits the use of soft exoskeleton in patients with minimal residual motor abilities. Some promising approaches to solve these problems rely on modelling the interaction dynamics between soft structures and user to improve control [101], increasing the stiffness of the textiles to maximize power transfer [78], and designing compatible sensors able to accurately and reliably measure joint kinematics and kinetics of soft structures [100, 102]. From an ergonomics standpoint, the soft structures of these exoskeletons have to be preloaded against the user body to limit undesirable motions [78]. As a result, the structure has to be tightly dimensioned on user height and morphology.
Only two out of the 52 reviewed exoskeletons included a combination of soft structure and compliant actuation. This choice is promising, and likely to lead to lighter devices with higher torque actuation performance with respect to current solutions (Fig. 4-B) [44, 73]. Nevertheless, due to the very few works identified in the literature, more research is needed to define the actual benefits and drawbacks of this option.
Choosing the right actuation type, the structure, and the physical interfaces and number of DoFs is a hard problem, which strongly depends on the application. Compliant exoskeletons with SEAs or VSAs are the most popular choice for ambulatory solutions for activities of daily living. This justifies the huge number of exoskeletons with this actuation type and the broad range of different existing designs. Pneumatic and soft exoskeletons with off-board actuators are preferred for non-ambulatory gait rehabilitation and assistance applications in the clinical setting. Soft structures are usually preferred in exoskeletons for gait restoration for individuals who still retain some walking ability (Fig. 6) [71]. Conversely, in people with strong body weight support needs, exoskeletons with rigid structures are the preferred option [25, 40].
A challenging issue is the ergonomic adaptation of exoskeletons to a wide range of patients within a clinical environment. In this case, designers have to use up-to-date anthropometric databases when defining the dimensions of the structures and interfaces. Scanning patient's limbs and obtaining the brace with additive manufacturing techniques is a promising, and low cost, solution [103]. A critical issue with 3D printing technology is the durability of the materials [104]. Thus, the study of materials and manufacturing techniques represent, in our opinion, a promising direction in order to get long-life and comfortable interfaces. However, how to address ergonomics and comfort is an unclear issue that has to be seriously taken into account in the design of exoskeletons. Under the DoF perspective, we found a strong variability across the reviewed works. This choice depends on several factors, mostly driven by the specific user needs, such as the level of reduced mobility of the user or the functional purpose of the exoskeleton, e.g. rehabilitation or assistance. Nevertheless, we could not find a clear relationship between these factors. Further research is needed to understand this issue.
A particularly relevant problem related to the use of exoskeletons in rehabilitation or assistance is their effect on balance. Users' balance may be compromised when using ambulatory unilateral or bilateral exoskeletons, due to the weight of the device and its behaviour. In addition, the loss of walking functions in most patients is frequently associated with balance disorders [105]. As a result, ambulatory exoskeletons are normally used in combination with crutches [25, 40, 70, 82]. The use of crutches may improve user's self-confidence, serve as a feedback tool, and reduce the risks of falls [82]. In clinical/research settings, non-ambulatory exoskeleton are usually supported with treadmill-based structures, standing structures or safety harness [26, 36, 38, 42, 59, 60, 65,66,67, 71, 84, 85]. Apart from the inclusion of these safety devices, balance is a topic that has been largely overlooked in the exoskeleton literature, and should be seriously considered in the future, both from the assessment and control point of view.
The studies carried out to date have seldom included the user's subjective perception of the exoskeleton in their evaluations. Including the users' opinion can shed light to the design process, in particular the design of the structure and the interface [74]. Satisfaction scales, such as Borg and QUEST scales, can be used to this end [106]. These scales have proven to have high reliability and validity [107]. Schiele studied subjective performance metrics and used a NASA TLX questionnaire to evaluate comfort in subjects using an upper limb exoskeleton [108].
Are you interested in learning more about Robot Joint Actuator? Contact us today to secure an expert consultation!