Pneumatic fluid-powered actuation is widely used in lower body rehabilitation and assistance. The McKibben pneumatic muscle is one of the popular choices for orthotic devices and lower extremity Exosuits [28, 30]. They are lightweight, cost-efficient and have low mechanical impedance and inertia. They can provide a reasonably large force (e.g. 370 N [30]) with a small pressure (e.g. 5 bar [30]) and exhibit a rapid response time (0.01s – 1s [30]). In addition, they are normally tethered and need the auxiliary power device to operate the system.
Pneumatic rotary and foldable actuators are also welcomed by the community [27, 33-38] due to their flexibility of motion. They can operate with a large range of bending degrees (e.g. 0° to 360° for a foldable actuator [37]), which are suitable for a wide range of rehabilitation and assistance applications, including simultaneous dorsiflexor assistance and plantar flexor assistance, TKA and post-surgery rehabilitation, and knee and ankle movement assistance.
Fabric-based pneumatic Exosuits have shown their advantages of varying stiffness, portability, and comfortability and ultra-lightweight [29]. Compared with McKibben pneumatic muscle-based devices, they generally operate at lower pressures (e.g. 0.5 bar [29]), in which lower forces are expected. These can be highly suitable for small assistance tasks, but they may not be sufficient for medical rehabilitation programmes due to limited force. However, the successful concept indicates a promising research direction of the development of soft wearable suit for our daily use. These assistive devices can significantly reduce the required muscle effort and empower our capability in daily activities.
Real-time feedforward and feedback control systems (e.g. PID control [28], force feedback control [31,33], human-robot interaction control [34, 36], patient-centred control [35]) have been developed for use as lower body rehabilitation and assistive devices. However, control accuracy and reliability continue to be key challenges, which are caused by the inherent compliance of the soft actuators. In addition, uncertain and inaccurate gait feedback can lead to large perturbation of the control systems, which can lead to wearer discomfort, or even injury. To address this challenge, advanced control algorithms and optimisation strategies can be employed.
New soft pneumatic fluid-powered actuators have also been developed and applied for upper extremity rehabilitation and assistance. Wilkening et al. designed a soft elbow trainer using soft pneumatic actuators as bending joints, as shown in Figure 3a [38]. The elbow trainer consisted of a bending joint, a passive rotation joint, and a passive translation joint to realize motion in both extension and flexion directions with an implicit self-alignment to the polycentric movement of a human joint. The bending joint was constructed using three pneumatic skewed rotary elastic chambers (sREC) actuators [35] connected in series with two antagonistic rotary elastic chambers. The bellows of the actuators were pneumatically connected, which were able to achieve a bidirectional rotary motion with a range of 120°. An artificial neural network (ANN) was used to estimate the trainer position using sensing signals from four flexure sensors. The estimated position from the ANN was experimentally examined and compared with the measured results from the inertial measurements units (IMUs), showing an average of the absolute signal difference of 1.25° with a standard deviation of only 0.85°. The adaptive assistance combined with the self-alignment capability can significantly reduce the device adjustment time. The adaptive position feedforward and feedback control enables the elbow trainer to provide adaptive assistance and human-robots interaction without using the torque and stiffness information.
For hand rehabilitation applications, Li et al. [39] developed an assistive glove that was combined with a pneumatic actuator fabricated using a latex material for post-stroke patients. The device consisted of a power pack, control electronics, a curvature sensor, and a soft “ball” shape pneumatic actuator. The actuator was embedded into the surface of the glove to control the bending angle of a finger, which could be measured by embedded curvature sensors. Sensors were calibrated based on a three-layer back-propagation neural network which achieved an accuracy rate of 91%. A closed-loop proportional controller was developed, and experimental results showed the device was able to quickly track the desired position in approximately 1 s. The device can only move all fingers together and the nonlinearity and resulting hysteresis of the system needs to be considered and addressed in the design of the controller. To independently control fingers and assist with functional grasp pathologies, Yap et al. [40] developed a soft robotic glove that was driven by fibre-reinforced elastomer actuators (FEAs). The FEAs were mechanically programmed to generate the desired motion paths by tailoring their morphology. The actuation force of the FEAs was up to 13.6 N at an actuation pressure of 153 kPa and each glove finger could be independently actuated by an individual FEA to realise a full range of hand motions. Radio-frequency identification (RFID) tags were combined with surface electromyogram (sEMG) electrodes in a brace to measure the movement of the forearm muscles and an EMG-RFID control strategy was developed where the RFID tags were used as non-physical switches for tasks such as palmar grasp and pincer grasp while the sEMG signals from the forearm muscles were used to monitor the states of the hand including activate, hold and release. The results showed that the glove could achieve palmar and pincer grasps corresponding to a Kapandji Score of 9. A Kapandji Score (score 1 to 10) was used to assess the movement of the opposition of the thumb with the other fingers. The EMG-RFID controller enabled the glove to detect user intention with the minimum number of sEMG electrodes on the device. To simplify the sensing system, a forearm band with three force-sensitive resistors was developed to capture the force myography (FMG) signals from three muscle locations [41]. An artificial neural network (ANN) was used to classify four hand configurations, including finger extension, finger flexion, wrist extension and wrist flexion from the measured FMG data. The results showed that the average real-time testing accuracy could reach 94.04% with a standard deviation of 2.06%, which is feasible and promising for successful user-interaction control for hand rehabilitation. The team also developed a bidirectional pneumatically driven glove that can provide both active flection and extension [42]. The glove is fully fabric-based and includes both the flexion and extension actuator fabricated by using ultrasonic welding layers of thermoplastic polyurethane (TPU) coated fabric. The flexion actuator can achieve a bending radius of 0.069±0.003 at 30 kPa and a block tip force of 14.3 N at 70 kPa while the extension actuator can provide an extension torque of 0.3 Nm at 70°. Tests on five healthy participants showed that the glove was sufficient for 90% of daily function activities. Compared with the fabric-regulated elastomeric gloves, the fully fabric-based gloves are much lighter and can achieve a smaller bending radius, a larger range of motion, and higher forces output at lower pressures. The use of the extension actuators can effectively mitigate the slow dynamic responses due to pneumatic valve discharging and achieve bidirectional motions, although fatigue tests on the fabric-based actuators and the evaluation of efficacy on impaired patients need to be conducted. For thumb rehabilitation, Shiota et al. developed a device by integrating the FEAs into a 5-digit assist system which is assembled onto a forearm socket [43]. The FEAs were able to bend over 170° in 0.5 s and return to a resting state within 0.4 s. The average maximum torque of the FEA was 36.1 cNm. An enhanced Kapandji test was conducted, and the results showed that the gloves can reach the majority of positions that are commonly used on a daily basis.
Powered wearable assistive devices are used for helping carers during repetitive and heavy work. Lower back pain (LBP) is a common symptom, which is related to repetitive and heavy lifting and bending. Soft wearable power assistance devices manufactured from soft pneumatic actuators provide a promising solution to meet this important societal need. A powered assistive device that used two types of pneumatic actuators for the tasks of lifting and holding to prevent LBP was developed by Cho et al. [44]. The device employed an elongation-type pneumatic actuator and layer-type pneumatic actuator to provide an assistive force for the lower back, as shown in Figure 3b. The elongation-type actuator was constructed using a rubber tube with a woven bellows sleeve cover with two ties at each end. The actuator was able to extend from an initial length of 320 mm to 490 mm at a driven pressure of 500 kPa. The layer-type actuator consisted of two thermoplastic polyurethane (TPU) balloons inside nylon pockets, which was able to extend its height from an original length of 2 mm to 50mm at a driven pressure of 250 kPa. The expansion force of the layer-type actuator could reach up to 450 N at a pressure of 60 kPa. A feedback PID control system was developed using the actuator pressure and body acceleration and the acceleration was measured when the human body was inclined. The results showed that the desired pressures could be achieved, and a maximum pressure of 500 kPa can be achieved with a rise time of 2.5 s. The deflation time from the pressure of 100 kPa to 0 was 1.7s. The device was used by three subjects, and the results showed it was able to effectively reduce the maximum muscle activity by 33.1% and the muscle activity time by 2.16 s. The device is lightweight, user friendly and can be worn on the body with everyday clothing, demonstrating the benefits of a non-rigid and soft skeleton frame structure.
Another soft body suit was reported by Abe et al. that was based on combining two novel concepts (i) a muscle textile and (ii) shifts in the balancing posture of the body [45], as shown in Figure 3c. The active muscle textile was fabricated by sewing multiple soft, thin pneumatic muscles into a flexible fabric, which provided high power density, high force, good flexibility and a lightweight nature. The suit consisted of 11 muscle textiles and was divided into two parts to support shoulder flexion and horizontal flexion, elbow flexion and extension, respectively. The suit weighed 2 kg and its high force capability was able to support and disperse the operating stress at the contact points between the human body and support unit, which effectively reduced the force on the human body and improved comfort. The flexibility and lightweight nature of the textile actuators enabled complex motions, such as twisting. To evaluate its performance, the device was used by a subject and the reduction of the body burden force was evaluated by measuring the integrated electromyogram (iEMG) signals of the muscles. The results showed a burden reduction of 33% in the biceps brachialis muscle and a suppression of 5% in the body sway. The muscle textile was recently further improved using a novel '18 weave' structure muscle [46], where Figure 3d shows the ‘18 weave’ structure that is weaved using thin McKibben muscles as warps and wefts. The ‘18 weave’ structure muscle was able to retract by 26.5% of its original length. It was observed that during use of the iEMG of posterior deltoid decreased by 23%, and the myoelectric potentials of the anterior deltoid decreased by 40%.
Ang et al. presented a 3D printed Soft Robotic Wrist Sleeve (SWS) with 2-DoFs in the flexion-extension and radial-ulnar direction for wrist rehabilitation [47]. Two fold-based soft pneumatic actuators were attached to the fabric sleeve to realise 2-DoFs motion, as shown in Figure 3e. The SWS was able to achieve a 71.1% range of motion on healthy wrists and could provide sufficient torque and bending curvature at an operating pressure of 200 kPa. The SWS is easy to fabricate using 3D printing technology and can assist with wrist in both flexion-extension and radial-ulnar directions.