A disabled lower-limb can be assisted with active or passive exoskeleton for rehabilitation purpose. Moreover, for proper balance of a patient while walking, the exoskeleton should repetitively provide assistance for the elevation of the centre of mass. Suitable selection of the actuator for the active assistance of exoskeleton during gait cycles is a challenging problem in exoskeleton design. The actuator selected for an exoskeleton should deliver high torque required for actuating the joints, and it should have compact and light weight drives. However, the actuators for exoskeleton require higher torques at higher speeds, and hence most of the commonly available actuators used for robotics are not suitable for exoskeleton applications [1]. For regular walking, an average torque of 35 Nm is presumed to be sufficient for the actuation of the hip joint of most patients [2]. Hydraulic cylinders, electrical motors, pneumatic cylinders (pneumatic artificial muscles (PAM)), and series elastic actuators (SEA) are suitable for actuating the joints of the exoskeleton to assist the patient during walking [2]. The most suitable actuator is selected based on the required joint angle pattern, joint forces, and power required for the lower limbs to get actuated.

An actuator is said to be efficient when its efficiency curve lies within the critical crossing point of the reference efficiency curve. As the ratio of moving time to the sum of moving time and stopping time of electric actuator crosses the crossing point, electric actuators are less efficient for its use as exoskeleton joint actuators. On the other hand, when a pneumatic cylinder has a short stroke and actuates vertically, it does not reach the crossing point easily. Hence the pneumatic actuators are highly efficient than electric actuators for actuating the exoskeleton joints. Also it is misleading information that pneumatic actuators are inefficient than electric actuators [3].

Mostly, electrical actuators are employed for torque generation in powered exoskeletons. But considering the economical aspect and weight to power ratio, pneumatic actuators are preferred for the design of robotic systems. The criteria for the selection of actuators include force and power requirements at the joints, required pattern of angle joint and the degrees of freedom of joints [4]. Many research works have utilized McKibben muscle as the torque generator for actuating the pneumatic orthosis [5]. Apart from the static and dynamic properties, McKibben muscles resemble the features of biological muscles [6]. However, in Mckibben muscles, reduction of stiffness is noticed when driving heavier load. The limitations of artificial muscles are overcome with the aid of pneumatic cylinders which are compact, rigid and occupy less space when pressurized. Ferris et al. implemented a pneumatically powered orthosis which was driven by artificial pneumatic muscles and delivered substantial plantar flexion and dorsiflexion torque at the ankle joints [7].

Powered ankle-foot orthosis (PAFO) assist the lives of several people with physical ailments by aiding the user's ankle joint during walking. A two degree-of-freedom (DOF) PAFO was developed and validated with healthy subjects and found to be kinematically stable [8]. The features of powered hip orthosis with pneumatic muscles were assessed in order to predict the paralyzed gait cycle [9]. In 2019, Doumit et al. from University of Ottawa designed a passive ankle exoskeleton (PAE1) based on braided pneumatic muscle [10]. In 2020, they reoptimized this exoskeleton PAE2, composed of an aluminum structure, a steel plantar clutch and artificial pneumatic muscles.

The pneumatic muscles contract and expand in response to the compressed air input [11]. The working concept of a pneumatic cylinder was described by Kazerooni [12] and Varseveld et al. described a basic control approach for pneumatic cylinders utilizing PWM technique [13]. Similar studies were also presented by Song and Liu [14]. This control approach is mainly concentrating on the control of actuator shaft's position. But for the real design of the orthosis, it is very essential to consider the velocity or torque control of the pneumatic cylinder. The torque control in joints is crucial as the torque varies in irregular surface conditions. Pneumatic actuators exhibit highly non-linear characteristics, due to the compressibility of air, the involvement of friction and the non-linearity of valves. Accordingly, the load of motor is non-linear and tends to vary throughout the operation. Actuator power and piston displacement can be controlled effectively to control actuator force. A control algorithm was developed to convert the pneumatic actuators into force generators for robotics control applications [12]. While pneumatic actuators are powered by an easily accessible source, they are subjected to significant friction forces, making position control challenging. It is possible to precisely position pneumatic actuators by using solenoid valves instead of huge and expensive servo valves [13], [14], [15]. In an orthosis, controlling of the solenoid valve is necessary for the transition between flexion and extension.

For controlling a pneumatically powered orthosis (Pneu-WREX) a suitable controller was designed, which moves the orthosis to the desired target with a low position tracking bandwidth [16]. The performance of pneumatic actuators can be enhanced by the usage of adaptive controllers [17], [18], [19]. A biped robot was controlled by pneumatic muscle and PID control strategy with appropriate gain constants was implemented with minimum error of −1.5° to 1.9° [20]. Zhao et al. depicted the characteristics of pneumatic actuator with a PID controller tuned using RBF neural network (RBFNN) [21]. The response of RBFNN PID controller was attained with minimum tracking error (−2% to +2%) whereas; PID had an overshoot of 20%. RBFNN has low training speed in recognition of gesture signals. So, PSO-RBFNN was proposed which enhances the feature recognition rate [22]. In many biomedical applications, PSO approach is employed for tackling nonlinear optimization problems. By tuning the PID controller using PSO, time domain specifications like rise time, settling time and peak overshoot of the system were altered. As a result, the computation efficiency of the system was improved [23].

The traditional PID control is vulnerable to muscle hysteresis errors, and it can be enhanced by adopting a cascaded model-based controller [24], [25]. However, due to the intricacy of the pneumatic muscle dynamics, conventional controllers may not be able to operate the entire robotic system [26], [27], [28]. As a result, using a sophisticated control method to govern pneumatic muscles is not the best choice. Integrated intelligent control techniques, along with conventional controllers, are necessary to attain the desired response [29].

The implementation of traditional controllers like PI and PID in an orthosis will affect the performance in tracking desired trajectory of the gait cycle in the presence of uncertainties and load disturbances. Lack of very fine tuning of gain constants could affect the response of the system and it can be improved by proper optimization techniques. In this study, a pneumatically actuated orthosis was developed with suitable controllers to mimic human gait. The gait analysis of thirty healthy subjects (men) and four afflicted subjects (polio affected men) were carried out in the previous work done by the same research group [30]. The developed pneumatic orthosis was customized for a polio affected person. While testing the performance of the controllers, the pneumatic orthosis was fitted onto thirteen volunteered healthy subjects and one polio affected person. To improve the stability of the system a stochastic approach, PSO was employed for achieving optimal controller gain constants.