Developing a Linear Quadratic Regulator for Human Lower Extremity Exoskeleton Robot

Abstract

During the last two decades, exoskeleton robot-assisted neurorehabilitation has received a lot of attention. The major reason for active research in robot-assisted rehabilitation is its ability to provide various types of physical therapy at different stages of physical and neurological recovery. The performance of the robot-assisted physical therapy is greatly influenced by the robot motion control system. Robot dynamics are nonlinear, but many linear control schemes can adequately handle the nonlinear dynamics with the help of feedback linearization techniques. In this study, the dynamic model of the human lower extremities was developed. A state-space form of the human lower extremity nonlinear dynamic model is presented. LuGre friction model was used to simulate the robot joint friction. A Linear Quadratic Regulator (LQR) was designed to control the human lower extremity dynamics. Dynamic simulations were carried out in the Matlab-Simulink environment. The designed controller's tracking performance was demonstrated in the presence of joint friction. The developed controller's tracking performance is assessed by comparing the results obtained using LQR with other linear and nonlinear controllers (PID, Computed torque control, and Sliding mode control). For performance verification, the same robot dynamics, friction model, and trajectories were used. The stability of the developed control system is also analyzed.

Radial basis function‐based exoskeleton robot controller development

“…The trajectory tracking performance of the Sliding Mode Controller [18], Computed Torque Controller [15], Adaptive Controller [16], LQR [22] and Model Reference Computed Torque Controller [17] is shown in Figure 28. All the abovementioned controllers provide very high trajectory tracking performances.…”

“…The complete form of the variables used in Table 3 and Table 4 are as follows, references [20], [18], [17], [19]. All four controllers are based on the same plant dynamics.…”

“…A robot can be controlled by linear, non-linear, robust, optimal and adaptive control systems. Most established control approaches [15][16][17][19][20][21][22] are based on model-based control strategies. For linearisation and control, different parts of the robot's dynamic equation of motion are used.…”

“…There are different robot control techniques are available. Linear [16,19], nonlinear [18], robust [20], optimal [19], adaptive control [17] schemes can be used for controlling a robot. By analyzing the established control techniques [16][17][18][19][20] Some prominent works that used neural networks for robot control applications are discussed below.…”

“…Different values of joints velocities, subject's weight and height used for input data generation 19. …”