Adaptive control of one-DOF portable rehabilitation robot for wrist training

Abstract

Stroke is one of the leading causes of severe disability. The application of rehabilitation robots is increasing rapidly to help in recovering this disability through rehabilitation training. By using robot, the patient may perform the training more frequently. Various rehabilitation robots have been developed with a set of rehabilitation training programs with different haptic modalities. Different controllers were applied to provide accurate motor control for the rehabilitation robot and PID controller is one of the commonly used controllers. A robot named CR2- Haptic, which is used to train upper limbs, was developed in UTM with a set of rehabilitation training programs with PID controller that was designed for the patients having standard weight and wrist flexibility. The robot is successfully being used for training of stroke patients. One of the limitations for the PID controller is that it is not able to adapt its controller if the load is over its capability, since the robot controller is tuned based on a set standard weight. Therefore, the robot controller was not able to adapt itself to rotate the patient’s hand for patient with high muscle stiffness which is common in stroke patient. Thus, it limits the use of the device to only patient with low muscle spasticity. Whenever the unknown and inaccessible load torque is imposed, the system will have the steady-and/or transientstate error. Therefore, in this project, a model reference adaptive controller (MRAC) which is able to adapt itself based on different patient conditions has been designed using Lyapunov method and implemented on the CR2-Haptic device to reduce the positioning error and make it more beneficial for wide range of stroke patients. The controller has been tested on subjects of different muscles stiffness. It has performed better for accurate positioning of the end effecter for patients with different weight and muscle stiffness. The results show that the designed controller is able to cope with the variations in limb’s stiffness of the patients without the aid of any additional stiffness detection sensors.