|M.Sc Student||Marjieh Fares|
|Subject||Hand Rehabilitation Robot Based on Impedance Control|
|Department||Department of Mechanical Engineering||Supervisors||ASSOCIATE PROF. Reuven Katz|
|PROF. Miriam Zacksenhouse|
|Full Thesis text - in Hebrew|
In this thesis, a forearm rehabilitation robot was built (movement restricted to pronation/supination). The robot is designed to work in two phases: (1) Free movement in which the patient has full control of the robot movement without any intervention from the robot. This phase allows the patient to attempt the movement on his/her own. (2) Controlled movement in which the robot assists the movement of the patient taking into consideration the resistance the patient is applying.
Rehabilitation robot is required to interact with its environment (patient). For a robot to interact efficiently it is not enough to control the torque it is applying to its environment or its movement. That is because movement cannot be controlled via torque control and vice versa. Therefore, the relation between torque and movement is the one that should be controlled. This relation is called Impedance.
In this thesis, position based impedance control method was used to apply impedance control in two robots. This control method requires measuring the interaction torque between the robot and the patient. Two mechanisms were used for measuring the interaction torque: (1) Electric torque sensor. (2) Series elastic actuator (SEA) mechanism.
The position controller was designed using loop shaping method with wide bandwidth and large phase margins. Prior to controller design, the system was modeled using system identification methods, and real system performance was compared to model predicted performance. Feedforward control was designed to compensate for the robot dynamics such as non-linear friction and gravity. Additionally to compensate for the environmental torque applied on robot.
Two robots were constructed using each of the two mechanism for measuring the interaction torques. A series of experiments were carried to quantify performance of each one of the robots. Experiments were divided to two categories: (1) Impedance accuracy, i.e., the ability of the controller to behave according to the desired impedance, with computer controlled interaction force. (2) Impedance accuracy with external, physical torques. This category includes two types of experiments: (2a) assisting in reaching specified angle through minimum jerk trajectory, and (2b) minimal torque required to start movement.
The experimental results show that for the constrained movement phase, performance of the robot with electric torque sensor are superior to the performance of the robot with SEA mechanism. That is because the torque electric sensor is more accurate and the positon feedback is better. The robot with SEA mechanism, however, is better for the free movement phase because it allows working with impedances with lower values.
In the future work, the rehabilitation robot will be used to characterize the behavior of the forearm in pronation/supination movements. It can also be used to estimate the forearm impedance model in pronation/supination motion when fast disturbance torque is applied.
In addition, an advanced autonomous robot will be designed to demonstrate the concept of brain-in-the-loop (BITL) for robot-assistive rehabilitation. The purpose of the method is to account for the underlying neural processing in the brain, which is critical both for motor re-training and for predicting patient response to rehabilitation.