Scientists Develop Improved Algorithm for Upper Arm Rehabilitation Robots

The proposed algorithm provides accurate solutions that avoid unnatural arm postures for robotic rehabilitation in stroke survivors

Robot-assisted exercise therapy is effective for stroke rehabilitation. But the design of rehabilitation robots is complicated by a peculiar problem related to “inverse kinematics,” where the angles of the robot’s joints have to be back-calculated from their desired final positions; often, the calculated angles are unnatural for the human body. Now, scientists from Japan have developed a faster, less resource-intensive method to provide naturally feasible solutions to inverse kinematics in upper arm rehabilitation robots.

Stroke is a major concern in countries with aging populations like Japan, where the risk of stroke can be as high as 20% in middle-aged people. It is one of the leading causes of motor function loss and disability. Appropriate rehabilitation can help restore function to the affected areas, but it can be expensive, often requiring several trained therapists for a single patient. Robotic rehabilitation offers a solution to this issue.
 
An important concern in the design of robots is the ‘inverse kinematics’ problem. A rehabilitation robot mimics the human limb and has the same number of joints. When it comes to upper limb rehabilitation robots, that includes the shoulder joint, the elbow, and the wrist. Extending these joints to a particular angle makes the arm move a certain way. But imagine the possible number of angles that the joints could bend in just to do something simple like picking up an object. This is where inverse kinematics comes in. It looks at where the arm needs to go (its end position and direction) and then calculates backwards to determine what angles the various joints would need to be bent at to achieve this. It seems like an obvious solution, but the mathematics behind it is incredibly complicated. In an article published in Artificial Intelligence Review¸ a group of scientists led by Dr. Tam Bui of Shibaura Institute of Technology, Japan,  propose an improved algorithm for solving the inverse kinematics problem. “The advantage of the proposed algorithm is the reduced complexity and volume of calculations in comparison with other methods,” says Dr. Bui.
 
In recent years, optimization-based approaches to solving the inverse kinematics problem have gained popularity. Optimization deals with the minimization or maximization of an objective function (such as execution time, or energy of execution). But most current approaches to solving the inverse kinematics problem do not look at the feasibility of the optimized joint angles, i.e., if the shoulder, elbow, and wrist of the human arm on which the robot is going to be used can move in the same way that the robot arm is moving based on the back calculations. Dr. Bui and team accounted for this variable when they developed their algorithm, titled ‘self-adaptive control parameters in Differential Evolution with search space improvement (Pro-ISADE).’ Like the name suggests, this uses an optimization approach called Differential Evolution (where the calculation is iterated until the best or ‘optimal’ solution is found) with a reduced search space, which is the range of angles the joints of the robot can bend in. Reducing the search space improves the calculation speed of the algorithm and ensures that the calculated joint angles for the robot are not unnatural.
 
Dr. Bui and his team—Dr. Trung Nguyen and Dr. Ha Pham from HUST, Hanoi, Vietnam—used their Pro-ISADE approach to solve the inverse kinematic problem for the human arm in two activities essential to daily life, drinking a glass of water and brushing one’s teeth. They also studied throwing and catching a ball. First, they performed the activities using a functional human arm and captured the angles of the arm using a measurement device they had created, called an Exoskeleton type human motion capture system (E-HMCS). The E-HMCS is strapped onto the arm and converts its motion into an electronic signal using sensors called potentiometers.
 
The researchers used the E-HMCS to record the path taken by the arm to perform the tasks, as well as the arm’s final position and direction. They then put these recorded values into the Pro-ISADE algorithm to see if it could accurately predict the angles the robotic joints had to make to accomplish the tasks. They found that the Pro-ISADE algorithm predicted values that were very close to the measured angles, proving its effectiveness at solving the inverse kinematics problem for natural human movements.
 
The Pro-ISADE algorithm is, thus, a great tool for rehabilitation robots, where mimicking the natural movement of the human body is essential to avoid injury to patients. According to Dr. Bui, “Robot-assisted rehabilitation allows for higher intensity training, longer duration, and more repetition. Upper limb rehabilitation robots could help post-stroke patients quickly reintegrate into their daily lives.”
 
The Pro-ISADE algorithm could also be used in industrial and server robots, showing the wide applicability and enormous potential of this approach.

Reference

Title of original paper:	Using proposed optimization algorithm for solving inverse kinematics of human upper limb applying in rehabilitation robotic
Journal	Artificial Intelligence Review
DOI:	10.1007/s10462-021-10041-z

Funding Information

This study was also supported by the Centennial SIT Action for the 100th anniversary of Shibaura Institute of Technology entering the top 10 at the Asian Institute of Technology.