Researchers from the University of Waterloo are using MapleSim to develop a musculoskeletal model of the human arm to in order to design new model-based controllers - ultimately leading to a better understanding of movement disorders of the arm.
In this project, researchers at the University of Waterloo, Borna Ghannadi and Dr. John McPhee, used MapleSim from Maplesoft to develop a musculoskeletal model of the human arm that provides the human action for an upper limb rehabilitation robot, in order to develop new model-based controllers for it. The controlled robot is tested in partnership with the Toronto Rehabilitation Institute (TRI) and Quanser.
Movement disorders in the upper extremities, which are common among post-stroke patients, demand effective rehabilitation procedures. Rehabilitation robots are now being used clinically, but because of emerging proposals for motor learning there is still much that can be done to improve the designs and control algorithms of these robots. For example, one of the neglected aspects in the design and development of rehabilitation devices is the modeling of human interaction with the robot.
The TRI/Quanser robot is an end-effector based planar robot, which performs reaching movements in the horizontal plane for therapy of the shoulder and elbow. The team decided that a fitting starting point was to develop a simplified planar 2D musculoskeletal arm model which consists of two hinged links and six muscles, and assumes no tendon compliance.
After evaluating tools from multiple vendors, the team selected MapleSim for their model development work. Describing their choice, Dr. McPhee said: “Taking into account simulation times and quality of results, MapleSim, because of its symbolic computation technology together with optimized code generation, performed better than the other software platforms. Therefore, we selected MapleSim for use throughout this project.”
The team then developed an impedance controller which can automatically adjust itself in a variable admittance environment, representing the variable levels of movement disorders affecting rehab patients. The controller was simulated running on the 2D model, in 4 different modes. The first two modes, simulating a healthy arm, were used to calibrate and tune the controller, while the second two modes which simulated a post-stroke patient’s arm, were used to evaluate its performance.
Hand position error and muscle activation levels were measured and compared during simulation runs in the different operating modes. The results were positive, and in line with expectations, demonstrating that it is possible to use musculoskeletal arm models to evaluate the planar robot.
During the next phase of the project, the team will develop an advanced 3D musculoskeletal arm model with integrated muscle wrapping. As musculoskeletal models become more detailed and life-like, engineers are able to enhance the design of control algorithms for upper limb rehabilitation robots, which ultimately improves the rehabilitation process for post-stroke patients.
Additional resources demonstrating how Maplesoft technology is used in other robotic and mechatronic applications are available here.
