Rehabilitation Robotics

Robotic Manipulandum for Reaching

We are using a robotic manipulandum to study how control subjects and patients perform actual reaching and reaching in virtual environments. A device prototype was designed and built that allows user interaction via the handle; users physically translate the handle in space or apply force to the handle in a static position measured by a force sensor. Either position or force input is mapped to the movement on a cursor on a computer screen, positioned horizontally over the hand's workspace. The robot can also apply forces to the user that assist or resist motion via two motors and a cable transmission system.

Using the system, we are investigating how people learn to perform reaching movements in a virtual environment that requires only static force input. We are comparing different control strategies for mapping force input to cursor movement, with the goal of selecting a strategy that is most similar to natural movement. Further experiments will determine how the motor system learns isometric tasks and ascertain whether the motor system forms an internal model in hand or joint space. Ultimately, we will test whether learning virtual reaching in an isometric environment will transfer to dynamic reaching, with the goal of developing patient-specific rehabilitation strategies. The use of isometric devices for rehabilitation represents a mechanically simpler and cheaper solution compared to other rehabilitation robots.

HAPI Bands: Haptic Augmented Posture Interface

The HAPI Bands system is a wearable device that performs the role of a posture coach, helping a user maintain static poses of the upper body. Motion of the user's body is tracked in space via the Microsoft Kinect range camera. While the Kinect records 3D position of upper body joints in space, it cannot capture rotation of the arm about its axis. Body-mounted accelerometers record arm orientation information and complete sensing in 6 DOF (degrees of freedom). To interact with the system, the sensing and software capture a desired upper body pose. As the user misaligns his body from the "target" he receives vibration feedback from eccentric mass motors mounted on the interior of 5 bands worn on the body. Vibration signals are unique in that they are applied to the misaligned body segment and indicate the direction that the user should move to correct the misalignment. Joint-errors from the target are measured in real-time and a single tactile display is given to help user correct the joint with greatest misalignment. A total of 15 DOFs are corrected. Initial testing in a pose-recall task indicated that the haptic feedback was as effective visual feedback in helping user's locate target positions in space.

Current work on this project includes the implementation of a wireless solution and feedback to multiple joints simultaneously. Future work includes the extension of the system to include both upper and lower body, in addition to correction of full motion trajectories.



  • National Science Foundation, Stanford University, Powell Foundation

Click here to return to the main research page.