Mike Schubert, Holly Liang, Caitlin Makatura, Bryan Solomon, Julie Walker
Abstract
Although recent improvements in prosthetic technology allow a user to control an upper limb prosthesis with some dexterity, a method for receiving proprioceptive feedback (the feeling of body position) has not yet been implemented in commercial hardware. The user must rely on visual feedback to know where his or her limb is in space. We designed and tested a skin-stretch device that rotates the skin about the circumference of the upper arm to relay proprioceptive information. The device consists of a rubber pad on a customized wheel that stretches the skin to the right or left by an amount proportional to the gripper aperture. The device has been incorporated into a system mimicking a typical myoelectric prosthetic arm, and tests have shown that users are easily able to discriminate levels of gripper opening with this skin-stretch feedback. This indicates performance operating a prosthetic gripper could be improved with this device.
Background
Technology for prosthetic devices is progressing rapidly, but very little progress has been made in replicating touch sensations. A lack of touch feedback means that an upper arm prosthesis user must constantly visually monitor his or her prosthetic device while using it. This requires higher mental effort than naturally receiving touch information. In a study surveying upper limb prosthesis users, 76% relied on visual feedback, particularly for grasping tasks and proprioception. 67% of users relied on auditory feedback, primarily for knowing if the limb is moving and how fast. In the same study, the users were asked to rate how important sensory feedback was and what types were most important. Sensory feedback was said to be absolutely important for 45% of the participants, and the most important type was grip force, followed by movement and then positioning (1).
Proprioception, which is the sense of body position and motion, is an important component of touch sensations. Typically proprioceptive information is sensed by stretching skin and nerves in joints and muscles. A person with an amputated arm lacks these receptors and afferent neural pathways to carry this information to the brain. Ideally, it would be possible to connect directly to a person’s neural network and naturally send touch information, but this is difficult and invasive. A simpler method of substituting the sense of touch will allow prosthesis users to benefit from greater dexterity without the complications that arise from a neural interface.
Haptic devices are being applied to relay information to people through their sense of touch. Previous attempts have been made to relay proprioceptive information by tendon vibration and by tactile feedback (tapping, dragging, squeezing, or twisting the skin). Tactile feedback appears to be a more effective and realistic form of feedback than tendon vibration. Dragging, tapping, squeezing, and twisting of the skin are more natural sensations and provide a more intuitive method of feedback to the user. Users noted in a study that gentle twisting of the skin to relay information felt very realistic without being painful after extended periods of time (2).
Twisting skin-stretch feedback systems were also studied by the research team of Karlin Bark at Stanford University. Bark and her team used a device that creates mild rotations and torques to provide feedback. The results of Bark’s experiments showed that skin-stretch feedback provided users with a wide range of easily distinguishable sensations (3). Applying a skin-stretch haptic feedback device to relay grip-aperture information in prosthetic hands could improve dexterity and overall quality of life for prosthesis users.
Problem Statement
This research effort aims to develop a wearable skin-stretch device that effectively relays gripper aperture from a prosthetic hand to a prosthesis user.
Design and Implementation
To successfully relay information to the user we developed an upper arm skin-stretch feedback device, the Rocker, that shifts the skin left and right around the upper arm. We reached this particular design after evaluating a wide variety of methods that could potentially convey proprioception. Among these were rotating (our final choice), squeezing, twisting, and linear movement along the axis of the arm. In the end, through user evaluations, we determined that rotating was the most feasible skin-stretch mechanism due to sensitivity to motion combined with comfort for the user.
Figure 1. The Rocker feedback device
The Rocker, as shown in Figure 1, is strapped to the user’s upper arm and gently stretches the skin beneath it to indicate the current gripper aperture. A Futaba micro servo motor rotates a curved contact piece, and a neoprene rubber pad on the bottom of the contact piece gently stretches the skin below it. The user will be able to tell that a rotation in one direction – for example, a clockwise rotation on the right arm – indicates that the gripper is opening, whereas a rotation in the other direction indicates that the gripper is closing. The extent of the skin-stretch is proportional to the change in gripper position from its natural rest state. The entire system is fixed in a 3D printed housing. Velcro straps attached to each side of the Rocker housing enable the device to fit on a range of arm sizes. The components are identified in Figure 2.
Figure 2. Labeled components of Rocker skin-stretch feedback device
A MSP430 control the gripper movement and relay this position to our feedback device. The MSP430 was selected because it consumes very low energy when the system is in standby mode (in this case, when the gripper is not moving). The price of an MSP430 is low, and it is easy to transfer to a self-designed PCB. The potentiometer in the robotic gripper outputs a position signal, which is used to generate a pulse-width signal to control the movement of the micro servo in the feedback device.
The gripper that was built to test the feedback system was created using an open source design by Yale University called the OpenHand gripper. This gripper has four fingers rather than just two forceps, as many simple grippers do, and so more closely resembles a human hand. This was selected to resemble modern prosthetic devices. The OpenHand gripper is made mainly from 3-D printed components and molded gels material. The entire prosthetic system designed and built study skin-stretch feedback is shown in Figure 3.
Figure 3. System overview of myoelectric control, OpenHand robotic gripper, and Rocker feedback device
Evaluation and Results
The device was assessed with both quantitative tests and a series of user trials. To be considered effective the device must react to a change in gripper position in less than 0.75 seconds. This time was measured through high speed video, marking the time difference between when the gripper moved and when the arm device moved. The feedback time was a mere 0.2 seconds. Another requirement was that the minimum level of “resolution,” defined as the number of different positions that the arm device could move to (note that this is different from the user’s ability to differentiate between these states) be at least 10 different positions. The device far exceeds that by providing a continuous range of positions. The final quantitative criteria was that the user could actually differentiate between positions of the feedback device. This was tested this by sending five different positions to the feedback device in a random order and asking the test subject to identify the current position. The minimum requirement for this test to be 70% accuracy across users. The result of the test exceeded this: the average score for 10 users was 82%, indicating a high degree of sensory feedback resolution. All of these quantitative tests demonstrated that our device could successfully react to gripper position and convey this information to the user.
A user trial was run to assess other components of the efficiency of the feedback system. This study was completed with able-bodied individuals rather than amputees because the touch sensations on the upper arm would be the same in both groups. The design criteria that were tested consisted of comfort of the apparatus, comfort of the feedback, intuitiveness of the feedback, feedback intensity, and consistent noticeability of the feedback. All of these criteria were tested using user scales from 1-5, a score of 3 being the minimum requirement for each category.
Both comfort of the apparatus and comfort of the feedback received a 4.4 out of 5, where a 1 was “very uncomfortable” and a 5 was “no discomfort”. For intuitiveness of the feedback, the users scored our device with a 4.6, where a 5 was listed as “I can understand how this device works in under a minute with no explanation.” For feedback intensity (with a 1 being “I cannot feel any feedback,” and a 5 being “I can clearly feel all motion that the device makes”), the users gave an average response of 4.7. For consistent noticeability, or the ability to identify the location of the gripper after it has been stationary on the skin for an extended period of time, our users scored a 4.4 out of 5, with a 5 meaning they could correctly identify the exact position of the arm device after over a minute of it remaining motionless.
Discussion and Conclusions
The results exceeded all of the set design criteria. The most critical criteria, the ability of the users to identify the exact position of the arm device, surpassed the 70% accuracy requirement by a large margin. This criteria reflects directly on the ability of the user to determine their grip aperture from the feedback alone, which is the core purpose of our design. Furthermore, the impressive 4.6 out of 5 that our users scored for “intuitiveness” means that not only can it relay information about gripper aperture, but it can do so naturally and without an extended learning period. This means that our device could potentially be adapted to an existing prosthetic limb, instead of being limited to those who are just learning how to use one for the first time.
Comfort also scored well, with a 4.4 out of 5 for both the stationary arm device and the feedback, where a 5 was “I feel no noticeable discomfort whatsoever.” The comfort of the device appears to be an easy target for improvement, as the prototype that was used for testing is made out of a hard 3-D printed plastic with no cushions or covering. The comfort of the feedback is a much more challenging criteria, as it retains a negative correlation with both feedback intensity and consistent noticeability. As the device’s motion on the user’s arm becomes softer and more comfortable, it also becomes more difficult to detect its motion. Furthermore, the users’ ability to determine its position after a long period of motionlessness is decreased if the skin-stretch, and thus feedback, is eased in order to make it more comfortable. As a result, this design intentionally emphasizes the strength of the feedback over its comfort, and the result of a 4.4 in the user study was deemed acceptable.
These trials and user evaluations show that the Rocker surpasses all minimum criteria essential for a robust skin-stretch feedback system. This should allow the device to be successfully used in research directed towards improving touch feedback for prosthetic devices. Finally, the success of these trials shows that skin-stretch is a viable method of relaying proprioceptive information to upper-arm prosthesis users.
References
1. Lewis, Soren, Michael F. Russold, Hans Dietl, and Eugenijus Kaniusas. (2012). User Demands for Sensory Feedback in Upper Extremity Prostheses. Medical Measurements and Applications Proceedings.1-4.
2. Stanley, Andrew A., and Katherine J. Kuchenbecker. (2011). Design of Body-Grounded Tactile Actuators for Playback of Human Physical Contact. Proc. of World Haptics Conference,563-568.
3.Bark, Karlin, Jason Wheeler, Gayle Lee, Joan Savall, and Mark Cutkosky. (2009). A Wearable Skin Stretch Device for Haptic Feedback. Proc. of Third Joint Eurohaptics Conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems, 464-469.
Contact
Mike Schubert (Rice University)
Schubes@rice.edu