Log in / Register
Home arrow Computer Science arrow Robotic Assistive Technologies: Principles and Practice


The flow of signals from a prosthetic user to his or her device allows a device to perform motions that enact the intent of the user. However, as in conventional control systems in machines of all kinds, the use of a robotic prosthesis is challenging or impossible for a user without a complementary channel of information flowing from the device back to the user. This information to the user, termed feedback, is a critical part of effective prosthesis control.

In its simplest form, feedback is provided to the user as a by-product of the form and regular operation of the robotic prosthesis. Users are able to see and hear how their limb is moving and interacting with other objects. Vibration, torque, or impact forces to the limb are transmitted through the chassis of the device to the interface with the user’s body. These mechanical sensations are well known to be an important way for users to interpret the operation of their prosthetic device, even when they are not looking at it. Similarly, the sound, vibration, and movement of all the actuators within a limb are conveyed to the user through the chassis and have been reported to be one of the most important components of feedback for users of commercially available limb systems (Lundborg and Rosen 2001). At present, the majority of available prosthetic technologies use intrinsic signals from the device to the user as their only form of feedback-sensation is not explicitly recorded and transmitted from a robotic prosthesis to the user. For lower limb robotic prostheses, vibrations, sounds, and impacts often provide a significant percentage of the information that a user needs to skillfully locomote; for upper limb prostheses, users often desire more information about what their prosthesis is feeling and how it is operating.

One limitation of the simple forms of feedback noted is that they do not capture the full range of sensations that might be accessible to a biological limb; thus, the type of feedback signals that help provide dexterous, natural control of an artificial arm and hand is missing. For example, in many cases it would be desirable for users to receive feedback about temperature, texture, the motion or position of the limb in space (proprioception and kinesthesia), and even damage or pain. These modalities are not typically present in commercially available devices but are the subject of significant research and development. Communicating the full range of perceptual information from a device back to the human is considered to be a significant remaining challenge for closing the loop between a user and a robotic prosthesis.

Some approaches to closing this loop aim to link actual actions of the device to specific sensations delivered to the user’s body that are perceived by a user in the same way as the information that was recorded by the robot (e.g., the user perceives the robot’s contact with other objects as touch or its proximity to a flame as heat). This feedback can be provided in ways that are matched or nonmatched in form and physiology (Antfolk et al. 2013; Schofield et al. 2014). True physiologically matched sensations would have the patient perceive the sensation on the robot arm as the exact same sensation on their now-amputated biological limb—that is, the patient perceives pressure on the robot’s index finger as pressure on his or her missing index finger. An alternative approach that leverages the body and brain’s ability to adapt is substitution, which delivers sensation to the body in forms other than the way they were recorded (nonphysiologically matched sensations). For example, force of contact with an object may be reported to the user by vibration in the socket of the limb or at another location on his or her body. In either case, physiologically matched or unmatched feedback, the intent is for the user to understand aspects of the operation of the device not otherwise readily perceivable through the user’s direct physical attachment to the device. The device is actively, as opposed to passively, transmitting information to the user.

Because a robotic prosthesis has internal information relating to its actuators, power system, sensors, and control system, devices may also provide a user feedback about things not directly relating to the physical interactions of the prosthesis with the environment. It is important to communicate information to the user about the operation of the prosthesis itself, for example, communicating to the user which functions or modes the user is currently operating or signaling to the user that the battery is getting low. These forms of information must be communicated in a way that the user immediately knows what they mean and in a way that is not distracting or irritating during constant use. Wireless links to external devices such as a smartphone or tablet are also possible ways for users to receive different forms of information about their prosthesis, its operation, and the kinds of signals it is perceiving from the user’s body (e.g., plots of the muscle activity signals being recorded in the socket or a schematic of the configuration of grasps that the user can currently select during their use of the device).

With these examples in mind, we can readily identify that the core principles of feedback and control of robotic prostheses are in fact similar to those of human-to-human communication and also to those of machine- to-machine communication—for example, information theory, communication theory, classical cybernetics, and a large body of work dating back to researchers such as Harry Nyquist, Ralph Hartley, Claude Shannon, Norbert Wiener, and Alan Turing. Signals should be clear and interpretable, contain an amount of information that is appropriate to their complexity, and readily preserve the intent of the sender on its interpretation by the receiver. Moreover, different signals need to be distinguishable as unique for the control system to understand that each signal has a different meaning (as well described in a mathematical sense by Shannon 1948). Practically, this means that the feedback delivered from a robotic prosthesis should [1]

  • • Capture the most useful or important aspects of the information being recorded by the prosthesis (saliency); and
  • • Optimize how the modes of sensation recorded from the robot are matched to the perceptual information perceived by the human body (perceptual alignment or matching).

  • [1] Not be too detailed or complex for the user to understand and bepresented in a way that it can indeed be understood (clarity andinterpretability); • Be delivered at a rate that is appropriate for the physical interface withthe user (frequency, timing, and timeliness);
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science