The Rehabilitation Robotics Research Program

Chameleon: A Body Powered Rehabilitation Robot


This project involves the development of a wheelchair mounted, gravity balanced, mechanical arm whose end-point is controlled and powered by a functional body part of a person who can not use his/her arms. Two important design features are adaptability to different user input sites and ability to provide external power augmentation.


A prosthesis is a replacement for a missing body part. For example, if someone's arm is amputated after severe trauma, a prosthesis may be attached to the remaining limb. The prosthesis may provide cosmetic value or it may provide some functional abilities such as grasping and wrist rotation. Two methods are commonly used to power these articulations. The most common is the use of a cable to apply forces from a body member (e.g., the shoulder) the person can still use, to one of the prosthesis' articulations. Recently, EMG (elecro-myograph) signals have been used to control an external power source (usually an electric motor) for powering a prosthesis' articulations. Cable actuated prostheses have been used successfully for many years. Three primary reasons for their success are the following:

  1. Extended Physiological Proprioception (EPP) - A prosthesis is an extension of a person's limb, thus the user knows its position without visual feedback.
  2. Force Feedback - Forces felt at the prosthesis' gripper are transmitted back to the user's activation site via cables.
  3. Low Cost and Design Simplicity - The simple design of these devices make them affordable and maintainable.

In 1963, researchers in New Zealand developed a spoon feeder called the Distaff. It was a spoon mechanically linked to a foot-operated lever. Reportedly [1], it was difficult to use. In 1983 the Oxford Orthopaedic Engineering Center (OOEC) developed a cable-driven mechanical feeder (called "Magpie"), where spoon position (output) was proportional to foot position (input). This device couples 4 input motions (1 at the knee, 1 at the hip, and 2 at the ankle) to 4 spoon motions. After one month of using Magpie, it was reported that 12 of the original 16 patients fitted still used it for feeding at least once a day. The magpie was an effort to apply the successful concepts (primarily those given above) from the prosthetics field to the rehabilitation robotics field.

The goal of the Chameleon project is to develop a device conceptually similar to the Magpie, but functionally more general. It should enable the user to perform tasks such as lifting light-weight objects, operating switches, feeding, and simple manipulation. A range of input devices should be available for controlling the arm, where each one matches the abilities of a particular user. The project has been named Chameleon because of its adaptability.


The goal of this work is to develop a technologically simple, wheelchair mounted manipulator to allow a person with no or very little arm function to interact with his surroundings. It has been shown that a system which combines visual feedback with sensory feedback is superior to visual feedback alone [2]. The population that would most benefit from this device are people with spinal chord injury, multiple sclerosis, and stroke. The following four features highlight the design objectives.

  1. Intuitive and easy to use - The inputs of the user should map in an integrated manner to the outputs of the manipulator: a proportional, three-dimensional mapping of the user's input signal to the position of the arm's gripper is desired. A direct connection between the user interface (referred to as the "master") and arm (referred to as the "slave") facilitates a system which is easy to use since proprioception and force reflection are naturally built into the control system. For example, the user nodding his head "yes" corresponds to the slave's arm moving up and down.
  2. Modular - The system will be modular in two senses. First, the arm will accept or be adaptable to several different user inputs. These inputs depend upon the available user body motions, which to a large extent, depend upon the user's disability. For example, if the best available user input is from the head, the arm needs to accommodate an interface unit is designed for head input. Alternatively, if the best user input is from the hand, the arm needs to accommodate whatever interface is designed for hand input. Another way the system needs to be modular is in its ability to accept power assist units. In this way, if the user can not supply sufficient power to the interface (master) to directly cause the arm to move, power amplifier modules will be added to specific joints to assist the user in operating the arm. In order to minimize the user's effort, the arm needs to be gravity compensated as best as possible.
  3. Cost - The high cost/usefulness ratio of most rehabilitation robots makes their use very limited. It is the goal of this project to maintain a simple design philosophy so costs can be kept to a minimum.
  4. Aesthetics - The arm is designed to geometrically and functionally resemble a human arm. The interface unit and transmission system are designed to be as unobtrusive as possible.


    To date, several arm (slave) prototypes have been developed. Additionally, several head interface (master) units and one hand interface unit have been developed. We decided to focus on head input, since this would serve the largest population of potential users. The general idea, for the case of head input, is that the user engages to the master by biting onto the mouthpiece. When the user moves his head, the master moves in a corresponding fashion. Cable connections and electrical connections exist between the master and slave so that movement of the user's head ultimately results in movement of the slave's arm. Below is a photograph showing one of our engineers using the device to lift blocks from the box and then stacking them.

    [The Chameleon]

    Slave (Arm) Design

    One of the design objectives is to have an end-point controlled, mechanical arm which kinematically resembles the human arm. To facilitate this objective, the slave's kinematics were designed around a spherical coordinate system {øy, øp, r} defined in the following drawing.

    [Diagram of the Slave]

    Yaw (øy) and pitch (øp) motions are achieved through pulleys fixed to shafts at these axes. The radial motion, r, is activated when ør1 and ør2 rotate at the same speed in opposite directions. This is achieved through the radial motor rotating about the r1 axis and the parallel bar linkage (only partially visible in the diagram) causing one of the gears to rotate about the other, thus causing relative motion between the two links. A gripper located at the arm's end can both rotate about an axis co-linear with the distal link (øg), and open/close (P). The slave is approximately counter balanced through out its range of motion by the weight shown at the back of the arm.

    Head Master Design

    The design of the head master (shown in the diagram below) was done to give the user as much control as possible over the slave arm. Two rotary joints are present allowing the user to pitch (Øp) and yaw (Øy) his/her head to move the master. These two joints are mechanically linked to their corresponding joints (øp & øy) on the slave arm through Bowden cable (i.e., a wire sliding inside a housing), thus requiring the user to provide all the effort to rotate the slave arm about these axes.

    [Diagram of the Master]

    The radial motion required by the user, denoted by R, is quite minute. In fact, this motion is close to isometric. The user applies anterior/posterior forces to the mouthpiece, causing deflection of a cantilever beam (located inside of the radial link), which in turn activates Force Sensitive Resistor (FSR®) sensors. The signals from the FSRs® are used to activate the motor located at the back of the slave (radial motor) to cause radial motion, r, of the arm. Since this motion (r) is controlled electronically, the amount of force required by the user to cause this motion can be adjusted. The radial link is made from clear Lucite® (an acrylic plastic) so that it interferes as little as possible with the user's vision. Between the radial link and the mouthpiece mount is a universal joint to:

    1. enforce that only forces (not moments) applied to the mouthpiece cause cantilever deflection
    2. provide extra degrees of freedom to make the yaw and pitch movements of the head easier when the axes of the master and user's head do not exactly align.

    The design of the mouthpiece is quite involved and therefore just a few of the important ideas are mentioned below. A mouthpiece comfortable to the user, stiff enough to transmit forces and torques from the user to the master, and machinable for easy attachment to the master and joystick were three requirements. We worked with Dodd Dental Laboratory [4] in making custom and generic acrylic mouthpieces (lined with silicone). A custom mouthpiece has indentations (for the upper teeth) that fit the dentition of the user (after molds of his/her teeth and modifications have been made). The advantage of a custom mouthpiece is that less biting force is required by the user to engage with the mouthpiece, thus reducing fatigue during operation. However, for the purposes of prototype testing, the cost and time associated with making custom mouthpieces for several testers/evaluators is not justified, so Dodd Dental Laboratories made several generic mouthpieces to meet our application needs.

    A MICROJOYSTICK is used as a tongue control input for controlling the gripper. The joystick is mounted on the mouthpiece mount and the joystick's shaft passes through the mouthpiece. The user presses with his tongue on a teflon ball placed on the shaft's end. The MICROJOYSTICK uses FSR® technology developed by Interlink Electronics; it is the same joystick used on many laptop PCs. Two of the co-linear directions (e.g., North and South) correspond to opening/closing the gripper while the other two co-linear directions (orthogonal to the previous ones) correspond to rotating the gripper, øg, as shown in the above diagram.

    Interface Between Master and Slave

    Bowden cables link the the yaw (Øy) and pitch (Øp) axes of the head master to that of the yaw (øy) and pitch (øp) axes of the slave, respectively. This enforces a direct position mapping at these joints between the two units. Currently, the cable system is comprised of three components. A teflon® coated steel wire runs inside of a polyethylene liner which in turn runs inside a long, 1/8" diameter extension spring. The components have been carefully chosen to obtain a flexible (not stiff) transmission system, but to minimize friction.

    Present Work

    1. Counter Balancing - As mentioned previously, a single weight was used to approximately balance the slave throughout its range of motion. Unfortunately, the balancing was not good enough at the extreme ranges of pitch motion, causing the user to have to apply too much force in certain positions. A more detailed static analysis was performed on the slave and a solution was analytically found to exactly balance the slave throughout its range of motion. The solution was subsequently modeled in Pro/Mechanica (dynamic and animation software from Parametric Technology) and applied to the slave. Although not shown in any of the current figures, it involves using two weights (rather than one) placed at specific positions.

    2. Control of gripper - We are currently changing the control of the roll gripper motion. Presently, this motion is activated by the tongue pushing the MICROJOYSTICK. We are changing it so roll rotation of the mouthpiece corresponds to roll rotation of the gripper. This is being done for two reasons. First, users seem to have difficulty controlling the four orthogonal joystick directions with their tongue. Secondly, it is non-intuitive to control the roll of the gripper with either a N-S or E-W joystick movement; this causes the user to become confused as to which input direction causes the gripper to roll and which causes the gripper to open/close.

    3. Force Amplifier - Due to the geometry of the system, a light object being gripped by the slave can feel heavy to the user. These forces can cause fatigue and discomfort at the interface after only a short period of use. Additionally, if the weight is excessive, the user may be unable to lift it. In order to eliminate these problems, a force amplification device is being designed to increase the forces applied about the pitch (lifting/lowering) axis.

      The following photo shows the test-bed setup with a motor, FSRs®, pulleys and weight. The user applies an input force (shown as the hand pulling up) which is sensed as a voltage at the FSR®. The voltage is then sent to the PC and the corresponding motor voltage is calculated and output to the motor. The current design utilizes LabVIEW®, a PC based program, to calculate a voltage to send to the motor. Recent improvements to the system include the reduction of oscillations of the motor, and verification of the control algorithm. Once this phase of the design is complete, the force amplifier will be incorporated into the Chameleon. The cables in the existing system will be replaced by the Bowden cables, and the pulleys at the pitch joints of the master and slave will replace the two pulleys in the test-bed.

      [Test-Bed of Amplifier]


      So far, the Chameleon has been tested on an informal basis by several people in our laboratory and one person with athrogryposis who is a mouthstick user. Since then, many changes have been made to address the concerns and comments made by the testers. Please refer to the following critique to see details of the evaluation. Although the robotic system's design has been guided by consumer input and the evaluator's comments, several issues still need to be resolved.

      • Safety - Large forces applied to the arm transmit large forces and torques to the human interface and hence, the user's head. A means to avoid this occurrence needs to be devised.
      • Donning/Doffing - Since this system will be secured to a wheelchair, adjustability for allowing the user to easily engage/disengage from the system is required.
      • Adjustability - To allow recalibration of the system and to allow the robot to fit many users, variables such as link lengths and cable connections need to be adjustable.
      • Power assistance - Although the design will try to reduce friction as much as possible, the user may not have sufficient strength to completely body power the arm. To compensate, mechanical power amplifier modules are currently being investigated to determine if they can be incorporated into the system.
      • Locking - The user needs the ability to lock the arm in position when rest is necessary or the input site is needed for another purpose.


      1. MAGPIE - Its development and evaluation, Internal Report: Oxford Orthopaedic Engineering Centre, Nuffield Orthopaedic Centre, Headington, Oxford, England, 1991.
      2. B. Hannaford, L. Wood, D.A. McAffee, and H. Zak, "Performance Evaluation of a Six-Axis Generalized Force Reflecting Teleoperator", IEEE Transactions on Systems, Man, & Cybernetics 21(3):620-633, 1991.
      3. Rahman T., Ramanathan R., Seliktar R., Harwin, W., "A Simple Technique to Passively Gravity Balance Articulated Mechanisms", Transactions of the ASME-Journal of Mechanical Design, vol. 117, December, 1995.
      4. We greatly thank Dodd Dental Laboratories for making several mouthpieces for this project free of charge.

      Project Staff

      Tariq Rahman, Ph.D. (Primary Investigator)
      Kelly McClenathan (Graduate Student)

      The counter below indicates how many people have accessed this site since January 15, 1998.


      [Robotics] [ASEL]

      Last Updated: January 15, 1998 by Sean Stroud <>