Fig. 1. VIDOF 6-d.o.f. Joystick Prototype.
Van der Loos, Hendrik F. Machiel
Palo Alto Veterans Administration Rehabilitation R&D Center
Reprinted from the Proceedings of the RESNA Annual Conference, San Diego, CA, June, 1983
Copyright RESNA Press, Inc. 1983
The applicability of robotic manipulation aids for the physically handicapped depends largely on the device used as the man/machine interface. Commanding a system to perform a task is possible through keyboards or by voice, but guiding the motion of a mechanical arm requires an analogic input device, such as a joystick, for effective, real-time control. The design of a six degree-of-freedomjoystick is discussed; a prototype implementation is described; and its role as an effective assistive device is postulated.
The impairment of motor function as a result of, for example, spinal cord injury leads to an individual's handicap in self-care and vocational endeavors. One recent application of robotics involves the restoration of the patient's capability to manipulate objects in his personal space. Through modalities, such as voice, joystick input, and head motion, a robotic arm can be guided through the logic and through the trajectories to accomplish complex manipulation tasks for its user. The generic term 'joystick' refers to any input device of a specified number of degrees of freedom that transduces motion and/or force to electrical signals.
In the use of a joystick, the operator is called upon to perform a mapping of the joystick inputs to the motions induced in the controlled device. In the case of a 2-axis electric wheelchair joystick, this is a simple mapping task; in the case of a six degree-of-freedom (6-d.o.f.) robot, a joystick must be designed to reduce the mapping to instrumented 'pointing' in the desired direction of motion of the gripper. The latter was the primary goal of the process leading to a 6-d.o.f. joystick design applicable to rehabilitative robotics in particular.
Interfaced to a voice-controlled robotic manipulator, the joystick permits analog control of the mechanical arm's motions by a user having limited manual dexterity, range of motion, and force.
Industrial robotics concentrates on developing fully automated alternatives to human function; telemanipulation research endeavors to produce a perfect mechanical extension of man for the handling of materials where he himself can not go. In between these 2 fields lies Interactive Robotics. The goal of this field is to combine direct human control of a mechanical arm with automation, by an on-board computer, of low-level operations such as obstacle avoidance, object centering and grasping, peg-in-hole insertions, and line-and plane-following. The concept of distributed decision-making involves human switching of operation modes through a symbolic information channel (keyboard or voice), and human analogic control of real-time processes through a joystick configuration (3,8).
In the field of rehabilitative robotics, R&D emphasis has been placed on endowing the manipulator system with pre-programmed features (9), restricting human input to a minimum. The search for a lowest-common-denominator input device for use by a large proportion of consumers with a range of disabilities is desirable for implementation standardization. On the other hand, devices designed on an individual basis and exploiting the remaining motor and expression functions could offer the user a much richer interaction. An example is the 5-d.o.f. hydraulic mouthstick with grip force feedback developed by Torfason (12). Fullrange and dexterous head motion is required, thus usable by C-4 and below quadriplegics. The device directly controls the actuators of a hydraulic telemanipulator. Hand-operated joysticks of more than 3 axes are not common in industry, and only very few 'full-access', i.e., 6-d.o.f., controllers exist even in R&D environments. Lower-level quadriplegics, arthritics, and those with birth defects may have residual use of one or both upper limbs. Limited manipulation ability can be rendered useful through special devices: velcro strips for wheelchair joystick attachment, strapon spoons, etc. Using residual ability for the control of a 6-d.o.f. joystick is deemed possible through extrapolation of proven industrial joystick design concepts and innovative control strategies.
Preference and performance tests have been conducted in tracking and positioning tasks. Through these and the last century's evolution of the design of control devices, certain concepts have emerged. The most basic is the preference of the user to be part of a linear control system (4). Non-linearities in the linkages or behavior of a system, such as backlash, stiction, delay in response, and saturation of the control input are diagnosed as being undesirable. Though actual performance may not always be diminished (10), increased concentration levels and frustration can be psychological deterrents to a system's effectiveness.
For a linear system, the magnitude of the gain factor between the input device signal and the observed effect is not strongly related to performance. Indeed, in one study it was shown that over a gain factor range of 100, the performance varied by two orders of magnitude less (11). In this study and elsewhere (13), it was noted that there did appear to be a local optimum corresponding to a specific controller gain. This observation points up the need for an optimization phase for any controller implementation if maximum performance is desired.
The second concept involves the role of proprioceptive feedback. Best performance results have been obtained for spring-return displacement sticks and isometric joysticks (7). The first category has an advantage in jolting controller environments, such as electric wheelchairs, where the hand's accelerations cause disturbance forces to be imparted to the control stick: the added position cue of displacement minimizes the disturbance effect (1). The second use for proprioceptive feedback is in distinguishing one axis of control from another. A controller with a center detente permits full linear behavior throughout the non-zero control space, while cueing the operator when the zero-position in any direction is attained. For an electric wheelchair, this provision allows for forward motion without controller-induced side-to-sidedrift.
The third concept involves directional stereotypes. As users of appliances, vehicles, and robots, we expect a forward push on a control stick to induce forward motion on the controlled device. Studies have shown that deviations from expected relationships are very difficult to incorporate permanently (6).
The above concepts of linearity, proprioception, and directional expectations have been used as the basis for the design of a 6-d.o.f. manual joystick for a computer-controlled mechanical arm.
The context of the joystick project is the Stanford University and PAVAMC RR&D Center Robotic Aid for the Handicapped (5). Based on an industrial manipulator, the Unimation PUMA-250, augmented by a Z-80-based host processor, and voice recognition and synthesis modules, and run on innovative control software (2), the system allows easy incorporation of additional I/O devices, such as displays and joysticks.
A prototype 6-d.o.f. joystick, VIDOF (fig. 1), employs 2 groups of 3 inputs, separated into major and minor axes (fig. 2). The 3 major inputs are managed by a wrist-centered handgrip, controlling the translations of the gripper of the robot along the 3 cartesian axes (x,y,z). The minor inputs are managed by the thumb and forefinger through a finger-stick mounted on the handgrip platform. These 3 inputs control orientation (pitch: rx, roll: ry, yaw: rz).
Each control axis is of the center-detente, displacement type, with the ranges of motion adapted to human anatomy. Force gain factors are adjustable.
Computer software to convert the 6 inputs to robot arm velocities permits the study of 1) effects of controller gains and saturation, i.e. maximum attainable speed, 2) the use of filters to remove or diminish the effects of spasms and other non-linearities induced by the user's disabilities, and 3) the optimum topological mapping of joystick inputs to robot control directions.
Fig. 1. VIDOF 6-d.o.f. Joystick Prototype.
A prototype has been built (VIDOF), implemented, and has undergone a preliminary phase of parameter adjustment for able-bodied users. Qualitative feedback from users indicates that the device requires a learning stage before confidence is attained. The joystick allows complex, high-speed maneuvers with the arm: similar to driving a race car, this is a scary adventure for the neophyte.
The implementation is in conjunction with the software package designed to allow voice control of the arm. The symbolic input, though also available for voice-piloting ("left", "forward", "turn", etc.) is primarily used to complement the analogic joystick input: mode changes and coordinate system modifications are done by voice commands .
As the development of VIDOF is only in its initial stages, a number of studies and tests are projected. The most salient is the study of characteristics and features required for optimal use by both the able-bodied and those with limited arm and hand capabilities. Spring rates, center detente levels, ranges of motion, software filter types and parameters, and mapping conventions are physical quantities targeted for observation.
Fig. 2. Control Allocation of the Six Joystick Inputs.
The development of analogic input devices for telemanipulator and robot systems is becoming feasible through the same advances in electronics that made robotics itself possible. The exotic field of rehabilitative robotics requires, by definition, energy to be spent on characterizing disabilities and utilizing remaining strengths. In addition to being a powerful environment for the design of assistive devices for the handicapped, this field can provide strategies and technologies for the development of multi-axis input devices in robot and telemanipulator stations in general. The joystick described here is an attempt to introduce a versatile device for a range of robotic applications .
This work is being supported by the Palo Alto Veterans Administration Medical Center and the Department of Mechanical Engineering of Stanford University in the context of a joint "Robotic Aid for the Handicapped" Project.
1) G.T. Black, Flying Qualities Design Requirements for Sidestick Controllers. Wright-Patterson AFB Rept. No. AFFDL-TR-79-3126, Oct., 79.
2) C.W. Buckley, L.J. Leifer, Fast Interactive Control of a Manipulator for the Severely Disabled. Proc. of the ACC, Arlington, VA, June, 82.
3) W.R. Ferrell, T.B. Sheridan, Supervisory Control of Remote Manipulation, IEEE Spectrum, vol. 4, Oct., 67, pp. 81-88.
4) D. Graham, Research on the Effect of NonLinearities on Tracking Performance. Aerospace Medical Research Laboratories, Wright-Patterson AFB, Ohio, July, 67.
5) L.J. Leifer, Rehabilitative Robotics, the Stanford Robotic Aid. WESCON '81, San Francisco, CA, Sept. 15-18, 81.
6) N.E. Loveless, Direction of Motion Stereotypes: A Review. Ergonomics, 1962, 5(2), pp. 357-383.
7) M.H. Mehr, Manual Digital Positioning in 2 Axes; A Comparison of Joystick and Trackball Controls. 16th Annual Meeting of the Human Factors Society, Oct., 1972.
8) J.L. Nevins, D.E. Whitney, Information and Control Issues of Adaptable, Programmable Assembly Systems for Manufacturing and Teleoperator Applications. Mechanism and Machine Theory, 1977, vol. 12, pp. 27-43.
9) W. Schneider, G. Schmeisser, W. Seamone, Computer-Aided Robotic Arm/Worktable System for the High Level Quadriplegic, IEEE Computer, January, 1981, pp. 41-47.
10) G.P. Starr, Comparison of Control Modes for Time-Delayed Robot Remote Manipulation. Ph.D. Thesis, Department of Mechanical Engineering, Stanford University, 1978.
11) C.L. Tipton, H.P. Birmingham, Influence of Control Gain in a lst Order Man-Machine Control System. Human Factors, 1959,1, pp. 69-71.
12) L.E. Torfason, F.C. Risteen, A Single Joystick Hydraulic Control System with Six Independent Simultaneous Velocity Proportional Degrees of Freedom. ASME publ. 70 MECH-54, 1970.
13) T.C. Way, Studies in Manual Target Designation: Isometric Control, Report D162-1025801, Boeing Co., Seattle, July, 1970.
VA RR&D Center (MS 153) 3801 Miranda Avenue Palo Alto, CA 94304.
L.J. Leifer, R. Sun, H.F.M. Van der Loos, Terminal device centered control of manipulation for a rehabilitative robot. Proc. of the Joint Automatic Control Conference, San Francisco, CA, September, 1980.