Safety Issues for Kinesthetic Interfaces in Assistive Robotics

Marcos Salganicoff ,Applied Science and Engineering Laboratories (ASEL), University of Delaware/Alfred I. duPont Institute,Wilmington, Delaware USA

Leslie Hersh ,Department of Mechanical Engineering ,University of Delaware,Newark, DE, USA

ABSTRACT

Kinesthetic displays are mechanical devices that are designed to apply forces and enforce positions at different contact sites on a user under computer control. These devices have great potential, and are finding application in a wide variety of rehabilitation related applications. Exam- ples of these devices include force-reflecting joysticks for wheel-chair control, six degree-of-freedom head-input devices for improved assistive robot dexterity, and haptic displays to aid in the visualization of mathe- matical surfaces for students with visual impairments. Since a kinesthetic display is effectively a robot which operates in intimate proximity to the user, it is essential that safety issues be taken into consideration in their design and installation so that their many advantages can be enjoyed while presenting minimal risk to the user and their surroundings.

BACKGROUND - SHARING THE WORKSPACE

Traditional uses of robotic systems have involved installing robots in controlled access locations within a manufacturing environment [1]. Safety principles used in manufacturing safety standards for robot installations include making sure that the operator is never in the work envelope of the robot when it is powered up and capable of moving and limiting maximum velocities during maintenance and programming tasks [2]. These constraints can be enforced through a variety of safety techniques such as sensorizing the area around the robot through capacitive and infrared thermal sensing, pressure sensitive floor mats, interlocked gates and other barriers such that when a human enters the area around a robot it is put into safe mode.

This type of controlled access to robot workspaces to ensure safety is in contrast to the requirements for rehabilitation robotic systems [4] such as powered orthoses, dexterous general purpose robots, assistive robots for vocational workstations and feeding robots. In order to be useful, these systems almost inevitably require that the consumer and his/her human associates be inside the robot workspace either intermittently or continuously.

Robots are migrating ever closer to the consumer in the form of kinesthetic force displays. These displays [6] have great potential for educational, entertainment (VR) and high-dexterity input to control assistive robots [5]. Kinesthetic interfaces apply forces and moments at the physical interface between the operator and the haptic display. The contact site for a kinesthetic display may be located on a fingertip, palm of the hand, residual limb or at the head/neck depending on the disability.

We have developed a system incorporating specially modified PerForceTM robot (Cybernet Systems, Ann Arbor, MI) that acts as a kinesthetic master in a master-slave telerobotic test bed. This system has been designed to assist people with severe spinal cord injuries by the use of head-movement and proprioception. The system applies the concept of a virtual headstick interface with extended physiological proprioception (EPP) where the user's head/neck experiences the same forces as the slave in order to make him/her feel as if he were in direct contact with the environment (see Figure 1).

PROBLEM STATEMENT

The ability to have force feedback in the system is a curse as well as a blessing. When properly used and scaled, the forces fed back from the slave manipulator can dramatically improve performance on force constrained tasks such as key-insertion and other activities for daily living. If uncontrolled or improperly limited, the applied forces and moments may represent a potential hazard to the operator or people nearby. We consider possible failure modes leading to excessive forces below.

ADDRESSING SAFETY CONSIDERATIONS FOR BILATERAL KINESTHETIC SYSTEMS.

Bilateral telerobotic systems (see Figure 2) allow forces to flow from the operator to the environment and back from the environment to the operator. Our system, which is a representative of this type of system has a number of possible failure modes that may lead to excessive forces being applied to the operator. By inspection of Figure 2, several possible primary failures modes become apparent including wiring faults in the sensing apparatus of the master or slave robots, structural failure of master/slave robots, failures in the communication channel, software faults and computer hardware failures. Although much care can be taken in engineering these systems, both from a software standpoint, such as using reliable software specifications [3] and use of redundant wiring, computing and hierarchical computing architectures and watchdog timers [7] failure in any combination of these subsystems might lead to large and possibly damaging forces being applied to the operator.

Figure 1: Bilateral Kinesthetic Head-Control of an Assistive Robot

Figure 2: System Components

One obvious approach to increasing force safety is to limit the maximum achievable forces and moments that can be generated by the master kinesthetic display through specification of small motors, or current limiting on power amplifiers. This is an appealing approach, but these limits on motor torques may be contradictory with robot control techniques that increase the bandwidth of kinesthetic displays and improve their usability and functionality through feed-forward control which compensates inertial and frictional effects.

The bilateral kinesthetic control architecture presents unique sources of large forces aside from the primary software/hardware failure modes described above. Due to the bidirectional flow of forces between the master and the slave, collisions with objects in the environment may lead to large impulsive forces on the operator. Oscillations may occur if stability criteria for the system feedback loops are violated due to sensor failure or software failure, which in turn changes system gains. Care can be taken to design control systems for telerobots such that they are unconditionally stable such as through ensuring of passivity of all system components from a control standpoint, but this does not guarantee that the system cannot store and release energy in a fashion which might lead to large forces.

Our system incorporates a number of safety features, such as kinematic singularity avoidance on the slave robot, consistency checking of communication packets, and software integrity checking through the use of a watchdog timer. However in order for it, and other systems incorporating kinesthetic master devices, to benefit user communities and gain user acceptance, it will be important to provide a simple force-safety system which is independent of the complex software/hardware subsystems of bilateral kinesthetic systems.

Ideally, we wish to have a breakaway mechanism which completely frees the user's head from the master in the event of extreme forces or torques; a type of "force fuse."

BIOMECHANICAL CONSIDERATIONS

In order to design a safety feature that will prevent injuries to the user, we must first understand how and why head and neck injuries occur. Once the force and torque limits and the mechanisms of injury are understood, a proper force fuse can be implemented.

Injuries at multiple levels of the spine, head, and neck can occur due to applied external loads or moments. The muscles in this region work to protect against injury, but these muscles are all voluntary. It has been found that the total time to maximum muscle force is on the order of 130- 170 ms and is likely to be too long to prevent injury in a surprise impact or moment situation [9]. Therefore, injuries are more likely to occur at some unguarded moment, while muscles are relaxed, and it is unlikely that the operator would have time to react by hitting a kill switch or other de-activating device.

During unexpected impact or torsion, the neck will pass through its normal range of motion and into the stress and trauma ranges with little resistance [10]. Injuries are then more likely to occur at this point because there is a lack of muscle resistance to reduce the momentum of the blow or to increase the time the neck travels through its range of motion.

Although this may be the most vulnerable time for the head, neck and spinal regions, many serious injuries can still occur with the muscles working to resist. This can especially be a problem when the user has a motor disability such as a previous injury which may have lead to a weakened musco-skeletal complex. Some preliminary work has been done in characterization of head-motion limits and maximum reaction forces for subjects with high-level spinal cord injuries [8] which provides a base-line for design specifications via safety-margins. Practically it will be necessary to measure such parameters in a fitting protocol to take into account the unique biomechanical ranges of each user.

DESIGN APPROACH AND DISCUSSION

Taking into consideration the biomechanics of head and neck injury, as well as the application of this device, a number of specifications, or criteria arise. First, the device must be able to release instantaneously, since muscles are most vulnerable in their relaxed stage. Also, it needs to release at the predetermined force and torque limits. Since this may be different for each user, an adjustable strength device would be ideal. The device should have precise and repeatable disengagement properties and not disengage before the proper limits. Finally, dealing with the self- sufficiency issue, the breakaway mechanism should be able to reattach easily under the control of the user. We are currently evaluating 3 different conceptual designs for the force-fuse, a friction based tensioning system which enforces normal forces between two disks with known frictional characteristics, a ski-binding based approach, and an analog electronic force/torque and accelerometer based approach. The expected strengths and weakness of different approaches are briefly summarized in Table 1.

REFERENCES

[1] J.H. Graham, Safety, Reliability and Human Factors in Robotic Sys- tems, Van Nostrand Rheinhold, New York, 1991

[2] OSHA, Guidlines for Robotic Safety, OSHA Instruction PUB 8-1.3, Washington DC, OSHA

[3] W. Harwin and T. Rahman, Safe Software in rehabilitation mecha- tronic and robotics design, RESNA `92, pp. 100-102, June 6-11, 1992

[4] H.F. Machiel Van der Loos, D. S. Lees and L.J. Leifer, Safety consid- erations for rehabilative and human-service robot systems, RESNA International `92, 1992

[5] M. Salganicoff, D. Pino, V. Jayachandran, T. Rahman, R. Mahoney, S. Chen, V. Kumar, W. Harwin, Virtual Headstick Rehabilitation Robot System,'1995 IEEE International Conference on Systems, Man and Cybernetics, Vancouver, B.C., Canada, 1995

[6] T.H. Massie and J. K. Salisbury, The PHANToM Haptic Interface: A device for probing virtual objects, Proceedings of the ASME Winter Annual Meeting, Symposium on Haptic Interfaces for Virtual Envi- ronments, Chicago, IL, November 1994

[7] L.A. Jaros, S.P. Levine, J. Borenstein, T.E. Pilutti, and U. Raschke, Fail-safe features of a mobile robotic platform,,Resna 13th Annual Conference, Washington,DC, 1990

[8] C.A. Stanger and A. C. Phalangas, Range of Head Motion and Force of High Cervical Spinal Cord Injured Individuals for the Design of a Test Bed Robotic System, ASEL Tech Report ROB9402, A.I. duPont Institute, Wilmington, DE

[9] G.W. Nyquist, and A.I King, "Spine", Review of Biomechanical Impact Response and Injury in the Automotive Environment, National Highway Traffic Safety Administration, pp. 45-92.

[10] S.E.Reid,, Head and Neck Injuries in Sports, Springfield, Illinois: Charles C Thomas, 1984.

Table 1: Design Comparison

ACKNOWLEDGMENTS

This research is supported by the U.S. Department of Education, Grant #'s H129E20006 from the Rehabilitation Services Administration and H133E30013, the Rehabilitation Engineering Research Center on Rehabilitation Robotics, from the National Institute on Disability and Rehabilitation Research (NIDRR) of the Department of Education. Additional support is provided by Nemours Research Programs. We gratefully acknowledge the assistance Professor Rami Seliktar in this project.

Marcos Salganicoff Applied Science and Engineering Laboratories A.I. duPont Institute 1600 Rockland Road, P.O. Box 269 Wilmington, DE 19899 USA

Internet: salganic@asel.udel.edu

Web: http://www.asel.udel.edu/~salganic