SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
SPIE Photonics West 2018 | Call for Papers

SPIE Defense + Commercial Sensing 2018 | Call for Papers




Print PageEmail Page

Sensing & Measurement

Magnetic levitation haptic interfaces will impact training and design

An interview with Peter Berkelman and Ralph Hollis, Robotics Institute, Carnegie Mellon University

From OE Reports Number 165 - September 1997
31 September 1997, SPIE Newsroom. DOI: 10.1117/26.199709.0002

Please define haptic.

Berkelman: Haptic means interacting through a sense of touch that combines tactile and kinesthetic sensations. Haptic interaction is a relatively new research field. Several research groups are developing various kinds of devices, including linkage devices, tension cable devices, or exoskeletons that all try to give the user the sensation or experience of physically interacting with some kind of virtual or remote environment.

The other wording is magnetic levitation. How does that work? You have a stator and flotor?

Berkelman: Yes. Flotor is a sort of coined term in this maglev idea. Instead of having a rotor in a motor that just spins, we have a flotor that actually moves and rotates in a full six degrees of freedom. So that's three degrees of translation and three degrees of rotation. The stator is the stationary base and the flotor is the levitated object (Figure 1).

Figure 1. The magnetic levitation haptic device is composed of a flotor and two stators. It is an interface device that provides a feel for reaction forces based on a freefloating flotor suspended due to Lorentz forces. (a) The device is embedded in a desktop-height enclosure so that the user can easily grasp the handle, which is attached to the flotor, while resting the wrist on the rounded top of the stator. The magnet assemblies are visible on the inner stator. (b) The circular stator rim is normally mounted to the top panel of the enclosure. The flotor handle is visible in the center and the magnet assemblies can be seen on the inner and outer stator bowls. One of the position sensor assemblies can be seen in the lower left of the picture.

Hollis: Having the user interact with a single, low mass, moving part that is free of friction and backlash will improve the fidelity of the haptic experience.

Which one has the coils of wire?

Berkelman: We put the coils on the flotor because they're lighter. So it's easier to levitate. However, we do need to run lightweight, flexible cabling to the flotor. So, it's practically contactless.

What are the cables for? Power?

Berkelman: And that's only to supply the current to the coils and also to the LEDs, which are used for position sensing.

An alternator and motor are based on Faraday's law. What are you basing your levitation on?

Hollis: Our device is based on a new form of magnetic levitation invented at IBM Research in 1987. It uses Lorentz forces rather than the usual Maxwell forces found in more familiar devices such as magnetic bearings (Figure 2).

Figure 2. Lorentz force actuation. Fixed magnet assemblies produce a magnetic field B so that when a flat, wound coil with current i is between them, a force f is produced in the direction shown.

You have the coil on the flotor. What's on the stator then?

Berkelman: The stator contains the permanent magnet assemblies. We have an inner stator and an outer stator with permanent magnet assemblies fixed on them (Figures 3 and 4). Both stators are fixed. During levitation, the coils on the flotor are suspended between the opposing magnet assemblies (Figure 2).

Figure 3. To generate arbitrary forces and torques in all directions, a total of 6 actuators are used in a tightly packed configuration. The magnet assemblies are mounted on inner and outer stator hemispheres (not shown) and the coils are embedded in a single hemispherical flotor shell (not shown).

How does the Lorentz force interaction come in here now?

Berkelman: The Lorentz force gives you a force that is the cross product of the current and the magnetic field. We've got a current that is in the plane of the coils and the magnetic field is perpendicular to the flat coils. So, the force that is generated from each combination of magnets and coils is in the plane of the coil and perpendicular to its longer axis (Figure 2).

I can visualize the translation forces, but what about the torque?

Hollis: There is a total of six Lorentz actuators arranged so that the sum of the forces from the six actuators produces an arbitrary force-torque wrench on the flotor (Figure 3).

Figure 4. Cutaway half-section of Lorentz magnetic levitation haptic interface device. The stator bowls enclose the flotor shell. A hole in the inner stator allows for the flotor interaction handle and a hole in the outer stator provides access for the flotor wiring to the current amplifiers.

Tell me about your optical sensors.

Berkelman: We get sensing from a combination of infrared LEDs and position sensing photodiodes. The LEDs are embedded in the flotor; they move. We have lenses that focus the light from each LED onto its position sensor. These sensors are fixed on the stator. With a combination of three planar sensors and the three LEDs (Figure 5a), we can calculate the position and orientation of the flotor. Each sensor gives you x and y, so from the six sensor signals you can calculate the six-element flotor position-orientation vector from that. The lenses on the sensor assemblies (Figure 5b) demagnify the motion of the LEDs by a factor of 2.5.

Figure 5a. Position sensing configuration: The three sensors are mutually orthogonal and each faces an LED on the flotor. Each sensor is a planar position sensing photodiode that measures the position of the LED light spot focused on it by a lens. The position and orientation of the flotor can be calculated from the combination of the six sensor signals (x and y LED spot position from each sensor).

Figure 5b. Position sensor assembly: Each sensor is mounted in a cylindrical can to shut out any external light. The position of the LED spot on the sensor depends on the angle of the LED from the axis of the lens and sensor centers. The motion of the LED at its average distance is demagnified by a factor of 2.5 by the lens.

Of what use is this?

Hollis: This allows the flotor a large motion range while keeping the LED light spots imaged on the position sensing photodiodes. The sensed position-orientation information is used by the control algorithm to provide stable levitation and to impart the correct forces and torques to the user's hand.

What are some of the possible applications?

Berkelman: Any application that involves a user physically interacting with some sort of simulated or virtual environment. The virtual environment could be a CAD (computer-aided design) system where you're trying to fit parts together, feel how they fit or slide or don't fit together, how much clearance, how much wiggle one part has with another. Another good one would be simulation of medical procedures. We could simulate what it feels like to grasp a scalpel and feel it cutting bone or tendons or other kinds of internal tissues. This is a general purpose device for these kinds of interactions. What you're interacting with would depend on the application software, whether it's mechanical parts or a model of a human body, etc.

So, you would need specific application software, for instance, to tell you how much friction there is? How much resistance, how much inertia?

Berkelman: Yes. We'd have physical simulation running on a different computer. Then, the controller of the device would interact with that physical simulation to reproduce those forces and sensations that you actually feel when, for example, you're turning a screw.

Hollis: Right. These kinds of systems have the real potential for changing the way people do certain kinds of jobs. For example, there will likely be significant impact in areas such as training and design.

Berkelman: Somebody would have to build a physical model or a physical simulation of that process. These sorts of operations are possible using things such as a data glove or some kind of instrumented sensing glove, but what our device does is it pushes back on you so you actually feel reactions.

How accurate is that? What kind of sensitivity?

Berkelman: The accuracy and sensitivity of the device is where the magnetic levitation comes in. Because we don't have linkages or transmissions of any kind that would introduce nonlinearities in the actuation, we can be very sensitive, accurate, and achieve very high position and force bandwidths.

So, we should be able to get the position sensing to the micron level and the device should be capable of position control bandwidths of a hundred Hz or so. That gives you the ability to feel the subtle surface texture, friction, and other effects that you wouldn't be able to feel with some kind of larger or slower more traditional mechanical device.

I don't understand how bandwidth applies here.

Berkelman: If you feel a rough or complicated surface or interact with a stiff object, you get a sort of rich pattern of sensation in your fingertips. If you're dealing with a rigid object, then a collision with that object would generate a very sharp force impulse. In order to reproduce that sort of immediate responsiveness, we need a high-bandwidth device. A low-bandwidth device would feel sluggish, spongy, or soft.

We're really trying to reproduce the behavior of physical objects to the highest fidelity possible.

So, the bandwidth response is felt through the flotor?

Berkelman: Yes. The forces exerted on the flotor are what is felt by the user. The actual device control rate is going to be at least 1000 Hz and the actual response bandwidth depends on the sensitivity of the sensors, how well the controller is tuned, the inertia of the flotor, and other factors. But you should be able to feel any changes at 100 Hz.

Hollis: In the next year or so we will be attempting to optimize the control strategy and accurately characterize the performance. This will include feed-forward terms that partially account for residual actuator nonlinearities, a new iterative scheme for solving the real-time kinematics solution worked out by Stella Yu here at CMU, and implementation of a new compliance control scheme proposed by Ernest Fasse at the University of Arizona.

What's the weight of that flotor?

Berkelman: The current flotor (Figure 6) is about 800 grams. We're planning to build another flotor that weighs half that by substituting aluminum coils for the copper coils.

What's the advantage? Lighter weight?

Figure 6. The flotor. The actuator coils are visible on the inside of the flotor shell and the flotor handle is mounted in the center.

Berkelman: Yes, much lighter.

What about current carrying ability or generation?

Berkelman: The resistance of the aluminum coils will be greater, but we found that the actuators we have now are efficient enough that heat dissipation really isn't a problem. We can support the flotor we have now with less than 15 watts of power. Because that 15 watts is dissipated over a large hemisphere, we don't really have to worry about overheating. With aluminum coils we'll probably need less than half of that power to support the flotor because of the reduced weight.

What of the future?

Hollis: We're pretty excited about the possibilities.

Berkelman: We think it's a technology, an approach, that really has a lot of potential for any kind of tool-based interaction simulation. We want to build several more of these devices and go into the application/simulation aspect of it more.

What about the virtual environment of video games?

Berkelman: It would be very dramatic in video games. For this to happen on a wide scale, however, the cost would have to be reduced. Meanwhile, research or training applications such as driving or flying virtual vehicles are a distinct possibility.


Peter Berkelman is currently a robotics PhD student at Carnegie Mellon University. His research consists of development of magnetic levitation haptic interface devices and real-time haptic simulation of dynamic environments. He received his BS and MS degrees in Mechanical Engineering from Massachusetts Institute of Technology in 1992. He has also completed research internships in robotics with Philips Laboratories, Fujitsu Laboratories, and the Toshiba Manufacturing Engineering Laboratory.

Ralph L. Hollis is a Principal Research Scientist in the Robotics Institute at Carnegie Mellon University. He received his BS and MS degrees in physics from Kansas State University, and his PhD in solid state physics from the University of Colorado. Before joining CMU, he had 20 years of experience at the IBM Thomas J. Watson Research Center and at North American Aviation. He is a member of the American Physical Society and a Senior Member of IEEE, and has served on numerous government panels and on the editorial boards of two journals. His current research centers on haptic interfaces and architectures for rapidly reconfigurable assembly systems.

They were interviewed by Frederick Su.