Paper Reading #15: Madgets: Actuating Widgets on Interactive Tabletops

Reference Information
Madgets: Actuating Widgets on Interactive Tabletops
Malte Weiss, Florian Schwarz, Simon Jakubowski, Jan Borchers
Presented at UIST'10, October 3-6, 2010, New York, New York, USA

Author Bios

• Malte Weiss is a PhD student in the Media Computing Group at RWTH Aachen University. His work focuses on interactive surfaces and tangible user interfaces.

• Florian Schwarz is a Diploma Thesis student in the Media Computing Group at RWTH Aachen University. His work is in interactive tabletop computing.

• Simon Jakubowski is a student assistant in the Media Computing Group at RWTH Aachen University. He is working on two other projects besides this one.

• Jan Borchers is the head of the Media Computing Group at RWTH Aachen University. He holds a PhD from Darmstadt University of Technology in Computer Science and now explores HCI.

Summary
Hypothesis
How can we create tangible magnetic widgets for use on tabletop computers?

Methods
The authors created a prototype of the Madgets system and various sample controls for it, including radio buttons and slider knobs. They considered general purpose widgets, height, force feedback, water wheel Madgets (which transfer energy), and mechanical audio feedback as possibilities.

Results
General purpose widgets provide both haptic and visual feedback. The physical configuration can be saved for later use. Multiple users can also use the system since the system is actuated by electromagnets, which enables a tangible presence of another user. They also support ad-hoc usage.

Height is a possible feature, as the electromagnets can keep a Madget in place while lifting parts of it, as in a radio button. A user can feel both the shape of the button and its current state. A clutch control can lock or unlock moving parts to disable it, similar to current GUIs "graying out" and option.

Force feedback can be generated through resistance to the moving of a part of a Madget. The algorithm used allows for Madgets to vibrate and also create dynamic notches when a user reaches a certain step in a scale

The water wheels transfer energy from the table to the Madget. Inductive energy transfer, performed through plates, allows for power to be sent to a Madget without additional components. Motors work through a rotational actuation of a part and can be combined to produce a more complex system.

Mechanical audio feedback can occur through a magnetic pulse that triggers a noise of some sort.

Prototypes are quickly producible and do not take much time to program into the system with dynamic mappings.

Contents
The Madgets created by the authors are translucent, tangible widgets that resemble common controls like sliders and are relevant for general usage. They have magnets attached to them that can be actuated independently through an array of electromagnets. Actuated tangibles include moving the Madget across the surface, and force feedback. They are low-power and low-cost. The devices are unpowered and passive, which makes them easy to produce and hide the underlying technology from the user. Controls can be relabeled dynamically.

The sensing technique requires uniform backlighting, provided by an electroluminescent foil. An array of electromagnets, controlled by an Arduino through shields that provide various output channels to generate pulse width modification, actuates objects separately. To track the physical devices, a visual sensing technique, which does not interfere with the electromagnets, was used. Diffused Surface Illumination detects touch events but is also precise enough to differentiate Madgets from fingers. The controls are illuminated through the LCD, so that they are dynamically labeled. They are mounted on cylindrical markers that are used to determine where the Madgets are and encode the type of Madget. Moving parts also have a marker. Permanent magnets are attached for actuation. Each rigid body in a Madget can have different actuation forces applied to its magnets tangentially or normally. The forces are computed to move a permanent magnet a certain distance based on polarization and the position of the magnet. Linear optimization is used to resolve their computations. The Coin-or programming library minimizes the desired function, and thus the total force, power and heat production. The weights are applied dynamically to optimize performance and reliability with a frame rate of 30fps. Overheating electromagnets are considered weightier than others to reduce their usage. The widgets have gradient fiducials to increase the resolution of each sensing dot based on the radius of the object, detected by its brightness. The table is pre-calibrated for this with no objects, a white sheet, and then a dark gray sheet successively placed on it to determine the thresholds. The algorithm can identify multiple touches and dragging gestures. It first detects widget footprints and then focuses on remaining input.

Discussion
The authors created a system that features magnetic widgets as a new, cheap form of interaction. While their results certainly suggest that more work in this field is viable, the lack of user testing concerned me. I don't doubt that users could quickly adapt to using the devices as an extension of touchscreens, but the effect of automatic actuation was untested, which leaves me concerned about the effectiveness of the system.

The authors proposed that work should be done in avoiding Madget collisions, which struck me as a very good idea. While a good interface should mean that this situation would never occur, designs should always assume that some error will be made. If the system inherently tries to avoid crashing Madgets, then it is far less likely that such a problem will occur.

I was particularly intrigued by the sheer cheapness of producing one's own widgets, especially with a 3D printer available. The system was specifically built to allow the average designer to easily program any device into the system, which increases its accessibility that much more.