MagBall: Magnetic Rollerball for Multi-Scale Contact Interactions on Diverse Surfaces

Paper Title

MagBall: Magnetic Rollerball for Multi-Scale Contact Interactions on Diverse Surfaces

Publication Info

• Topic area: Tangible input devices for human-computer interaction
• Keywords: Tangible input, magnetic sensing, Hall-effect sensors, rolling mechanism, displacement estimation, force sensing, surface-independent interaction, wearable devices, stylus pen, smart tools

Background and Problem

• Problem / challenge: Existing input devices face a trade-off between compactness and spatial expressivity. Point-contact devices are compact but limited in spatial input, while surface-based devices require large, instrumented surfaces, increasing size and complexity.
• Significance: Addressing this trade-off enables more versatile and portable input devices, expanding interaction possibilities across diverse surfaces and applications.
• Motivation and related work: Prior work includes capacitive and magnetic input devices, trackballs, and optical-flow-based pens. However, these systems are constrained by surface requirements, limited spatial expressivity, or inability to resolve yaw rotation. This paper aims to overcome these limitations with a compact, surface-independent solution.

Solution

• Proposed approach: MagBall, a magnetic-ball sensor that uses a rolling magnet-embedded ball and Hall-effect sensors to estimate displacement and force across diverse surfaces.
• Novelty:
  1. Compact design (13 × 13 × 10 mm) enabling multi-scale, surface-independent interactions.
  2. Integration of asymmetric magnetic fields for yaw compensation and robust sensing.
  3. Machine learning pipeline for real-time displacement and force estimation.
  4. Demonstration of applications in stylus pens, wearable trackballs, and smart massage tools.
• Procedure and key techniques:
  • Magnetic ball with embedded magnets rotates over a Hall-effect sensor array.
  • Machine learning models (LSTM, ETR) process sensor data to estimate displacement and force.
  • Design optimization through simulation of magnetic fields and evaluation of ambiguity, rotational sensitivity, and force sensitivity.

Results

• Concrete findings:
  • Displacement estimation RMSE: 0.15 mm; long-term drift: 2.31 mm/s.
  • Force estimation RMSE: 0.67 N.
  • Real-time inference latency: 4.9 ms.
• Advantage over baselines:
  • Operates on diverse surfaces (e.g., glass, metal, fabric, skin) without additional instrumentation.
  • Resolves yaw rotation, unlike traditional trackballs.
  • Outperforms optical-flow pens on transparent and soft surfaces.
• Experiments / evaluation:
  • Validation of sensing performance through simulation and experiments.
  • User study comparing MagBall to commercial devices (stylus pen, mouse, trackpad) on various surfaces.
  • Chamfer distance analysis for path accuracy across six surface types.
• Limitations and future work:
  • Susceptible to strong magnetic interference.
  • Long-term drift due to lack of absolute position reference.
  • No off-contact tracking; potential solution includes integrating inertial measurement units.
  • Future directions: wireless communication, MagBall arrays for robotic skin, computational design optimization.

Summary

MagBall is a compact, magnetic-ball-based input device that bridges the gap between point-contact and surface-based systems by enabling multi-scale, surface-independent interactions. It achieves sub-millimeter displacement precision and robust force sensing using a machine learning pipeline. Applications include a stylus pen, wearable trackball, and smart massage tool, demonstrating versatility across diverse surfaces. While limitations like magnetic interference and long-term drift remain, future enhancements such as wireless communication and off-contact tracking could further expand its applicability in human-computer interaction and robotics.