Octopus-Inspired Embodied Tactile Intelligence: A Universally-Calibrated Miniature Vision-Based Suction Cup
We developed a miniaturized, vision-based suction cup that gives robotic manipulators a sense of touch — enabling autonomous alignment and the safe grasping of delicate, misaligned objects.
Here are our main contributions:
- Miniaturization: Shrank the system to a compact 22 × 40 mm footprint by integrating a micro-camera.
- Simultaneous Multimodal Sensing: Real-time detection of 3-axis pressing forces and contact orientation for closed-loop surface alignment.
- Universal Calibration: Our neural network applies universally to identically designed suction cups, eliminating the need for per-unit recalibration.
- Enhanced Mechanical Performance: Achieved 36.3 kPa adhesion with just a 2.3 N preload to safely handle delicate, curved objects.