Open-source tactile/shear telemetry + contact flags / grip scoring for grippers (SO-ARM101 + Teensy)

Developer 0o8o0-blip has released an early-stage, open-source project called 'tactile_kit' designed to add sophisticated tactile sensing to robotic grippers. The hardware setup pairs an SO-ARM101-style gripper—fitted with custom TPU (Thermoplastic Polyurethane) pads for compliance and sensitivity—with a Teensy 4.1 microcontroller. This combination streams high-frequency 3-axis tactile and shear force data, capturing not just contact but also directional slip forces, which are critical for dexterous manipulation.

The core innovation is in the software stack, which processes this raw sensor data into actionable insights. It outputs two primary data streams: a raw CSV of sensor readings and an annotated CSV. The annotations automatically generate simple contact flags, a heuristic 'grip_score' that quantifies grasp stability, and a 'baseline_quality' metric. This transforms raw telemetry into structured labels that can be directly used as reward signals, termination conditions, or phase indicators in reinforcement learning (RL) training pipelines, significantly lowering the barrier to integrating touch into AI-driven robotic control.

The project explicitly aims to decouple tactile sensing from complex motor control integration, making it a standalone tool for debugging and benchmarking gripper performance. By providing a standardized way to score grasps and flag contact events, it allows researchers and developers to quantitatively compare different gripper designs, control policies, and object manipulation strategies. The open-source release includes a demo GIF showing a full grip cycle, illustrating how the metrics change during approach, contact, hold, and release phases.