Shoggoth Mini: An Expressive Soft Tentacle Robot

This post details the creation of Shoggoth Mini, a soft tentacle robot designed for expressiveness. The author iteratively improved the hardware, solving cable tangling issues and adding calibration scripts. Control is achieved through a simple 2D mapping for intuitive manipulation, combined with GPT-4 and reinforcement learning for various control modes, including manual control, vision-based tracking, and open-loop behaviors. The author explores the relationship between expressiveness and the perception of 'aliveness' in robots, concluding with future research directions.