A compact tabletop humanoid torso designed for natural social interaction, expressive gesturing, and active desk assistance. Built on an acoustically silent differential silicone drive with hardwired ToF cliff safety and hybrid edge-cloud AI.
Every mechanical and electrical system in DeskBot is optimized for workspace constraints — acoustic silence, absolute fall protection, and zero-latency reflexes.
DeskBot traverses desktop surfaces using an acoustically silent, high-traction differential silicone drive base with a low-friction rear ball caster. Four downward-pointing Time-of-Flight (ToF) sensors trigger hardwired motor cutoffs in under 5ms, ensuring tabletop edge safety.
A 2-DOF Pan/Tilt neck delivers full 360° social gaze tracking to maintain natural eye contact. Dual expressive limbs are built upon the Agazie Modular Joint (AMJ2) dual-bearing standard for smooth, compliant, and durable physical interaction.
An ESP32-S3 executes deterministic low-level motor loops, while an onboard Raspberry Pi 5 runs edge vision and audio processing, linked to streaming multimodal LLMs for deep conversational reasoning without sacrificing reflex speed.
DeskBot decouples low-level deterministic motor reflexes from high-level cognitive intelligence — creating a responsive tabletop companion that understands context, reacts in real time, and collaborates seamlessly at your desk.
