Plexus is a camera-on-wheels Unmanned Ground Vehicle (UGV) that autonomously learns to navigate indoor environments using neural networks trained on the user's own data. The robot consists of a WiFi-enabled ESP32 microcontroller, a camera, two differential-drive motors, and a minimalist chassis. All neural computation runs on the user's existing PC GPU over WiFi - an architecture we call "Disaggregated Intelligence" - which keeps the robot hardware radically simple and affordable. The robot captures camera frames, streams them to the user's laptop, where a neural encoder builds a learned spatial map and a neural policy navigates toward user-specified goals. This pipeline currently achieves 87% autonomous goal-navigation success in real homes. Visit embrained.ai for more information.
