Botronka – Social Companion Robot
My BSc thesis: a Raspberry Pi companion robot that sees, listens, talks, and moves — vision, audio, and an LLM fused in one event-driven system.
The Problem
A social robot needs to perceive who it is talking to, hold a natural conversation, and act safely — all on constrained embedded hardware.
Approach
Designed a modular, thread-based architecture where workers publish and consume messages over a shared queue: face detection and identification with per-person trust levels, VAD-gated speech recording and transcription, an LLM agent deciding replies and actions, TTS output, and trust-gated motion control with ultrasound distance sensing.
Results
A working prototype that recognizes people, lets the owner register new faces, converses naturally, and executes safe movement commands — with OLED and buzzer feedback.