An autonomous security drone system with real-time perception, LLM-powered reasoning, and a native iOS companion — built to think, see, and act in the physical world.
DARREN is a fully autonomous embodied AI system — a drone that doesn't just fly, but thinks, perceives, and reasons about its environment in real-time.
Unlike conventional drones that require constant human input, DARREN operates with a multi-layered intelligence stack: a YOLOv8 perception pipeline for object detection, an LLM brain (powered by Ollama) for natural language understanding and contextual reasoning, and an ArduPilot flight controller for precise autonomous navigation.
Everything talks through ROS 2, orchestrating 10+ concurrent nodes that handle vision, speech, emotion, navigation, and safety — all running locally with no cloud dependency.
Real-time object detection via YOLOv8 with class tracking, confidence scoring, and threat assessment through a multi-layer perception pipeline.
GPS-guided waypoint navigation, patrol patterns, geofenced boundaries, and autonomous return-to-home with ArduPilot-SITL integration.
Local Ollama-powered reasoning engine with 3-step response chain: emergency bypass → LLM inference → keyword fallback. Full contextual conversation memory.
Native SwiftUI app with live video feed, thought panel showing reasoning process, emotion indicators, and natural language voice commands.
Personality-driven speech synthesis with 5 emotional states, real-time facial expression mapping, and context-aware response generation.
Emergency keyword bypass for instant response, geofence enforcement, battery monitoring, automatic RTL triggers, and hardware-level safety interlocks.