The Machine Learning Team develops the perception and decision-making models that allows our aircraft to autonomously interpret complex environments. By transforming raw sensor data into actionable information, the team enables reliable object detection, classification, and mission-aware responses in dynamic conditions.
The team designs and trains models using Python and frameworks such as PyTorch to process visual and sensor inputs from onboard systems. This includes object detection, target identification, and environmental awareness necessary for autonomous navigation and task execution. Because competition environments are unpredictable, models are built to remain robust under variable lighting, motion blur, and shifting terrain conditions.
Machine Learning works closely with Autopilot, Airdrop, and Hardware to ensure models operate efficiently within onboard compute limits. Accuracy, latency, and system reliability are balanced through iterative testing in simulation and live flight environments. The objective is not only detection performance in isolation, but dependable real-world autonomy.
