Real-Time Atmospheric Risk Detection Using Embedded Machine Learning with Low-Cost Aircraft-Based Sensors

Keywords

machine learning, embedded systems, Atmospheric sensing, sensor fusion, embedded systems

Presenter Abstract

Rapidly shifting atmospheric phenomena, such as microbursts and clear-air turbulence, pose significant risks to low-altitude flight operations. These risks are pronounced for unmanned aircraft systems (UAS) and general aviation aircraft, including fixed-wing platforms and turboprop aircraft. Current forecasting systems usually lack the spatial and temporal resolution to detect localized threats in real time. This research focuses on the design and development of an aircraft-mounted atmospheric sensing system. It uses machine learning to classify flight risks. The system combines barometric pressure, temperature, and humidity data from a DPS310 pressure sensor and a BME180 environmental sensor. An embedded computer allows continuous in-flight data collection. Flight testing is conducted on a Bushmaster UAS platform. This setup provides real-time atmospheric data at various altitudes and flight conditions. The system computes derived properties, such as pressure gradients, temperature rate of change, and humidity trends, in real time. This enables incremental training of machine learning models during flight. The model runs autonomously on the embedded device. It updates and categorizes atmospheric data by risk level. No external communication or cloud processing is needed. Data is collected concurrently for post-flight verification and performance evaluation. Preliminary testing evaluates system responsiveness, repeatability, and susceptibility to transitory atmospheric fluctuations in controlled and field flight environments. This technology shows that combining low-cost sensors with embedded machine learning increases situational awareness for unmanned aerial systems and traditional aviation. The system offers a scalable way to provide real-time atmospheric intelligence. It enables safer, more autonomous flight in areas lacking traditional meteorological data sources.

Presentations

Presented in Session 6: Sensors II

Share

COinS
 

Real-Time Atmospheric Risk Detection Using Embedded Machine Learning with Low-Cost Aircraft-Based Sensors

Rapidly shifting atmospheric phenomena, such as microbursts and clear-air turbulence, pose significant risks to low-altitude flight operations. These risks are pronounced for unmanned aircraft systems (UAS) and general aviation aircraft, including fixed-wing platforms and turboprop aircraft. Current forecasting systems usually lack the spatial and temporal resolution to detect localized threats in real time. This research focuses on the design and development of an aircraft-mounted atmospheric sensing system. It uses machine learning to classify flight risks. The system combines barometric pressure, temperature, and humidity data from a DPS310 pressure sensor and a BME180 environmental sensor. An embedded computer allows continuous in-flight data collection. Flight testing is conducted on a Bushmaster UAS platform. This setup provides real-time atmospheric data at various altitudes and flight conditions. The system computes derived properties, such as pressure gradients, temperature rate of change, and humidity trends, in real time. This enables incremental training of machine learning models during flight. The model runs autonomously on the embedded device. It updates and categorizes atmospheric data by risk level. No external communication or cloud processing is needed. Data is collected concurrently for post-flight verification and performance evaluation. Preliminary testing evaluates system responsiveness, repeatability, and susceptibility to transitory atmospheric fluctuations in controlled and field flight environments. This technology shows that combining low-cost sensors with embedded machine learning increases situational awareness for unmanned aerial systems and traditional aviation. The system offers a scalable way to provide real-time atmospheric intelligence. It enables safer, more autonomous flight in areas lacking traditional meteorological data sources.