The ASTRO project advances technology to build a platform for autonomous data-driven mobile sensing via networked drones. We target proof-of-concept demonstrations for a diverse set of unprecedented sensing capabilities including (i) high-resolution distributed mobile laser-spectroscopy gas sensing to identify, localize, and track health and environmental hazards in real-time and (ii) automated mobile radio-frequency spectrum analysis and usage via distributed diverse-spectrum virtual arrays.

The main objective of ASTRO is to realize high-resolution pollution sensing by exploiting the ability of our drones platform to dynamically move sensors in 3-D according to real-time measurements. We mainly target the detection, tracking, and modelling of high VOC concentration levels following extreme events.

We are designing ML algorithms for tracking VOC plumes, which will enable the creation of a first-of-its-kind warning system, that can identify hazardous areas and alert residents as those can rapidly change. Moreover, data collected from plume tracking will yield a deeper understanding of the dynamics, reactivity, and evolution of individual VOCs in the atmosphere.

ASTRO has the following unique features. We realize networked drones that communicate and coordinate among themselves. They form a dynamic mesh and employ software defined radios (SDRs) to adapt carrier frequency in order to realize longer range as needed to maintain connectivity with other drones, at the potential cost of less bandwidth being available at lower frequencies. In addition, ASTRO is tetherless, it does not employ ground control stations for sending and receiving control signals and/or data. As a consequence of tetherless operation, ASTRO drones do not require any communication infrastructure, enabling flight in areas not served by Wi-Fi or cellular networks. Moreover, ASTRO realizes data-driven sensing missions via online (without prior training) light-weight on-drone machine learning. Drone’s decisions and flight paths are adapted in real-time according to measured sensor data.