Unmanned aerial vehicles are widely used in mapping, aerial photography, rescue operations, shipping, law enforcement, agriculture, among other things. In addition, drones hold considerable potential in relation to human exploration, enabling humans to take a closer look into areas that are far to dangerous for people to visit. Behind this is better built and controlled drones, something made possible through advances with machine learning. Three recent research examples are considered.
Drones to monitor volcanoes
An area where it is dangerous for humans to move too close is within the vicinity of a volcano. A new development has shown how drones can monitor explosive volcanoes, protecting humans and at the same time gathering valuable data. To this end, researchers have created novel 3D models with centimeter resolution. These devices have been tested out close to a volcano in Guatemala.
To this end, scientists from GFZ GeoForschungsZentrum Potsdam, Helmholtz Centre, have successfully deployed a drone for a series of repeated survey flights with optical and thermal imaging cameras. The drone took the images around the Santa Maria volcano in Guatemala. The biggest challenge is with taking the various 3D images and putting them together to create a holistic overview of the volcanic activity, so that meaningful inferences can be drawn.
Based on the success of the study, it is hoped that a regular and systematic survey of dangerous volcanoes with drones will become part of regular exploratory activities.
The development of the 3D imaging drone has been published in the journal Scientific Reports. The supporting article is titled “UAS-based tracking of the Santiaguito Lava Dome, Guatemala.”
Drones used to forecast volcanic eruptions
Building on the above news, it has been reported from University College London that especially-adapted drones have been used to gather data from never-before-explored volcanoes. The objective here is to assist local communities in vulnerable areas to forecast future eruptions. This is based on tests conducted at Manam volcano in Papua New Guinea.
In addition, this type of research has the benefit of showing the extent that volcanoes contribute to the global carbon cycle, which is key to sustaining life on Earth. This has been made possible by incorporating miniaturised gas sensors, spectrometers and sampling devices into the drone structure.
AI drives drones over rough terrain
One problem with drones is that either human intervention is required (something that becomes problematic at distance) or rudimentary computers try to control the drone. This latter activity is not always successful and crashes often result. To make drone operation more efficient for exploratory purposes, researchers have been developing suitable artificial intelligence applications.
Through the application of machine learning, drones are now able to scan terrain and excavations without human intervention. A recent study has shown how artificial intelligence to take over drones. The algorithm developed id capable of predicting and counteracting the wind forces acting on the drone body, thereby minimizing the risk of drone damage. The designed Gaussian process model behind the algorithm allows wind encountered to be compensated for.
The study comes from Aarhus University.
