Is it possible to consider such an aircraft merely a “flying camera” if it is making instant decisions on its own? Such an understanding is quickly becoming outdated throughout the commercial UAV market. A new breed of drones is on the cusp of becoming intelligent aerial robots—that is, machines not only capable of capturing images with extremely high resolution but also interpreting such images instantaneously.
“Modern drones are defined by what they do rather than how they fly: their payload capabilities, onboard intelligence, and data processing efficiency,” says Vladimir Spinko, the founder of Aery Bizkaia, observing this transition in the design of his AI-powered radar and sensor solutions for surveying by drone and the detection of humanitarian mines. This is further fueled by developments in semiconductors, such as the CV5 AI-enabled System on a Chip by Ambarella, now in use in the Antigravity A1 360 drone.
“Drones are quickly transforming into intelligent edge aerial robots, capturing, processing and understanding the world in real time,” states Fermi Wang, Ambarella’s CEO. Michael Shabu from Antigravity Drone Technology further adds, With this functionality, our drones are able to ‘provide intelligent functionality at the edge while remaining focused on efficiency and performance.’ Ambarella’s A1 features upgraded infrared sensors for “deliver intelligent capabilities at the edge while staying focused on efficiency and performance,” while its AI core is responsible for detection, recognition, and decision-making.
Such a paradigm shift in the intelligence available on-board these vehicles has altogether changed the way drones were being applied across sectors before. In infrastructure analysis tasks, for instance, drones with SoCs integrated with AI can automatically detect defects on a bridge or a tower and point them out to the concerned team for repairs. In the area of energy provision, analysis of wind turbines or transmission cables can detect potential disruptions before they happen and shut them down before a blackout results. In public safety measures, analysis of thermal and optical images enables drones to pinpoint a possible location of a missing person or a risky area without putting human lives at risk.
Precision agriculture remains one of the most attractive application areas in this development. Intelligent drones with multispectral and thermal cameras are capable of scouting entire fields in one pass and even identifying early indicators of plant disease or dehydration. The ZenaDrone 1000 is an example of platforms merging AI and machine-learning algorithms in order to label plant sexes in their application in cannabis plant farming and monitor livestock location through GPS-temperature signature mapping. Even weather effects would be predicted based on environmental measurement tools carried by these platforms.
From a purely engineering perspective, the CV5’s edge AI technology is in line with trends in autonomous robotics in aerial platforms. Real-time processing enables drones to dynamically alter flight plans in response to sensor data input and undertake beyond visual line-of-sight flight in challenging environments. Such is necessary in survey or observation missions, such as scanning rough terrains or observing offshore platforms, in scenarios with limited connectivity and in which human observation is impractical or impossible.
The inclusion of AI in drone hardware also provides the possibility for multi-role UAVs. With a single drone capable of performing film imaging, industry inspection, and agricultural spraying with a mere change in the modules, the AI included in the CV5 will be able to change parameters based on the operation. This will be facilitated by the CV5 having the capabilities to deal with different data streams, ranging from HD video to multi-spectral analysis.
So long as the edge AI semiconductor technology keeps improving at a rapid pace, the division line between the airborne sensor and the self-contained robot will become less clear. Drones carrying intelligent SoCs, starting from the Ambarella CV5 chip, are more than just passive data-acquiring tools. They become fully-fledged participants of industrial processes who can decide on their actions immediately and wisely. For aeronautical engineers, robotics AI experts, and all drone enthusiasts, it means that the future will define drones less by flight dynamics and more by the level of intelligence and complexity of tasks they perform.