Drones Everywhere in the Cloud

Pilots must continue to maintain full control of the drone despite the autonomous nature of these BVLOS flights, for liability reasons, similar to that of the watchful human at the driverless car’s driver seat. In high-risk conditions, such as operating in urban Singapore, the system must achieve 10-7/ flight hour safety standards, that is, only 1 catastrophic failure for every 10 million flights.

As such, it is critical to have first-person-view (FPV) video streaming live to a drone operations center (DOC), where drones are operated. To avoid the massive investment in another nationwide RF network for drones, many manufacturers including GR are utilizing cellular networks as the primary network for BVLOS use cases. We host media servers in the cloud to record the video stream, and re-broadcast it to various consumers beyond the pilot.

4. 5G Network Slicing

The current generation of public cellular networks cannot support mission critical operations such as aviation, due to the lack of service level guarantees(SLG).

In the impending 5G rollout, one new feature is network slicing—the ability for a prioritized SIM to temporarily reserve a dedicated frequency spectrum, avoiding congestion originating from consumer devices. This makes connection to the cloud even more reliable than connecting via Wi-Fi from a GCS, as it avoids competing for use of spectrum from ubiquitous Wi-Fi devices.

To make this reliability even more apparent, Garuda CoPilot—our on-board companion system with LTE— is designed to have two or more LTE connections from diversified telcos, including private LTE base stations were available. We also secure the link with the latest cybersecurity technologies, such as having physical, secure elements on drones for authentication and encryption.

5. Artificial Intelligence (AI)

The cloud has unlimited scale when it comes to compute resources, power that a drone does not have on-board. By connecting drone sensors, video, and pilot commands to a cloud infrastructure, we can execute compute intensive machine learning (ML) and computer vision (CV) algorithms to help with the drone’s navigation or mission goals.

Although the primary use cases for ML/CV now are navigation and surveillance, the potential is endless. For example, GR is heavily involved with vision-based analytics in agriculture setting where we census crops and detect issues. Maybe one day, drone AI will verify your identity by scanning your face before lowering the pizza you ordered to you.

Cloud Players in the Drone Market

A brave new world is emerging at the confluence of aviation, infocomm, and public safety, one that promises better utilization of the 3rd dimension of space above our cities, but also one that challenges many status quos.

End Users of drone operation platforms, such as government agencies, facility managers, logistics companies, and security agencies, should not shy away from investing in full end-to-end trials, as part of the learning journey. Existing SOP should not get in the way of studying new drone centric concept-of-operations.

On the other hand, many traditional IT businesses are looking to increase their customer’s productivity with workflow built around drones and robotics, while retaining the cloud investments made earlier. Garuda Robotics hopes to help make that transition easier by providing a BVLOS Platform-as-a-Service (PaaS), drone test sites in Singapore, as well as ready-to-fly products such as Plex Horizon, a flight control system for BVLOS operations.

These pre-made software applications and components are, however, no match for a timely investment to fly, to learn to fly from professionals, to see the different points of view from manned aviation and regulators, and to build solutions from the perspective of the operator. Safety first.