Drone Technology: Today's Trends and What's Next

Ease of use, few mechanical parts, and the general public's acceptance make this platform the most widely used. The compactness of the newest quadcopters is coupled with impressive sensor packages. Now, every job in the field can carry a drone in its toolbox. The same advancements that make drones more compact allow larger multirotors to fly longer and carry more. We're getting closer to that field-demonstrated one-hour flight time for electric multirotors with a payload, with larger aircraft reaching 40–45 minutes regularly.

If range is the priority, wings work best. When you combine the simplicity of an airplane with the complexity and utility of the multirotor, you get VTOL platforms. These combine the flight time of a fixed-wing aircraft with the simplicity of taking off and landing like a multirotor. For example, the Quantum-Systems Vector is a large airframe with a significant payload capacity that converts to a VTOL platform by rotating the motors on the wings to transition between vertical and forward flight. And the WingtraOne Gen II, which takes off and lands on its tail, flips to horizontal once it has reached the desired altitude. Regardless of the VTOL design, users love the benefit of being able to fly in areas where traditional airplanes can't.

The standard image size used to be 12 megapixels; now, 20 megapixels is the absolute minimum for most use cases. Larger platforms are carrying 40-, 60-, and even 100-megapixel RGB sensors. This allows for more efficient flight paths, as the drone doesn't have to be as close to the subject to get the same resolution, or ground sample distance (GSD). This means fewer flight lines, higher altitude, and fewer images collected.

Much like their RGB counterparts, these sensors have gotten cheaper and lighter and have more resolution. Surprisingly, there is slow uptake on their use. There could be a couple of reasons for this. Unlike an RGB camera, accurate use of multispectral sensors requires an additional manual step of calibrating the images. This might be limiting the market to the agricultural sector, which is more familiar with this process. And there are very few plug-and-play multispectral packages for the more popular aircraft platforms. The number of packages is certainly improving but more slowly than expected.

Companies have been working on putting lidar on drones since the advent of drones. Two of the major challenges have been the weight of the sensor and the accuracy of the inertial measurement units (IMUs). Recently, these two problems have been solved. Laser pulses can penetrate vegetation, allowing for a more accurate digital surface model (DSM). This leads to the challenge of managing the resultant point cloud data. The future here is cloud management. This trend mirrors how aerial mapping from crewed aircraft transitioned from imagery to lidar, overcoming similar challenges but in a compressed time frame.

In the past, each drone manufacturer had its own development team for autopilot navigation and aircraft control. Now, drone hardware manufacturers are relying on autopilot companies to handle the command-and-control piece of the system. Auterion is one such manufacturer, creating Skynode, an all-in-one autopilot that handles the autonomous aspect of the flight. Skynode also has a cellular data connection that essentially turns every drone into a capable IoT device.

In a dramatic shakeup, the US Department of Defense (DoD) banned Chinese DJI drones used by the department. In response, the DoD's Defense Innovation Unit (DIU) began evaluating drone hardware and software to be approved for use. The DIU's current list includes hardware from Freefly, a US drone manufacturer that has been developing drones for movies and TV for years and is in position to be a reliable mapping platform as well. 

As sensors get larger and contain more data, the digital size of datasets gets heavier. Processing datasets used to be solely done on desktop computers. This is shifting to a cloudcentric processing method. Offline desktop processing will always have its place—primarily in rapid field processing for immediate decision-making. The raw data then makes its way to the cloud, where there can be higher resolution and larger datasets for the rest of the organization's uses. The cloud can autoscale processing to suit the needs of the larger file sizes and deliver the result to the user's enterprise workflow.

An example would be during a disaster event—drones are often the first aircraft on the scene collecting data. If internet connection isn't available, local rapid processing can give first responders the information needed to make decisions. Later, the devices connect to the internet and the raw data is uploaded to the cloud, where it is processed into larger datasets. Stakeholders can analyze these datasets online without having to transfer them from computer to computer.

Turning raw images into 2D and 3D datasets will always take time, but what is exciting is the growth in real-time or near real-time analysis of the data. FMV is a technology originally developed by the military for metadata-enriched video coming down from specialized (and expensive) gimbals. As the cost of these sensors decrease, it becomes easier and cheaper to have FMV data.

If you have a sensor package that supports FMV data, like the Quantum-Systems Vector, you can have FMV livestreamed to a geographic information system (GIS) such as ArcGIS Pro. Or, if using a low-cost consumer drone supported by Esri's Site Scan Flight for ArcGIS app, you can pull video files straight from the drone and play back enriched videos by dragging-and-dropping them.

As the volume of information from drone collections increases, it's more important than ever to leverage machine learning to help identify assets and problems. The requirements for success are training data, high-resolution input information, and compute power. All of these are being improved by the hardware and software trends we're seeing. AI can be trained on raw images and video from drones as well as the 2D and 3D outputs of the processing engine. 

As drones become part of the standard enterprise workflow, they are managed like computers or IoT devices, but also like vehicles—with regulatory requirements, maintenance, and flight tracking. There are many solutions for tracking drone flights, but winning technologies will capture this information automatically, allowing operators to do their jobs without interruption. In the US, for example, if flying within a certain radius of an airport or other controlled airspace, you must request authorization from the Federal Aviation Administration (FAA). Airspace Link is one company automating this process by sending requests directly to the FAA as an authorized Low Altitude Authorization and Notification Capability (LAANC) provider. Within Esri's Site Scan Flight app, operators can log in with a free Airspace Link account and request this authorization.