By 2036, drones will represent a $30 billion industry according to a recent study. Drones have emerged as a significant force for optimising and enhancing business operations in today’s society. From monitoring remote infrastructures, such as oil pipelines, power transmission towers and hydroelectric dams, to maintaining a birds-eye view of solar panels, large companies are turning to drones for getting the job done. However, despite an increase in use, there are a few barriers we need to overcome before we see their full potential.

Today, drones are used in nearly every sector. Besides taking pictures and videos from high up in the air, over water and on land, they collect data, and that is where their real value lies. Thanks to advanced sensors and imaging capabilities, drones can, for example, help firefighters obtain information about an accident’s location, provide insights about crops to farmers that allow them to reduce crop damage, or facilitate the inspection of remote tracks and bridges by train companies. These are just a few of the examples amongst hundreds of how drones may be utilised.

Whilst drones are proving to be beneficial to many industries around the world; their full potential remains unlocked. Rohit Gupta Vice President & Head of Manufacturing, Logistics, Energy & Utilities – Europe at Cognizant, explains the three main hurdles that prevent drones from ‘flying higher’ and being fully integrated into everyday life.

1. Flight regulations and security protocols 

As manned and unmanned aviation share a single sky, they will always be intertwined. Many countries are implementing one-time, short-term solutions that vary from country to country across Europe. As a result, the European Aviation Safety Agency (EASA) is actively involved in bringing together these fragmented set of regulations into one cohesive set of key safeguards. 

An important step in this direction is the Drones Amsterdam Declaration, which was adopted in November 2018 by European and national stakeholders. This declaration prioritises supporting the Member States in implementing technical rules to enable drone flights over longer distances, using the European U-space Demonstrators Network to carry out safe, secure and green drone operations and to develop a U-space system.

On other fronts, the EU is also pushing to adapt the regulatory environment to the increased presence of drones. SAFIR, a European Commission consortium of 13 public organisations and private companies aiming to assist the EU in establishing a traffic management system commercial drones, will also be running tests and demonstrations in Belgium in 2019.

Another area of focus will also be on training the pilots in processes, and it is safe to say that the foundation already exists in the current model for today’s aviation operations with a human crew. This provides a solid start for automated ‘identification and separation’ services. All in all, the question is not ‘if’ a solution can be found but ‘when’. 

2.Use of remote IDs 

As a general rule, all aircraft must be identifiable. Remote ID technology is in place and largely used in manned aviation; however, this is still a challenge when talking about drones. The Open Drone ID Project is a testimony to this, and, in summary, the biggest issues are: 

• Costs: developing ad-hoc technology to meet performance standards issued by different national authorities is very costly. Due to a large number of existing drones, it is possible that economies of scale could be applied to undertake the necessary tests to meet international standards. 
• Power:  remote ID requires transmitting information across significant distances, a process that consumes power, the most precious resource in a drone. A second concern is lighting, which also impacts energy saving. To solve this problem, power storage (batteries) and power generation (onboard combustion fuel, fuel cell or solar solutions) go hand-in-hand.
• Weight: just like in manned aviation, every bit of extra weight added to an aircraft translates into less room for payload, as well as less time in the air, and will use more power

3.True learning and Artificial Intelligence (AI)

Drones gather large amounts of data, including detailed videos and pictures; however, the ability to learn and use AI while still flying is under question. Challenges include:

• Vantage Point: There has been considerable work done over the past 30+ years related to AI and image analytics, much of which can be transferred to the drone industry. However, drones have a very different vantage point than, say, a robot on an assembly line. While the transfer training tools exist, what is missing is a large body of drone imagery (like maps, point clouds, 3D models and Drone Elevation Models).  Additionally, while satellite and manned aviation images exist, they do not compare to the resolution and density of images collected by drone. Over time, this drone image learning database will grow to enable efficient transfer learning in the AI sector.  
• Storage and Streaming Limitations:  At the moment, commercial drones collect data and either store ‘high-quality imagery’ onboard for post-flight processing or streaming less dense data for in-flight analytics. While there are exceptions (high end / high cost commercial and military applications) this is the current state of play across many companies when adopting drone technology.  Consequently, if a business needs to collect data for processing post-flight, it is limited by the amount of storage memory within the drone. While drone memory can be extensive, it ultimately is finite. 

While streaming of lower quality imagery is supported, for many applications, this lower density data is insufficient for remote analytics processing.  

The ideal solution would be to limit onboard processing for navigation and collision avoidance and develop the capacity to stream higher quality imagery for off- craft analytics. This way, CPU cycles are dedicated to very specific real-time processing needs, while more in-depth analytics can be performed on big data-based, large scale, parallel processing systems. 

• Each drone could be equipped with a trained model: essentially, the information core of an organisation.  If the drone is lost (or it falls into the wrong hands), a company could potentially lose confidential data. 

Fortunately, this data can be effectively encrypted.  Furthermore, while others could use a trained model, it gives no direct insights into the training weights and sample images that went into developing the model, thereby protecting both intellectual property and the data stored aboard.  Alternatively, another option would be to make sure that sensitive information is migrated to a physical data centre or a cloud-based secure server to enhance security should the craft be lost. This option limits the data at risk to just the images that remain on the craft for the mission at hand.

Drones will soon be taking off to change the way packages are delivered, and even how farmers grow their crops. Tapping into their full potential will demand technology and regulations to evolve hand-in-hand. One thing is certain, however. The use of commercial drones is inevitable, and there is no putting the drone back in the box.