DaaS — bringing together drones, cloud, AI and 5G
Currently there is a massive increase in the commercial use of drones to realise new innovative use-cases — which are only now possible. In 2018, the drone market was worth $25.59 billion. In 2028 it is expected to reach $70.28 billion(Statista). Now organisations are starting to realise the real benefits having of UAV drones in their business arsenal, I will discuss what this could look like technically speaking.
DaaS
Introducing the term DaaS (Drone as a Service); which is the implementation of several technologies to enable the operation of smart connected UAV drones commercially.
In this series of articles, I will delve deeper into each of the individual technologies/components and provide context on how they are used to enable the operation of the UAV drone. In this article specifically, I will ‘fly’ ;) over the technical concepts.
These include;
- Artificial intelligence
- Cloud computing
- Edge computing
- DevOps
- 5G
Drones
First a brief history of drones — where did they originate from and why are they now disrupting industries?
The first use of drones were to serve military purposes in WW1, the UAV in question was named the Hewitt-Sperry Automatic Airplane — used to deliver missiles aerially without a pilot, which looks a lot different to the drones in question in recent times.
The Kettering Bug managed it’s flight in a mechnical manner. Before flight, a flight time was calculated and total engine revolutions were determined and when this value was reached the engine would shut off. This ensured that the aircraft would reach its intended destination, although this worked — it meant that each aircraft would be one-time use. With each Bug costing around $400 USD in 1918 it simply wasn’t cost efficient.
The aircraft’s high cost inspired the use of radiowaves to remotely control the path of flight, this work was a basis of today’s remote controlled UAVs.
Without getting too sidetracked with history, the first commercial use of drones was in 2006 with much R&D taken from historic use in the military — this year was also the first time the FAA issued a commerical drone permit. Due to Beyond Visual Line of Sight (BVLOS) innovations, organisations are slowly starting to realise the benefits of drone adoption for their unique use-cases.
Probably one of the most notable mentions of recent commerical adoption is Amazon Prime Air, who started R&D into this domain in 2013 with FAA approval in 2020 alongside Wing (owned by Alphabet inc.) in addition, a handful of logistics focused entities. Amazon Prime Air is aiming to deliver packages to customer’s doorsteps within 30minutes or less. They successfully completed their first delivery on December 7th 2016 in a 13min timeframe.
Another early pioneer into the drone domain was Vodafone, who also started to realise new innovative use-cases — also in 2016. Vodafone are focusing on drones for their own internal use-cases, but more crucially how they can provide the network capability to allow connected drones to operate effectively. They are introducing mechanisms to deal with issues such as interference, long-distance control and to ensure allowed fly zones are adhered to — to name a few.
More recent innovations also include the pairing of realtime AI and drones, allowing organisations to plan routes and not be required to manage flight or inspect any incoming payload — instead an AI model can be implemented to complete various tasks.
Artificial Intelligence
So by now we can see the potential of drones and their future in the industry, this can be further increased by implementing the use of AI on the drone. The AI software used can be implemented in many ways; with different payloads harvesting data for various use-cases. In the context of this article, the AI application will be used to process object detection on live video.
Although this presents some challenges as AI operations tend to have strict hardware requirements (such as requiring a GPU) to be able to run; each frame from the live video has to be processed by the AI model, without the horsepower this wouldn’t be possible.
In order to run this we have a couple of options:
- Equip the drone with GPU hardware
- Offload the data to a GPU located at an edge server (more on this in later articles)
Cloud Computing
Cloud computing is also a consideration for this drone solution as we want to apply cloud native principles to the development of the AI & drone management software we will deploy.
Cloud native principles; such as:
- CI/CD
- Containers
- Microservices
- Elasticity
Using cloud native principles, we can automatically deploy new software versions to be implemented on the drone via a CI/CD pipeline. The software will be deployed as a container, which can be redeployed in place of the old container with very minimal/no disruption to the user. In addition, its possible the demand for use of drones can fluctuate, so applying elastic scaling principles found in the cloud; we can dynamically scale up or down the applications deployed to serve the operations of the drone.
Without going too low-level into the creation of CI/CD pipelines and containers etc — we will leave it at a high-level and go lower level in further articles.
4G & 5G
Mobile connectivity enables worldwide communication with the fleet of drones currently in operation; to retrieve flight data (such as location, altitudes e.g.) and also any payload data (such as video detections, sensor information e.g.).
The difficult thing to ensure here is the reliable delivery of data. One type of data is the payload data; which includes real-time video —being sent to the user or the edge; which could host an application which assists the drone’s operational capacity.
In addition, the flight data must be reliably delivered for control and management of the fleet of drones (or singular drones). Without this, we would not be able to ensure safe flight and this would certainly be a major issue; with drone flight laws being implemented by many nations.
Considering the network architecture of 4G and 5G deployments, we can be certain that we will not face such issues. In later articles, we will explore how we plan to implement specific 4G/5G software and hardware options to overcome such issues with drone flight and data transmission.
Edge Computing
One facet of the 4G & 5G considerations is edge computing; which aims to bring the application closer to the source of data which could be a user or self-driving vehicle for example.
As you can see in the image above, the data network (DN) is brought closer to the user (UE), being closer geographically we can see vast reductions in latency and bandwidth strain. You can now see at the edge it is referred to as LADN (Local Area Data Network) and you can see the traditional DN (Data Network)
More specifically known for its operation in 5G, to allow LADN communication with User Equipment (UE). However one large factor to its success is the ability to be deployed not only in 5G networks but also 4G networks — included is the ability to easily upgrade the deployment to available 5G infrastructure via software changes.
Specifically to drone operations; the edge will host the application which increases the operational capacity of the drone — such as AI execution, collision avoidance, analytics gathering and so on; the similarity of these applications is that they require ultra-low latency and high bandwidth enabled by edge computing.
Summary
The purpose of this article was to introduce all of the technical concepts which can be incorporated together to enable a smart connected drone operations possible, with the aim of delving deeper into each of those concepts more specifically in following articles; to provide less vagueness and give real examples of how they can be implemented via specific software and hardware options.
