© istockphoto.com/narvik


The principles behind autonomous aircraft

“We know what you want” – the tagline used by author Marc-Uwe Kling in his novel, QualityLand, for an online e-commerce platform called TheShop. Data evaluations predict what customers actually want – consciously or subconsciously. Products are then sent to them by drones, without even needing to place an order. And as they open the package, the “unboxing” is filmed. This video is then immediately posted online through a platform called Everybody. Finally, customers rate their experience on a touchscreen held up by the drone. [1] We’re not a million miles away from this sci-fi scenario, especially given current rates of technological development. Companies like Amazon and DHL are already working on a reliable system that will make it possible for drones to make deliveries.

The term for the underlying principle that will make such concepts possible is autonomization. According to Kant, autonomy was tantamount to “sensible self-determination,” whereby the freedom to make decisions is subject to certain rules. Autonomy is measured according to a specific environment – the place in which individuals define themselves or act with self-determination in order to fulfil their sense of purpose. It thereby revolves around individuals’ norms and rules. [2] When systems interact with their environments, they organize themselves without relinquishing self-determination. Systems use interactions with their environment to expand their sphere of influence and they exploit potential through multi-dimensional interactions. They attempt to achieve self-defined goals by deciding which data to use, by processing and evaluating unstructured data, and by adapting decision-making structures. They strive to safeguard their own power to act and, if possible, extend it. According to scientific definitions, autonomy can be extended to include the ability to remain functional in unknown environments. This ability is referred to as robustness and it necessitates a certain ability to self-organize.

From a holistic standpoint, systems are driven by macro-level goals of self-optimization and self-preservation. There are three features of an autonomous system that can be derived from this. The system’s actions are self-determined and self-organized, and it has the power to take action; to achieve this, a system gives itself an identity or is provided with one by a system designer; and it interacts within a trusted network according to global governance dictating system interactions. [3]

The Ferdinand-Steinbeis-Institute’s definition of autonomization in the area of business information systems is the implementation of (sub-)processes in business according to these three features. For this to work, different types of digital technology and information systems can be used, and these facilitate systems comprising agents whose role is to make use of autonomous processes. This is the case if, for example, an autonomous smart home system orders something by itself, books medical appointments, or adjusts the air-conditioning in a house to make things comfortable for occupants. The system organizes things itself to achieve this and interacts with stakeholders in a trusted space at home using its own identity. It carries out valuable transactions itself and reacts flexibly to unknown situations according to consistent system behaviors.

Autonomization: model showing different levels

Introducing autonomous processes requires a large volume of data which then needs to be analyzed. With data-driven approaches, models are developed based on heterogeneous, multi-dimensional data records. These are required to make the information available that is needed to make decisions and thus carry out processes autonomously. Compared to mathematical methods, data-driven methods make it possible to assess larger volumes of data spanning multiple factors – the type of information generated by the advent of the internet of things. In comparison to automation, with autonomous (sub-)processes, unstructured decisions have to be made. Systems decide for themselves which factors and information sources should be included to make a decision.

These may differ depending on the context and time factors and different outcomes may be achieved for different clients despite similar maximization premises. This contrasts to structured decisions, whereby the same outcomes may be achieved for different clients despite similar maximization premises. Technologies such as deep learning and distributed ledger offer potential ways to achieve the principle of autonomization. The FSTI is looking into these technologies in order to develop concepts aimed at autonomization. Accordingly, experts at the FSTI interpret the three features of autonomization thus:

  • Self-determination: Systems define their own goals and (sub-)processes for achieving goals in keeping with their own rules and standards.
  • Self-organization: This comprises the selection of factors and data sources for unstructured decision-making, processing and evaluating large volumes of multi-dimensional data, and adapting the system’s own decision-making structures (= self-learning).
  • Ability to take action: This refers to how systems implement decisions while interacting with their environment and safeguard their robustness.

People want to come up with systems for introducing autonomous processes in many areas. But it is difficult to work out whether they mean autonomous, automated, or digital systems when they say this, because there are so many different definitions.


According to Nicholas Horbaczewski, the founder of Drone Racing League, drones will have the same ability to cause disruption as smartphones. They can transport objects short distances into areas that are difficult to access. From delivering packages to helping with medical emergencies, there are a variety of ways to use drone technology. [4]

In the future, collaborative drones could become components within systems that will take on a variety of tasks according to the principles of self-determination and self-organization. As unmanned aerial vehicles (UAVs), drones offer an important complement to manned aircraft because they can be used for dangerous missions, but also monotonous tasks. Examples include using UAVs to access remote locations in disaster areas, carry out inspections, or transport goods in bad weather. [5] The aim in the future will be to raise the automation levels of manned aircraft to provide pilots with support, or even take on tasks as autonomous UAVs. Accordingly, airspace will be a mixture of manned aircraft and UAVs. This will require coordination mechanisms, and progress will be dictated by the maturity levels of emerging technology.

Autonomous flying (schematic representation)

The system architecture of a UAV comprises one or several on-board computers containing software for controlling flight electronics. The number of on-board computers depends on the required level of modularity and partitioning. There are also sensors and actuators on board to allow drones to be steered and propelled. Then there is a special unit for interpreting sensor data. Defined actions are then carried out by actuators. Drones also contain communication systems so they can interact with other objects or units on the ground. They have special systems for carrying out controlled landings in an emergency. [6] In his novel, Marc-Uwe Kling expands this physical level of drones by adding a digital level so they can immediately make complex decisions based on data models. On the digital level, the flight system in QualityLand uses different types of digital drones and other data sources to carry out autonomous processes – including weather data, a shopping database, and a platform called Everybody. Each step of the process, from ordering to shooting the unboxing video, is self-selected, self-organized, and carried out by the system itself, which has the power to act as it needs to. Events within this process are not fixed according to a rigid timeline; the system plans steps itself to adapt to the changing situation – it’s a self-learning system.

Projects like Uber Elevate and Uber Air will allow Uber to experiment with air transportation for the next five years. Uber’s aim with Uber Elevate is to operate flights between fixed points using autonomous and electric technology. Uber Air is an on-demand service offering flights on adaptable routes. [7] Such projects require further development when it comes to autonomization, however, in order to provide suitable autonomous aircraft.

Technological concepts such as the internet of things, artificial intelligence, and data processing will also need further development. Nonetheless, the situation is less complicated with autonomous aircraft compared to developing autonomous road vehicles, because airspace is highly regulated – there are fewer unknown variables to take into consideration (like pedestrians). [8] One thing we can be sure of, however, is that some things will never change: “Could you possibly look after these two packages for your neighbors?” [1]



Daniel Burkhardt (author)
Research Assistant
Ferdinand-Steinbeis-Institute (FSTI) (Stuttgart)

[1] Cf. Marc-Uwe Kling, QualityLand, Ullstein Buchverlage GmbH, 2018, Seite 18
[2] Cf. Kant 1968, BA 87, Buss 2011
[3] Cf. Littlewood 1996, Froese, Virgo and Izquierdo 2007
[4] Cf. https://medium.com/foursquare-direct/in-ten-years-the-future-of-ai-and-ml-fa68a527f378
[5] Cf. https://www.dlr.de/ft/desktopdefault.aspx/tabid-1377/1905_read-3360/
[6] Cf. https://www.dglr.de/publikationen/2018/480238.pdf
[7] Cf. https://www.adweek.com/brand-marketing/uber-elevate-ambitions-flight-sharing
[8] Cf. https://www.gartner.com/smarterwithgartner/5-trends-emerge-in-gartner-hype-cycle-for-emerging-technologies-2018/, https://www.itworldcanada.com/blog/5-trends-on-gartners-hype-cycle-foremerging-technologies/423379