Autonomous Systems and Artificial Intelligence
Rapid conversion of bulk raw data into relevant information is very important in context of decision making and decision support. The volumes of registered and available data grow exponentially every day. Artificial intelligence (AI) is one of the different Big Data processing techniques. AI includes self-learning systems. Once fed with (large) quantities of appropriate data, these systems can autonomously take or propose certain actions or decisions (whether physical or non-physical), and thus demonstrate a form of (controlled) real artificial intelligence.
In recent years, quantum computing is becoming more important. Quantum computing is very different from classical computing. The way in which quantum computers work, allow them to solve certain problems which are considered difficult for conventional computers in much less time. The consequences of this new method of computing are broad. One example is the effect of quantum computing on cybersecurity (Research area 05), rendering frequently used asymmetric key distribution systems vulnerable (hence the importance of post-quantum cryptography). Another example is the robustness in image processing where classical machine learning fails (for example, classic machine learning methods for image recognition can be tricked by changing one pixel). Quantum computing methods are more robust to this type of countermeasures. While there are many more examples, the two examples above indicate that quantum computing is important for Defence.
The trend towards autonomy1 of operational systems has direct consequences for Belgian Defence. Intelligent autonomous systems are not to be confused with unmanned vehicles (UxVs2) or unmanned systems (which are not necessarily autonomous). Autonomous systems and system of systems are part of the answer to the increasing risks of modern combat with more and more urgent constraints. Among these constraints are reducing costs while increasing efficiency, alleviating the lack of military personnel, and reducing risks for and/or by human actors in operational missions. Belgian Defence is therefore looking at (semi-)autonomous system-based solutions aiming at accomplishing specific challenging types of missions (air, land, sea, cyber and/or space). The latter two types are covered respectively in the Research Areas 05 and 01.
One of the invaluable advantages of intelligent autonomous systems is that they can overcome the physical limitations of human beings. They will fundamentally change the way armed forces operate, by increasing situational awareness, reducing the soldier’s physical and cognitive workloads, improving sustainment, facilitating movement and manoeuvre, increasing reach and range and force protection. In addition, intelligent autonomy will enable capabilities that are currently not possible (such as long-duration unmanned underwater observation and surveillance, where the platform must be able to work for months without human intervention or communication). Autonomous systems technology will also provide a significant opportunity in the training and education of armed forces to improve learning and provide cost-effective and realistic training.
The concept of intelligent autonomous systems must be viewed in a broad sense. Autonomous systems can respond to uncertain situations by independently composing and selecting among different courses of action to accomplish goals based on knowledge and a contextual understanding of the world, itself, and the situation. The world, situation and course of action in this context can be physical or virtual. To this end, data yielded by sensors and sensor systems can be used (Research Area 03). Autonomy is characterized by degrees of self-directed behaviour (levels of autonomy) ranging from manual, over automatic and automated, to fully autonomous. In general, these levels of autonomy3 are applicable to only a part of the system, rather than to the complete system. Intelligent refers to the ability of these systems to perform tasks that normally require human intelligence (recognising patterns, learning from experience, drawing conclusions, making predictions, or acting) whether digitally or as the smart software behind autonomous systems.
The ability to process and interpret large quantities of data from different origins, and take action accordingly, allows autonomous systems to interact with the environment, whether or not in a multi-agent context (human-system (Research Area 11) or system-system). Representing the processed data in a human-understandable manner (even at the lowest operational levels) enables staff to secure and supervise areas. In urban areas, the scene of an increasing number of military operations, information from vast numbers of sensors is available in real time. Processing, analysing, interpreting, verifying and validating, rendering and representing this information, contributes to the development of this Research area. Developments in terms of IT and AI for the use and analysis of hybrid systems succeed each other at a high pace. To increase the autonomy of systems, including interaction and multi-agent collaboration, very specific technological aspects require further exploration and in-depth examination. Some examples of these aspects are decision support algorithms, cybersecurity, procedures to evaluate vehicle/operator/team system performance as a function of platform autonomy and the design of these systems in terms of hardware.
The study, development, design and/or use of Big Data, AI and Autonomous systems entails ethical and legal issues, considering the intended purpose, the effective use, the desired, required or allowed level of autonomy, the rules of engagement and rules of command and control, which will be addressed within the Research Area 11. The energy aspects of autonomy will be addressed in Research Area 09.
1 In its simplest form, autonomy is the ability of a machine to perform a task without human input. (NATO ACT – Autonomous Systems – Issues for Defence Policymakers)
2 “Unmanned Vehicles (UxV’s) may be remotely controlled by a person or may act autonomously depending on the mission. Applications include allowing for access to unreachable areas, persistent surveillance, endurance, robots in support of soldiers, and cheaper, automated logistics deliveries” – (NATO Science & Technology Trends 2020-2040)
3 Ref: NATO Science & Technology Trends 2020-2040