Artificial Intelligence (AI) has been widely taken up in defense circles, mainly for use in unmanned vehicles, both aerial and terrestrial. Nowadays, too, AI is also coming into its own in command and control systems (those that support operational command in driving tasks and monitoring of forces assigned to any mission), with uses ranging from data-acquisition and -analysis to presentation of the information to the operator. This article glances at AI’s present and future contribution in all the abovementioned fields.
First of all we’ll look at data acquisition, where AI is showing great potential in two of its main components: adaptive sensor management and sensor data fusion. Adaptive management involves an adjustment of measuring parameters and processes and a coordination of sensors to suit each particular mission’s requirements. The new emerging scenarios (like hybrid warfare) and the new generations of sensors (biometric, social, urban) are prompting new advanced forms of adaptive management. Techniques like data mining in social media and real-time video analysis, to quote only two examples, will facilitate management based on a more general analysis of the context, going well beyond traditional techniques based on control theory. For their part, the modern deep learning algorithms, used in conjunction with classic statistical techniques (like the Bayesian methods, which stand right on the edge of what we currently understand by Artificial Intelligence), push back the frontiers of data fusion, coming into their own for the combination of unstructured data (like voice data or natural-language text) from a wide range of different sources. The main challenges they raise are correct alignment or association of the data and management of inter-data conflicts. By way of example, to deal with information on the same tactical component coming from different sources the AI algorithm would take on the task of establishing a relationship between them, solving the divergences on the basis of an assessment of the credibility of each source in different situations.
Another command and control area where AI is proving highly useful is aid in decision making. At the time of writing the idea of completely replacing human beings in important decision making still seems a long way off, but AI-based tools are already crucial for cutting down the decision-makers workload. Prime examples are situation assessment and choice of the best line of action.
Situation assessment refers to an understanding and interpretation of the operational scenario and possible threats, for the purpose of achieving tactical and strategic targets. Experience and knowledge of the enemy, as well as own and enemy doctrine, are therefore still essential for a proper assessment of the situation. AI, nonetheless, makes the task easier by helping out the human being in the previous diagnosis, drawing on the abovementioned data fusion and data preprocessing and analysis. This gives the operator the necessary information in a structured and understandable way.
Selection of the line of action, based in turn on the assessment of the situation, involves an analysis of the possible alternative lines and forecasting the consequences of each one. Once again, the decision maker’s experience and knowledge are the key factors here. Nonetheless, algorithms like those of machine learning can come in very handy in terms of designating the best decision maker on the basis of the information to hand (i.e., when the obtaining of additional information would not impinge significantly on the decision in itself); this would be based on a past record of similar experiences. This is crucial. In command-and-control, after all, timing is everything, and the chances of success are much higher if the decision has been taken at the right moment. On the basis of the abovementioned past record of experiences (consisting basically of a prior knowledge of background facts, decisions and consequences in comparable situations), the tools are also capable of making a quantitative estimate of the impact associated with each possible alternative, together with the working hypothesis and the corresponding explanation, allowing the human operator to understand why a given decision might give rise to a given result.
To wind up this nutshell account, we will touch on a key element of the presentation of information in command and control systems, i.e. the Common Operational Picture (COP). The COP takes on responsibility for showing the significant operational information, such as the positions of friendly and enemy units; it is typically associated with a map showing the objects of interest graphically. In the classical systems these objects are fed in manually by a human operator and then displayed to everyone with authorized access to the COP. Machine learning techniques mean information can now be added to the map automatically, drawing, for example, on an analysis of satellite images or of a drone sent to reconnoiter the operations area. In this way the objects of interest, such as enemy troops or important infrastructure, can be automatically detected, analyzed and presented to the operator.
As already pointed out above, as things stand, human judgement will continue to be the overriding factor in command and control decision-making for some time yet. Even so, the efficacy now shown by AI-based tools and their sheer speed of development will enable the human being to be freed of some of the more accessory tasks, allowing him or her to concentrate solely on the areas where human beings still outperform machines.
Author: Raúl Valencia
Las opiniones vertidas por el autor son enteramente suyas y no siempre representan la opinión de GMV
The author’s views are entirely his own and may not reflect the views of GMV