Agents in network-centric joint-force operations
In modern, network-centric joint-force operations, there is a high demand for intelligence, surveillance, and reconnaissance (ISR) resources, which must be defined within a limited time frame and in such a way as to provide the most suitable assets for a mission. As stated in Joint and National Intelligence Support to Military Operations Joint Publication 2-01 (2004), providing the most efficient resources to meet mission requirements is a major challenge. Pechoucek, Thompson, and Voos (2008) discuss a solution that involves analysing the mission definition and the available resources to match these in the most advantageous way.
The solution appears to be agent-based as an individual computer system performs actions and makes decisions automatically in an environment to meet user-defined design objectives on the user’s behalf and in its best interest (Wooldridge and Jennings, 1995). It uses the model provided by the Missions and Means Framework (MMF), first classifies the mission into operations and tasks, then builds the ISR resources from components, systems, and platforms for the tasks identified in the first step.
For example, given a military mission planning environment that includes various components, for example, cameras, unmanned aerial vehicles (UAVs), and a search-and-rescue mission definition. Based on this information, the agent divides the mission into separately identifiable tasks and prepares a descriptive list of the requirements necessary to approach the assignment. The system must gather information from these resources to match them with the mission. After evaluating the assets, the agent recommends that an EO/IR camera is needed, built into a camera tower, and mounted on a UAV, but it does not change the military mission planning environment with the decision it makes.
If taking into consideration the definition of expert systems (Wooldridge, 2003), it is debatable whether the discussed solution fits the agent definition. An expert system can solve problems and provide advice or feedback based on information coming from a user. What distinguishes the solution from an expert system is the sensor data collection, reasoning, and coordination that are performed to successfully carry out matchmaking in the most advantageous way.
The solution can be distinguished from regular computer systems in that it must independently find the most appropriate resource, while the user does not give instructions for this, only a dedicated goal that must be achieved.
A matchmaker service does not fit the definition of an intelligent agent because it is not reactive, proactive, and does not have social capabilities (Wooldridge and Jennings, 1995). While there is a knowledge-sharing relationship between the agent and the various assets, this communication is not an approach to a coordinated solution to a problem, the gathered information is used by the agent to make decisions. The solution is not implemented in such a way as to react to changes, it only strives to achieve a goal defined by the user and does not need to negotiate with other agents, the decision is based on sound reasoning.
The study by Pechoucek, Thompson, and Voos (2008) tries to deal with situations when different agents with their own goals and points of view need to come to a common agreement by coordinating their activities and negotiation. A military environment is only partially observable, so wrong information can easily be obtained. In addition, agents must reason based on the information they receive from sensors, which assets are fallible and may provide unreliable data resulting in ineffective operations. The reasoning method is faster and involves gathering more knowledge than the competing method. Finally, the agents must produce arguments based on which they can form a common opinion. This solution would fit the definition of an intelligent agent.
There is a tremendous need to assign assets quickly and efficiently for network-centric joint-force military operations. Once assets have been delegated to a particular mission, the agent has no further information about the consequences of its decision. It is not clear to the system whether the action was successful or whether there were unintended victims. Simulations conducted by the U.S. military are designed to reduce the number of assets used and the number of deaths (Allen, 2011). In this way, agents can face situations close to real life and receive feedback on their decisions.
An agent-based solution was discussed, which analyses a user-defined mission and evaluates the applicability of resources. The solution could fit the intelligent agent definition by adding argumentation. In this way, the agents would negotiate by reasoning until they reach a common agreement. Such a solution involves processing a large amount of data, cyber-technology, and autonomous weapons. These systems become increasingly vulnerable to cyber-attacks, which could lead to disruption of operations, theft of sensitive information, manipulation of data, or even unintentional civilian casualties. Moreover, the complexity of such a solution increases, which could lead to a decrease in management and maintenance efficiency.
It can be concluded that the discussed system fits the definition of an agent because it is able to make decisions automatically based on the collected information from the environment where it is situated, without the need for instructions from its owner.
References
Allen, T.T. (2011) Introduction to Discrete Event Simulation and Agent-based Modeling.
Joint and National Intelligence Support to Military Operations Joint Publication 2-01 (2004).
Pechoucek, M., Thompson, S.G. and Voos, H. (2008) Defense Industry Applications of Autonomous Agents and Multi-Agent Systems Editors Whitestein Series in Software Agent Technologies and Autonomic Computing. Available at: www.whitestein.com.
Wooldridge, M. (2003) Introduction to MultiAgent Systems. Wiley. Available at: http://ebookcentral.proquest.com/lib/liverpool/detail.action?docID=139837.
Wooldridge, M. and Jennings, N.R. (1995) ‘Intelligent agents: Theory and practice’, The Knowledge Engineering Review, 10(2), pp. 115–152. Available at: https://doi.org/10.1017/S0269888900008122.