Model Dynamics

[NOTE: Freewriting first draft for the "Military Modeling" chapter of the upcoming book, Dynamic Modeling, edited by Paul Fishwick, to be published by CRC Press in 2006.]

To this point we have focused on defining and categorizing military modeling according to its application. Those categorizations were meant to illustrate the unique situations, problems, and interests of the developers and customers for military models and simulation systems. In this section we will describe the most dominant forms of dynamic modeling that are used in the community. Because military systems and problems are so diverse and such a large investment has been made in exploring them, there are many more unique forms of dynamic modeling than can be captured in a single chapter or an entire book. However, the forms that are described here are some of the most commonly used. They are also presented as general categories that cover a number of unique implementations.

Dynamic modeling is military simulation often focuses on activities like:
· Movement,
· Perception,
· Exchange,
· Engagement,
· Reasoning, and
· Dynamic Environment.

In this section we describe the dynamics that are included in each of these categories. This is followed by a section that explores multiple approaches to modeling these dynamics.

Movement

Dynamic representation of movement captures the change in an object’s position over time. Models may represent position as a coordinate in two-space, three-space, or a velocity vector. Two-space coordinates usually include a position in X and Y, such as latitude and longitude. For models that represent only ground-based vehicles like trucks, tanks, and foot soldiers, this can be sufficient. The object may have no variation in elevation, or the elevation may come from the underlying elevation of the terrain on which it sits. Position may also include orientation, which in two-space would be limited to a 360 degree angle around the vehicle. A common reference system for this angle is with the zero point being aligned with true north and proceeding clockwise with 90 degrees being east, 180 being south, and 270 being west.

In three-space, the coordinate system includes a representation of elevation. This third dimension may be height above the local terrain, elevation above mean sea level, or distance from the center of a sphere that represents the Earth. The latter measurement evolved during the creation of distributed heterogeneous simulation systems. When networking multiple simulations, differences in the terrain representation within each system led to significant differences in vehicle position with respect to the terrain. Therefore, a non-terrain referenced coordinate system was needed to overcome these differences. When a three-space orientation is added to this model, it includes the pitch, roll, and yaw of the object, creating a six degree-of-freedom (6-DOF) model. When represented as a vector, this may also include the velocity of the vehicle along the axis of orientation.

In their basic form, movement models change these position and orientation coordinates according to a logical or physical representation of movement, as described in the next section. However, most implementations go further to include the effects that movement has on the object and the environment. The movement model may be linked to a model of the fuel consumption of the vehicle. This adds a limiting factor that can stop movement when the fuel is depleted. The inclusion of a fuel model leads to the need for the system to represent a process for replenishing the fuel consumed. Otherwise, the objects in the simulation will eventually grind to a halt. In military modeling, the addition of these details leads to the need for many more models to drive the additional variables that were added. Systems can grow far larger than can be developed, funded, or hosted on a computer through the poor management of these modeling details. Many authors have warned against this gradual creep in features that leads to the eventual failure of the system being developed (Law and Kelton, 1991). This type of growth is not limited to movement modeling, but can occur throughout the system if the designers do not control it.

A movement model may also calculate the number of hours of operation that the object has been used. This information is the root of most system failure and maintenance models. This drives a mean time between failure (MTBF), repair (MTBR), or other similar models.

The interaction of object movement with the terrain can generate environmental changes that trigger yet another model, such as the generation of smoke or dust clouds in the wake of a vehicle. If these changes to the environment are represented, then they call for specific environmental models that can calculate the size and density of the cloud created, as well as its drift and dispersion over time.

Perception

Military objects move about the environment in order to interact with other objects. One of the first steps in this interaction is to perceive or detect other objects. This is the process of applying a sensor to detect the existence, position, and identification of the other object. Sensor models capture the signatures of those objects. A visual sensor will capture reflected light from an object to the sensor. In most cases, the sensor model does not actually represent the path of a light vector, but instead considers the range and orientation between the target object and the sensor and calculates whether the target is potentially detectable based on the effective range and field-of-view of the sensor. A sensor model may also include information about the environment in which the detection is being attempted. For a visual sensor, atmospheric factors like the presence of smoke, dust, fog, and lighting may be used to diminish the possibility of detection. Also, environmental features like hills, trees, and buildings may be interposed between the target and the sensor and impact the detection of an object. The physical characteristics of the target may also be considered. Its size, contrast with the background, movement, and composition may significantly impact its detectability. Larger targets may be easier to see than smaller ones. Targets may have a higher or lower degree of camouflage, changing the ability of the sensor to separate them from the background image.

In military simulations, visual sensors are just one of a large variety of sensors that are available. Many systems include sensor models that collect signature information in the infrared spectrum, sound, emitted radio and radar signals, magnetic properties, and movement and vibrations. Models of each of these can be constructed at a number of different levels of detail, but each must determine whether to include the properties of the sensor, sensing platform, paired geometry, environment, target, and external interference. As illustrated earlier, as the sensor model becomes more complex, it drives the complexity of the entire system. Including all of the categories just listed would trigger the need for additional detail in the sensor model, but also the need for additional details in all target objects and the environment. Often the limitation in creating a high-fidelity sensor model is not driven by our understanding of the sensor, but, rather, by our ability to represent the characteristics of the target and environment that are needed to create such a model. In a military simulation system, the detail included in a model may be limited both by the needs of the customer and by the desire to keep the entire system balanced, not allowing one model to drive all of the others into become larger and more detailed (Pritsker, 1990).

Exchange

After moving and detecting, models are needed to allow objects to exchange materials and information with each other. Battlefield operations often lead to the depletion of materials like fuel, ammunition, food, medical supplies, and people. A logistics model may be used to represent the ability of the military to constantly deliver these materials to objects in operations. Such models are often based on an understanding of the rates of consumption, the pre-deployment of supplies to locations that are close to the operation, and the constant replenishment of supplies through a network of supply nodes. Replenishing supplies within n object on the battlefield is the culminating model of a much more complex representation of the logistics infrastructure that can stretch across an entire country or even around the world. The logistics model must also include mistakes and interference that cause it to breakdown and deprive the military objects of the supplies that keep them operating. A logistics model may be driven by textbook ratios of consumption or it may be include specific messages from the military objects about the levels of supplies consumed. In the latter case, a communications model is needed to carry information about what materials that are being consumed, by whom, and where they are located.

Communication is another model of exchange. The thing being exchanged is information rather than physical items. In the modern military, the amount of information that is carried around in a physical form, such as a book, letter, or paper map, is quite small compared to the amount that is transmitted in digital form. Therefore, modern models focus on communications in the form of digital computer and analog radio networks. A model of radio communications, like that of a sensor, may include the characteristics of the transmitter, transmitting platform, environment, the receiver, the geometry between the sender and receiver, and interference by other objects. Details in the representation of the radio or the signal it generates call for corresponding details in the receivers, environment, and countermeasures.

Military models of digital computer communications are similar to the tools used to study Internet traffic. They represent the senders, receivers, relay nodes, interference from competing traffic, multiple paths for the information to travel, and the loss of a message or the failure of a network. Modeling how people, objects, and units respond to the receipt of this information is included in the section on reasoning.

Engagement

Engagement is strictly a form of exchange. The item being exchanged is a weapon and the effect is the degradation of the operational capabilities of the target. Most military simulations perform movement, perception, and exchange specifically so they can put themselves in a position to engage an enemy target. Engagement has historically been the pivotal centerpiece of a simulation system and one of the most important models in the system. Certainly, not all objects engage the enemy, but those that do not are often referred to as support elements whose mission is to make engagement possible for combat equipped units (Smith, 2000).

An engagement model typically includes the exchange of weapons or firepower from a shooter to a target. This exchange decrements the capability of the shooter by expending ammunition in one of its many forms (e.g. bullets, missiles, bombs, rockets, grenades, artillery rounds). Just as in the perception and communication models described above, this exchange is usually impacted by the geometry between the shooter and the target. The engagement may also be mitigated by the environment and characteristics of the target. Trees, terrain, water, and buildings may interfere with the optimal delivery of the weapon and reduce its impact on the target. The target may also contain defensive systems that counter the effects of the engagement. A defensive model may represent the effects of flares or chaff in deceiving and misleading a guided missile or the protective effects of armor to deflect the weapon.

If the weapon successfully impacts the target and is powerful enough to overcome any interference or defenses, then a level of attrition must be calculated for the target. Different approaches to modeling attrition are described in the next section. Attrition is usually directed at the model state variables that control its ability to perform its primary functions. These may include health or strength, fuel levels, communications capabilities, and mobility. Models may also make a binary decision about whether a vehicle, human, or unit it completely destroyed or not.

The attrition model may be linked to communications and medical models. Communications models propagate the outcome of an engagement so that units or operators are aware that an engagement has occurred. These communications may trigger a medical model that will attempt to conduct extraction and provide medical treatment to simulated humans that are wounded. It may also trigger the logistics model to extract and repair vehicles.

Reasoning

Within large military simulation systems, there are usually many models of human decision-making and behaviors. These have become more prevalent as systems have grown in both the breadth of coverage and the depth of detail of the battlefields that are represented. Representing human thinking and even some computer reasoning are one of the most challenging parts of the current practice of military modeling. This type of information processing is largely not understood and general approximations and simplifications are captured in models.

Reasoning models often rely on the techniques developed within the Artificial Intelligence field. Techniques like finite state machines, expert systems, rule-based systems, case based reasoning, neural networks, fuzzy logic, means-ends analysis, and others are used to organize information and create decisions that are similar to those of living humans. FSM are currently the most widely used technique for both military models and those inside of commercial games. These reasoning models are challenged to perform a wide array of operations, to include commanding subordinate units, decomposing and acting on commands from higher level units, reacting to enemy attacks, selecting maneuver routes, identifying threats and opportunities for engagement, fusing information, and extracting meaning from intelligence reports. Each of these functions can be extremely complicated and require significant computing resources to execute. Reasoning models must balance their level of realism between robotic reactions to stimuli and detailed consideration of the situation prior to selecting an action.

The variety of reasoning models that are required on a battlefield cannot be fit to a single modeling technique. In practice, multiple techniques are required, each applied to a reasoning problem for which it is best suited.

Dynamic Environment

Earlier we described the evolution of the simulated environment from static state structures to dynamic representations of features and their interactions with military objects. Military objects interact with the environment both through direct intention and through accidental collocation. An engineering unit may be tasked to destroy a bridge or a road. This is an operation in which the effects on the environment are the specific intent of the action. In another case, an aircraft may bomb a convoy of trucks moving on a road. In this case, the trucks are the primary targets, but the road may sustain damage because of its collocation with the trucks.

Until recently, military simulations seldom included impacts on the environment. However, with the current focus on precision operations, there is much more interest in destroying specific buildings, roads, bridges, communications equipment, and pieces of the social infrastructure. Since this data is usually found in the environmental database, models that accurately modify environmental information are needed.

For decades, military organizations have worked on models that accurately represent the engagement that takes place between two tanks, airplanes, or ships. It is becoming necessary for those models to also impact the trees, terrain, and roads in the vicinity of these engagements. This means that information on the effects of weapons on trees is necessary, as well as their effects on buildings, roads, bridges, and a host of other types of surrounding terrain.

Even though these models are making the environment a dynamic part of the simulation system, the type and level of damage done to a tree is seldom the focus of the experiment or exercise that is being conducted. Therefore, the detail in these models is not as critical that in the models that govern the dynamic changes to other military objects.