As mentioned earlier, fidelity refers to the realism or degree of similarity between the training situation and the operational situation being simulated. The two basic measures of fidelity are physical and functional characteristics of the training situation (Hays and Singer, 1989). In the case of a manned model, the model contributes to both. In the case of computer-based simulation, the mathematical model contributes to functional characteristics. Fidelity is determined subjectively. The level of fidelity required is determined by the training objectives, which, in turn, are based on task needs and training analysis (Chapter 3; Hays and Singer, 1989). Determinant measures may be used to aid in assessing the level of fidelity in a given simulation. Accuracy is inherently a determinant measure of how close something is to being exact. The accuracy of a trajectory prediction model is determined by measuring variations of the predicted trajectory with the actual trajectory.
In many respects, fidelity is more difficult to address than accuracy because it involves a subjective assessment of how real the simulation is. Balancing accuracy of trajectory modeling with fidelity of motion in visual scenes, for example, is very challenging. It is possible to provide a believable simulation using a simple trajectory model that, with a few minor validating adjustments, can appear to be realistic to be realistic to pilots and mariners in the specific harbor/ship situation. Yet, performing slightly different maneuvers than those used for validation can result in quite inaccurate trajectories. Indeed, all models have limited accuracy in various modes of which the trainer may be unaware. In general, this issue has not been addressed by simulation providers except to try to use the most accurate modeling approach economically available.
The accuracy of trajectory prediction models available to drive a simulation can be compared with the level of fidelity specified by the training analysis as necessary to achieve training objectives. The accuracy of trajectory prediction, for instance, is less important in courses where vessel maneuvering behavior is not an instructional objective than in courses where maneuvering is required to achieve the goal of certain learning situations or is the primary instructional objective.
It is sometimes possible to enhance training effectiveness by departing from realism. As a general rule, in marine simulation, departures from realism are driven by limitations in training resources, rather than a conscious attempt to optimize training effectiveness. The most notable exception is the initial development