Simulation

Simulation consists of constructing a model of an existing system to be studied or a system to be built and then executing actions allowed in this model. The model can be a physical entity like a scale clay model of an airplane or a computer representation. A computer model is often less costly than a physical model and can represent a noncomputer entity such as an airplane or its components as well as a computer entity such as a computer system or a program. A computer model also can represent a system with both computer and non-computer components like an automobile with embedded computer systems to control its transmission and brakes.

This physical or computer model is called the simulator of the actual system. A simulator can carry out simulated executions of the simulated system and display the outcomes of these executions. A physical model of an airplane in a wind tunnel shows the aerodynamics of the simulated plane that is close to the actual plane. A software simulator on a single-processor system shows the performance of a network of personal computer workstations under a heavy network traffic condition. A software simulator can also be designed to simulate the behavior of an automobile crashing into a concrete barrier, showing its effects on the automobile’s simulated occupants. Sometimes a simulator refers to a tool that can be programmed or directed without programming to mimic the events and actions in different systems. This simulator can be either computer-based (software, hardware, or both) or noncomputer- based.

Simulation is an inexpensive way to study the behavior of the simulated system and to study different ways to implement the actual system. If we detect behavior or events that are inconsistent with the specification and safety assertions, we can revise the model and thus the actual system to be built. In the case in which we consider several models as possible ways to implement the actual system, we can select the model that best satisfies the specification and safety assertions through the simulation and then implement it as the actual system.

Different levels of details of the actual system, also called the target system, can be modeled and its events simulated by a simulator. This makes it possible to study and observe only the relevant parts of the target system. For example, when designing and simulating the cockpit of an aircraft, we can restrict attention to that particular component by simulating only the cockpit with inputs and outputs to the other aircraft components, without simulating the behavior of the entire aircraft. In the design and simulation of a real-time multimedia communication system, we can simulate the traffic pattern between workstations but need not simulate the low-level signal processing involved in the coding and transmission processes if the performance is not affected by these low-level processes. The ability to simulate a target system at different detail levels or only a subset of its components reduces the resources needed for the simulation and decreases the complexity of the analysis of the simulation.

There are several simulation techniques in use, such as real-time-event simulation and discrete-event simulation. A real-time-event simulator like a physical scale model of an automobile performs its actions in real-time and the observable events are recorded in real-time. An example is the crash testing of the physical model of an automobile. Such a simulator requires recording instruments capable of recording events in real-time. When the physical model is actually an implemented target system, this is no longer a simulation but rather a testing of the actual system.

A discrete-event simulator, on the other hand, uses a logical clock(s) and is usually software-based. The variety of systems that can be represented by such a simulator is not limited by the speed at which the hardware executes the simulator since the simulated actions and events do not occur in real-time. Rather, they take place according to the speed of the simulator hardware and the instructions in the simulator program. Examples include the simulation of a network of computers in a single-processor system or the simulation of a faster microprocessor in a slower processor as in the design of popular next-generation personal computer microprocessors. Here, the appropriate actions and events “occur” at a particular logical time (representing real-time in the target system) depending on previous and current simulated actions and events.

Variations of simulation include the hybrid simulation approach, in which the simulator works with a partial implementation of the target system by acting as the non-implemented part. As in the case of using only a simulator, the simulator here makes it possible to predict the performance and behavior of the target system once it is completely implemented.

One main disadvantage of simulation as a technique to analyze and verify realtime systems and other systems is that it is not able to model all possible event-action sequences in the target system where the domain of possible sequences of observable events is infinite. Even when this domain is finite, the number of possible events is so large that the most powerful computer resources or physical instruments may not be able to trace through all possible sequences of events in the simulated target system.