Modeling state-dependent systems can quickly become unwieldy as the number of states and transitions increase. This paper introduces a reusable approach to modeling state transitions that results in simpler models and allows designers to focus on a design's functional behavior rather than the details of state transition. Introduction A signi cant challenge associated with modeling state transitions, in both UML and CPN, is that the complexity of the models tends to grow exponentially [1]. E orts to model state transition behavior typically result in one-to-one conversions of UML elements into arcs, transitions and places [4, 1], resulting in complex and unweildy models. This paper introduces an approach to modeling state transitions that results in simpler models and allows designers to focus on functional behavior rather than the details of state transition. A reusable component is modeled in CPN Tools [2]. Related Research Andre, Benmoussa and Choppy proposed a formalization of UML state machines using Colored Petri Nets [1]. Meghzili, Chaoui, Strecker and Kerkouche presented an approach to transforming State Machine Diagrams into CPN models, and proved certain structural properties in the transformation [3]. Pettit and Gomaa described an approach to systematically map state dependent objects and their corresponding state charts from UML into CPN [4]. This work di ers from others in that state transition behavior is abstracted into a reusable CPN component that de nes state transitions and their associated actions in metadata, requiring no complex network de nition. Modeling State Transitions UML is the de facto standard for modeling systems [1], but it is not executable. Petri nets are executable, however models of state transitions become increasingly complex because they are typically modeled using an explicit one-to-one correspondence with the State Transition Diagram. CPN's hierarchical capabilities facilitate the building of models bene ting from abstraction and separation of concerns. The functional behavior associated
1`"Insert Disk" @10 +++ EVENT 1`"Play" @20 +++ 1`"Next" @30 +++ 1`"Prev" @40 +++ 1`"Pause" @50 +++ 1`"Pause" @60 1`"OFF" Current
State STATE All
Trans
TRANS
State Machine 1`[{s="OFF", e="Insert Disk", n="READY"}, {s="READY", e="Play", n="PLAYING"}, {s="READY", e="Eject", n="OFF"}, {s="PLAYING", e="Eject", n="OFF"}, {s="PLAYING", e="Next", n="PLAYING"}, {s="PLAYING", e="Prev", n="PLAYING"}, {s="PLAYING", e="Pause", n="PAUSED"}, {s="PAUSED", e="Pause", n="PLAYING"}, {s="PAUSED", e="Eject", n="OFF"}] with processing events, issuing actions, and changing state can therefore be encapsulated within a reusable CPN, allowing the supported event-driven state changes to be de ned in dynamically programmable meta data. In other words, rather than de ne a new net for each state driven scenario, each new scenario can simply be programmed into the reusable state machine. Figure 1 depicts the external view of the reusable state machine, with events and transitions de ned to model a CD player.
Conclusions and Future Work The approach simpli es modeling in a number of ways. For example, the reusable State Machine eliminates the the need to explicitly de ne the state transition relationships with a massive number of arcs and transitions. The designer need only de ne the transition data, and then focus on the functional aspects of their design.
This paper addresses state transitions in only the most general sense. That is where the modeling complexity is rst and most apparent. The incorporation of the Entry, Exit and Do actions associated with each state results in similar complexity, and early e orts to model them using the same simpli ed approach described above show similar improvements. Therefore, the main future work includes incorporating transitional actions.