Systems controlled through state machines can often be de-composed into largely separate sub-systems. It is definitely a reasonable strategy to model the sub-systems with separate state machines and coordinate them through events.

However the state chart formalism provides the possibility to model such systems within the same state machine diagram through the concept of orthogonal sub-regions. This allows to keep the system description on the visual level.

An orthogonal sub-region is an additional separate region within a composite state. Entering the composite state enters all sub-regions collectively. The fundamental extension is that the state machine may now be in multiple states at the same time. This introduces a further level of possibilities for transition source and target states within the state machine.

The below state diagram has a state with two orthogonal regions and shows a few typical cases for transitions.

Figure 1: State machine diagram with orthogonal regions, fork transitions and join transitions.

Figure 2: Tree structure for the above state machine with orthogonal regions.

All transitions that target the composite state ultimately bring the state machine into a state where two states are active at the same time:

If the transition forks and ends at individual sub-states, the intended states to enter are obvious. If the transition ends at the composite state border, the initial states of the involved sub-regions are entered. And finally a transition, that explicitly targets only a subset of the sub-regions while leaving other sub-regions unspecified, is executed by entering the initial state for each unspecified sub-region and entering the defined states for all directly targeted sub-regions.

Similar situations arise for transitions that leave composite states with orthogonal sub-regions:

If the transition joins from the individual sub-states it will be considered as reached when all of these sub-states are active. The transition can then execute if the corresponding event arrives or the guard condition is fulfilled. If the transition starts on the composite state border, it will be considered as reached if the state machine is in any configuration of states of the orthogonal sub-regions. And finally, if only some of the orthogonal sub-regions are defined as source states while other sub-regions are unspecified, the transition is considered as reached when the configuration of defined sub-states matches while the other unspecified sub-regions are not considered and may be in any state.

Transitions may also be both join and fork in case the source state and the target state are composite states with orthogonal sub-regions.

As all these transitions enter or leave multiple states, the state machine also executes the corresponding entry and exit actions of the sub-states when the transition is executed. The sequence of the sub-regions in the code determines the sequence how entry actions are executed. Exit actions are executed in reverse order.

Orthogonal sub-regions of a composite state are added through multiple Region..EndRegion statement blocks, embedded within the State..EndState statement pair of the composite state, after any Transition statements of the composite state.

The Transition statement accepts arrays of state names instead of a single state name, in order to specify individual sub-states within orthogonal sub-regions. Arrays of state names can be passed in for both the source and the target state of the transition.

The source state must be specified in case it references a sub-state of the composite state where the transition belongs to.

StateMachineTemplate t = new StateMachineTemplate();
t.Region(StateA, false);
    t.State(StateA, null, null);
        t.Transition(Transit1, new string[] {StateB1B, StateB2B}, Event1, null, null);
        t.Transition(Transit2, StateB, Event2, null, null);
        t.Transition(Transit3, StateB1B, Event3, null, null);
    t.EndState();
    t.State(StateB, null, null);
        t.Transition(Transit4, new string[] {StateB1B, StateB2B}, StateC, Event4, null, null);
        t.Transition(Transit5, StateC, Event5, null, null);
        t.Region(StateB1A, false);
            t.State(StateB1A, null, null);
                t.Transition(Transit99, StateB1B, null, null, null);
            t.EndState();
            t.State(StateB1B, null, null);
                t.Transition(Transit6, StateC, Event6, null, null);
            t.EndState();
        t.EndRegion();
        t.Region(StateB2A, false);
            t.State(StateB2A, null, null);
                t.Transition(Transit98, StateB2B, null, null, null);
            t.EndState();
            t.State(StateB2B, null, null);
            t.EndState();
        t.EndRegion();
    t.EndState();
    t.State(StateC, null, null);
    t.EndState();
t.EndRegion();

In case a sub-state of an orthogonal sub-region is itself again a composite state, then all entry actions of this composite state are executed before any entry actions of the next sibling orthogonal sub-region.

Unlike the other transitions, Transit4 explicitly specifies the source states StateB1B and StateB2B to indicate that the transition starts from these sub-states.

Note

The OMG UML Specification describes the possibility to build transitions as a directed acyclic graph with multiple intermediate vertices including choice and junction vertices, named compound transitions. While this provides an enormous flexibility and comfort for the possibilities to define transitions, there are also downsides like the undefined precedence of execution in case of conflicting transition paths. StaMa provides a rigorous reduced model, where each transition has a single event signal and guard condition per transition but still allows to specify source and target states of orthogonal sub-regions.