This chapter defines language elements to define clocked state machines. These state machines have a similar modeling power as Statecharts (Harel, 1987) and have the important feature that at one clock tick, there is only one assignment to every variable (for example, it is an error if state machines are executed in parallel and they assign to the same variable at the same clock tick; such errors are detected during translation). Furthermore, it is possible to activate and deactivate clocked equations and blocks at a clock tick. An efficient implementation will only evaluate the equations and blocks that are active at the current clock tick. With other Modelica language elements, this important feature cannot be defined.
The semantics of the state machines defined in this chapter is inspired by mode automata and is basically the one from Lucid Synchrone 3.0 (Pouzet, 2006). Note, safety critical control software in aircrafts is often defined with such kind of state machines. The following properties are different to Lucid Synchrone 3.0:
Lucid Synchrone has two kinds of transitions: strong and weak transitions. Strong transitions are executed before the actions of a state are evaluated and weak transitions are executed after the actions of a state are evaluated. This can lead to surprising behavior, because the actions of a state are skipped if it is activated by a weak transition and exited by a true strong transition.
For this reason, the state machines in this chapter use immediate (= the same as strong) and delayed transitions. Delayed transitions are immediate transitions where the condition is automatically delayed with an implicit previous.
Parallel state machines can be explicitly synchronized with a language element (similarly as parallel branches in Sequential Function Charts). This often occurring operation can also be defined in Statecharts or in Lucid Synchrone state machines but only indirectly with appropriate conditions on transitions.
Modelica blocks can be used as states. They might contain clocked equations. If the equations are discretized, they are integrated between the previous and the current clock tick, if the corresponding state is active.
Any Modelica block instance without continuous-time equations or continuous-time algorithms can potentially be a state of a state machine. A cluster of instances which are coupled by transition statements makes a state machine. All parts of a state machine must have the same clock. All transitions leaving one state must have different priorities. One and only one instance in each state machine must be marked as initial by appearing in an initialState statement.
The special kinds of connect-like equations listed below are used to define define a state machine.
| Expression | Description | Details |
|---|---|---|
| transition(, , , ) | State machine transition between states | Operator 17.1 |
| initialState() | State machine initial state | Operator 17.2 |
The transition- and initialState-equations can only be used in equations, and cannot be used inside if-equations with conditions that are not parameter expressions, or in when-equations.
The operators listed below are used to query the status of the state machine.
| Expression | Description | Details |
|---|---|---|
| activeState() | Predicate for active state | Operator 17.3 |
| ticksInState() | Ticks since activation | Operator 17.4 |
| timeInState() | Time since activation | Operator 17.5 |
Arguments and are block instances, and is a Boolean argument. The optional arguments immediate, reset, and synchronize are of type Boolean, have parameter variability and a default of true, true, false respectively. The optional argument priority is of type Integer, has parameter variability and a default of 1.
This operator defines a transition from instance to instance . The and instances become states of a state machine. The transition fires when if (this is called an immediate transition) or previous() when (this is called a delayed transition). Argument priority defines the priority of firing when several transitions could fire. In this case the transition with the smallest value of priority fires. It is required that and that for all transitions from the same state, the priorities are different. If , the states of the target state are reinitialized, i.e., state machines are restarted in initial state and state variables are reset to their start values. If , any transition is disabled until all state machines of the from-state have reached final states, i.e., states without outgoing transitions. For the precise details about firing a transition, see section 17.3.
Argument is the block instance that is defined to be the initial state of a state machine. At the first clock tick of the state machine, this state becomes active.
Argument is a block instance. The operator returns true if this instance is a state of a state machine and this state is active at the actual clock tick. If it is not active, the operator returns false.
It is an error if the instance is not a state of a state machine.
Returns the number of ticks of the clock of the state machine for which the currently active state has maintained its active state without interruption, i.e., without local or hierarchical transitions from this state. In the case of a self-transition to the currently active state or to an active enclosing state, the number is reset to one.
This function can only be used in state machines.
Returns the time duration as Real in [s] for which the currently active state has maintained its active state without interruption, i.e., without local or hierarchical transitions from this state. In the case of a self-transition to the currently active state or to an active enclosing state, the time is reset to zero.
This function can only be used in state machines.
[Example: If there is a transition with immediate = false from state A1 to A2 and the condition is ticksInState() >= 5, and A1 became active at 10ms, and the clock period is 1ms, then A1 will be active at 10ms, 11ms, 12ms, 13ms, 14ms, and will be not active at 15 ms.
]
[Figure 17.1 shows the recommended layout of a state machine.]
The annotation for graphics of transition has the following structure: annotation(Line(), Text()); and for initialState(): graphical-primitives(Line()); with Line and Text annotations defined in chapter 18.
[Example:
]
The Text annotation representing the transition condition can use the notation %condition to refer to the condition expression.
The extent of the Text is interpreted relative to either the first point of the Line, in the case of immediate = false, or the last point (immediate = true).
In addition to the line defined by the points of the Line annotation, a perpendicular line is used to represent the transition. This line is closer to the first point if immediate = false otherwise closer to the last point.
If the condition text is somewhat distant from the perpendicular line, a dimmed straight line joins the transition text and the perpendicular line. (See the rightmost transition above.)
If reset = true, a filled arrow head is used otherwise an open arrow head. For synchronize = true, an inverse “fork” symbol is used in the beginning of the arrow. (See the rightmost transition above.)
The value of the priority attribute is prefixing the condition text followed by a colon if priority > 1.
The initialState line has a filled arrow head and a bullet at the opposite end of the initial state (as shown above).
For the purpose of defining the semantics of state machines, assume that the data of all transitions are stored in an array of records:
The transitions are sorted with lowest priority number last in the array; and the priorities must be unique for each value of from. The states are enumerated from 1 and up. The transition conditions are stored in a separate array c[:] since they are time varying.
The semantics model is a discrete-time system with inputs {c[:], active, reset} with t being an array corresponding to the inputs to the transition operator, outputs {activeState, activeReset, activeResetStates[:]} and states {nextState, nextReset, nextResetStates[:]}. For a top-level state machine, active is always true. For sub-state machines, active is true only when the parent state is active. For a top-level state machine, reset is true at the first activation only. For sub-state machine, reset is propagated from the state machines higher up.
The state update starts from nextState, i.e., what has been determined to be the next state at the previous time. selectedState takes into account if a reset of the state machine is to be done.
The integer fired is calculated as the index of the transition to be fired by checking that selectedState is the from-state and the condition is true for an immediate transition or previous() is true for a delayed transition. The max function returns the index of the transition with highest priority or 0.
The start value of c is false. This definition would require that the previous value is recorded for all transitions conditions. Below is described an equivalent semantics which just require to record the value of one integer variable delayed.
The integer immediate is calculated as the index of the immediate transition to potentially be fired by checking that selectedState is the from-state and the condition is true. The max function returns the index of the transition with true condition and highest priority or 0.
In a similar way, the Integer delayed is calculated as the index for a potentially delayed transition, i.e., a transition taking place at the next clock tick. In this case the from-state needs to be equal to nextState:
The transition to be fired is determined as follows, taking into account that a delayed transition might have higher priority than an immediate:
nextState is set to the found transitions to-state:
In order to define synchronize transitions, each state machine must determine which are the final states, i.e., states without from-transitions and to determine if the state machine is in a final state currently:
To enable a synchronize transition, all the stateMachineInFinalState conditions of all state machines within the meta state must be true. An example is given below in the semantic example model.
A state can be reset for two reasons:
The whole state machine has been reset from its context. In this case, all states must be reset, and the initial state becomes active.
A reset transition has been fired. Then, its target state is reset, but not other states.
The first reset mechanism is handled by the activeResetStates and nextResetStates vectors.
The state machine reset flag is propagated and maintained to each state individually:
until a state is eventually executed, then its corresponding reset condition is set to false:
The second reset mechanism is implemented with the selectedReset and nextReset variables. If no reset transition is fired, the nextReset is set to false for the next cycle.
When a state is suspended, its equations should not be executed, and its variables keep their values – including state-variables in discretized equations.
The execution of a sub-state machine has to be suspended when its enclosing state is not active. This activation flag is given as a Boolean input active. When this flag is true, the sub-state machine maintains its previous state, by guarding the equations of the state variables nextState, nextReset and nextResetStates.
The entire semantics model is given below:
[When a state class uses an outer output declaration, the equations have access to the corresponding variable declared inner. Special rules are then needed to maintain the single assignment rule since multiple definitions of such outer variables in different mutually exclusive states needs to be merged.]
In each state, the outer output variables are solved for and for each such variable a single definition is formed:
last is special internal semantic operator returning its input. It is just used to mark for the sorting that the incidence of its argument should be ignored. A start value must be given to the variable if not assigned in the initial state.
A new assignment equation is formed which might be merged on higher levels in nested state machines.
[The causal connection semantics of Modelica for non-state machines are generalized to states of state machines, using the fact that only one state is active at a time.]
It is possible to connect outputs each coming from different states of state machines together – and connect this with other causal connectors. These outputs are combined seen as one source of the signal, and give the following constraint equations,
with yi being outputs from different states of the state-machine and ui being other causal variables. The semantics is defined similarly to section 17.3.5:
[Example: Consider the hierarchical state machine in figure 17.2. The model demonstrates the following properties:
state1 is a meta state with two parallel state machines in it.
stateA declares v as outer output. state1 is on an intermediate level and declares v as inner outer output, i.e., matches lower level outer v by being inner and also matches higher level inner v by being outer. The top level declares v as inner and gives the start value.
count is defined with a start value in state1. It is reset when a reset transition (v >= 20) is made to state1.
stateX declares the local variable w to be equal to v declared as inner input.
stateY declares a local counter j. It is reset at start and as a consequence of the reset transition (v >= 20) to state1: When the reset transition (v >= 20) fires, then the variables of the active states are reset immediately (so count from state1, and i from stateX). The variables of other states are only reset at the time instants when these states become active. So j in StateY is reset to 0, when the transition stateX.i > 20 fires (after state1 became active again, so after the reset transition v >= 20).
Synchronizing the exit from the two parallel state machines of state1 is done by checking that stated and stateY are active using the activeState function.
The Modelica code (without annotations) is:
Figure 17.3 shows the behavior of the state machine.
The transition from state1 to state2 could have been done with a synchronize transition with condition=true instead. The semantically equivalent model is shown below:
]