12th IFAC/IEEE Workshop on Discrete Event Systems

Cachan, France. May 14-16, 2014

On Relative Observability of Timed

Discrete-Event Systems
Kai Cai ∗ Renyuan Zhang ∗∗ W.M. Wonham ∗∗∗
∗
Urban Research Plaza, Osaka City University, Japan. This work was
supported in part by Program to Disseminate Tenure Tracking System,
MEXT, Japan. kai.cai@ur-plaza.osaka-cu.ac.jp.
∗∗
Department of Traffic and Control Engineering, Northwestern
Polytechnical University, China. ryzhang@nwpu.edu.cn.
∗∗∗
Department of Electrical and Computer Engineering, University of
Toronto, Canada. This work was supported in part by the Natural
Sciences and Engineering Research Council, Canada, Grant no. 7399.
wonham@control.utoronto.ca.

Abstract: We study new observability concepts in supervisory control of timed discrete-event

systems (DES) under partial observation. Known concepts in the literature are timed observ-
ability and weak observability; neither, however, is closed under set union, and consequently
the supremal (weak) observable sublanguage of a given language need not exist in general. In
the first part of this paper, we extend our previous work on relative observability to timed
DES: a fixed, ambient language is given, relative to which timed observability is tested. Relative
observability is stronger than timed observability, weaker than normality, and closed under set
union. An algorithm is presented for computing the supremal relatively observable sublanguage
of a given language. In the second part, we identify relative weak observability by exploiting
forcible events to preempt the clock event tick. Relative weak observability is stronger than
weak observability, weaker than normality, and closed under set union; we present an algorithm
for computing the supremal relatively weakly observable sublanguage of a given language.

Keywords: Timed supervisory control, partial observation, relative observability, relative weak
observability, timed discrete-event systems

1. INTRODUCTION need not exist the supremal weakly observable sublan-

Recently we introduced the new concept of relative observ-
ability (Cai et al. (2013)) in supervisory control of discrete-
event systems (DES) under partial observation. Relative
observability is stronger than observability, weaker than
normality, and preserved under set union; hence there
exists the supremal relatively observable sublanguage of
a given language. In this paper, we study relative observ-
ability of timed DES in the Brandin-Wonham framework
(Brandin and Wonham (1994); Wonham (2013)).
Observability is a familiar concept in untimed supervi-
sory control under partial observation: Lin and Wonham
(1988), Cieslak et al. (1988). This concept was extended
(Lin and Wonham (1995)) to timed supervisory control in
the Brandin-Wonham framework. Like its untimed coun-
terpart, timed observability is not preserved under set
union; consequently, the supremal observable sublanguage
of a given language need not exist in general.
Timed observability was further studied in Takai and
Ushio (2006), where the concept of weak observability
was proposed for a distinct class of timed supervisors. In
particular, the observability requirement for the special
event tick is relaxed by exploiting choices of forcible events
(formal definitions are given below). Weak observability,
however, is also not preserved under set union, and there
guage of a given language.
In this paper, we introduce relative observability and
relative weak observability for timed supervisory control,
extending respectively Lin and Wonham (1995) and Takai
and Ushio (2006). First, we introduce timed relative ob-
servability, and prove that it is stronger than timed ob-
servability, weaker than normality, and closed under set
union. An algorithm is presented for computing the supre-
mal relatively observable (and controllable, Lm (G)-closed)
sublanguage of a given language. Second, we introduce
relative weak observability, and prove that it is stronger
than weak observability, weaker than normality, and closed
under set union. We present an algorithm for computing
the supremal relatively weakly observable sublanguage of a
given language. The concepts proposed and the relations
proved, together with those of Lin and Wonham (1995)
and Takai and Ushio (2006), are summarized in Fig. 1.
The rest of the paper is organized as follows. In Section 2
we review the basics of the Brandin-Wonham timed su-
pervisory control framework. Section 3 introduces timed
relative observability and investigates its properties; an
algorithm is presented to compute the supremal relative
observable sublanguage. Section 4 introduces relative weak
observability and studies its properties and computational
issue. Finally in Section 5 we state our conclusions.

978-3-902823-61-8/2014 © IFAC 208 10.3182/20140514-3-FR-4046.00005

WODES 2014
Cachan, France. May 14-16, 2014
normality
relative observability relative weak observability
observability weak observability
(Lin and Wonham, 1995) (Takai and Ushio, 2006)
Fig. 1. Observability concepts and their relations in super-
visory control of timed DES under partial observation.
2. PRELIMINARIES ON TIMED DES
This section reviews the timed DES (TDES) model pro-
posed by Brandin and Wonham (Brandin and Wonham
(1994); Wonham (2013),Chapter 9). First consider the
untimed DES model
Gact = (A, Σact , δact , a0 , Am ). (1)
Here A is the finite set of activities, Σact is the finite set
of events, δact : A × Σact → A is the (partial) activity
transition function, a0 ∈ A is the initial activity, and
Am ⊆ A is the set of marker activities. Let N denote
the natural numbers {0, 1, 2, ...}. We introduce time into
Gact by assigning to each event σ ∈ Σact a lower time
bound lσ ∈ N and an upper time bound uσ ∈ N ∪ {∞},
such that lσ ≤ uσ ; typically, lσ represents a delay in
communication or in control enforcement, while uσ is often
a hard deadline imposed by legal specification or physical
necessity. With these assigned time bounds, the event set
˙ rem (∪˙
Σact is partitioned into two subsets: Σact = Σspe ∪Σ
denotes disjoint union) with Σspe := {σ ∈ Σact |uσ ∈ N}
and Σrem := {σ ∈ Σact |uσ = ∞}; here “spe” denotes
“prospective”, i.e. σ will occur within some prospective
time (with a finite upper bound), while “rem” denotes
“remote”, i.e. σ will occur at some indefinite time (with
no upper bound), or possibly will never occur at all.
A distinguished event, written tick, is introduced which
represents “tick of the global clock”. Attach to each event
σ ∈ Σact a (countdown) timer tσ ∈ N, whose default value
tσ0 is set to be
½ is one that preempts event tick: if, at a state q of G,
uσ if σ ∈ Σspe ,
tσ0 := (2)
lσ if σ ∈ Σrem .
When timer tσ > 0, it decreases by 1 (counting down) if
event tick occurs; and when tσ = 0, event σ must occur
(resp. may occur) if σ ∈ Σspe (resp. if σ ∈ Σrem ). Note
that while tick is a global event, each timer tσ is local
(with respect to σ). Also define the (integer) timer interval
Tσ by ½ Here designating both Σhib and tick controllable is to
[0, uσ ] if σ ∈ Σspe ,
Tσ := (3)
[0, lσ ] if σ ∈ Σrem .
Based on (1)-(3), the TDES model G is given by
G := (Q, Σ, δ, q0 , Qm ), (4)
Q Q
where Q := A × {Tσ |σ ∈ Σact } ( denotes Cartesian
product) is the finite set of states, a state q ∈ Q being of
the form q = (a, {tσ |σ ∈ Σact }) (i.e. a (1 + |Σact |)-tuple);
˙
Σ := Σact ∪{tick} is the finite set of events; δ : Q × Σ → Q
is the (partial) state transition function; q0 = (a0 , {tσ0 |σ ∈
Σact }) (tσ0 as in (2)) is the initial state; and Qm ⊆ Am ×
209
Q
{Tσ |σ ∈ Σact } is the set of marker states. Starting from
q0 , TDES G executes state transitions in accordance with
its transition function δ. Let q = (a, {tα |α ∈ Σact }) ∈ Q
and σ ∈ Σact ; δ is defined at (q, σ), written δ(q, σ)!, if δact
of Gact is defined at (a, σ) (i.e. δact (a, σ)!) and timer tσ
satisfies (i) 0 ≤ tσ ≤ uσ − lσ when σ ∈ Σspe , and (ii)
tσ = 0 when σ ∈ Σrem . The new state q 0 = δ(q, σ) is given
by q 0 = (δact (a, σ), {t0α |α ∈ Σact }), where t0σ is set to be
its default value tσ0 as in (2); for other timers tα , α 6= σ,
the reader is referred to detailed updating rules given in
Brandin and Wonham (1994); Wonham (2013),Chapter 9.
On the other hand, δ(q, tick)! if no timer of a prospective
event is zero, and q 0 = δ(q, tick) = (a, {t0α |α ∈ Σact }), i.e.
there is no change in the activity component a of q, while
the rules for updating timers are again referred to Brandin
and Wonham (1994); Wonham (2013),Chapter 9.
Let Σ∗ be the set of all finite strings of elements in Σ =
˙
Σact ∪{tick}, including the empty string ². We introduce
the languages generated by TDES G in (4). The transition
function δ is extended to δ : Q × Σ∗ → Q in the usual way.
The closed behavior of G is the language L(G) := {s ∈
Σ∗ |δ(q0 , s)!}, and the marked behavior is Lm (G) := {s ∈
L(G)|δ(q0 , s) ∈ Qm }. Let K ⊆ Σ∗ be a language; its prefix
closure is K := {s ∈ Σ∗ |(∃t ∈ Σ∗ )st ∈ K}. We say K is
Lm (G)-closed if
K ∩ Lm (G) = K. (5)
TDES G is nonblocking if Lm (G) = L(G).
To use TDES G in (4) for supervisory control, it is neces-
sary to specify certain transitions that can be controlled
by an external supervisor. First, as in the untimed theory
Wonham (2013), we need a subset of events that may be
disabled. Since disabling an event usually requires prevent-
ing that event indefinitely from occurring, only remote
events belong to this category. Thus let a new subset
Σhib ⊆ Σrem denote the prohibitible events; the supervisor
is allowed to disable any prohibitible event. Next, and
specific to TDES, we bring in another category of events
which can preempt event tick. Note that tick may not be
disabled, inasmuch as no control technology can stop the
global clock indefinitely. On this basis let a new subset
Σf or ⊆ Σact denote the forcible events; a forcible event
tick is defined and so are one or more forcible events,
then tick can be effectively erased from the current list
of defined events (contrast with indefinite erasure). There
is no particular relation postulated a priori between Σf or
and any of Σhib , Σrem or Σspe ; in particular, a remote event
may be both forcible and prohibitible. It is now convenient
to define the controllable event set Σc := Σhib ∪˙ {tick}.
simplify terminology. We emphasize that events in Σhib
can be disabled indefinitely, while tick may be preempted
only by events in Σf or . The uncontrollable event set Σu is
Σu := Σ − Σc = Σspe ∪(Σ˙ rem − Σhib ).
We introduce the notion of controllability as follows. Let
K ⊆ L(G) and s ∈ K; define the eligible event subset
EK (s) := {σ ∈ Σ|sσ ∈ K}. (6)
We say K is controllable wrt. G in (4) if, for all s ∈ K,
WODES 2014
Cachan, France. May 14-16, 2014

 E ˙
(s) ∩ (Σu ∪{tick})
 L(G)
if EK (s) ∩ Σf or = ∅,
EK (s) ⊇ (7)
 EL(G) (s) ∩ Σu
 (i) K is observable (in (9));
if EK (s) ∩ Σf or 6= ∅.
Thus K controllable means that an event σ is eligible
to occur in K if (i) σ is currently eligible in L(G), and
(ii) either σ is uncontrollable or σ = tick when there is
no forcible event currently eligible in K. Controllability
plays the central role in the TDES supervisory control
framework developed in Brandin and Wonham (1994) for
the case of full-event observation.
3. SUPERVISORY CONTROL OF TDES WITH
PARTIAL OBSERVATION
The Brandin-Wonham TDES framework has been ex-
tended to the partial-event observation case by Lin and
Wonham (1995), where the concepts of observability and
normality of TDES were introduced. This work is first
reviewed. Then we introduce relative observability, which
is weaker than normality, stronger than observability, and
closed under set union. Computation of the supremal rel-
atively observable (and controllable, Lm (G)-closed) sub-
language of a given language will be discussed.
3.1 Lin and Wonham’s Result
Let Σo ⊆ Σ be a subset of observable events. Define the
natural projection P : Σ∗ → Σ∗o according to
P (²) = ², ² is the empty string;
½
², if σ ∈
/ Σo ,
P (σ) = (8)
σ, if σ ∈ Σo ;
P (sσ) = P (s)P (σ), s ∈ Σ∗ , σ ∈ Σ.
As usual, P is extended to P : P wr(Σ∗ ) → P wr(Σ∗o ),
where P wr(·) denotes powerset. Write P −1 : P wr(Σ∗o ) →
P wr(Σ∗ ) for the inverse-image function of P .
A supervisor V under partial observation is any map
V : P L(G) → P wr(Σ). Denote by V /G the closed-loop
system where G is under the supervision of V ; then the
closed language L(V /G) ⊆ L(G) is defined inductively
according to
(i) ² ∈ L(V /G);
(ii) s ∈ L(V /G), σ ∈ V (P s), sσ ∈ L(G) ⇒ sσ ∈ L(V /G).
The marked language Lm (V /G) of V /G is defined by
Lm (V /G) := L(V /G) ∩ Lm (G).
A supervisor V is nonblocking if Lm (V /G) = L(V /G),
and admissible if for each s ∈ L(V /G),
(i) Σu ⊆ V (P s);
(ii) EL(G) (s) ∩ V (P s) ∩ Σf or = ∅ & tick ∈ EL(G) (s) ⇒
tick ∈ V (P s).
˙
Let K ⊆ Lm (G), and recall Σc = Σhib ∪{tick}. We say K
is observable (wrt. G and P ) if for every pair of strings
s, s0 ∈ Σ∗ with P s = P s0 there holds
(∀σ ∈ Σc )sσ ∈ K, s0 ∈ K, s0 σ ∈ L(G) ⇒ s0 σ ∈ K. (9)
Theorem 1. (Lin and Wonham (1995)) Let K ⊆ Lm (G)
be a nonempty language. There exists a nonblocking,
210
admissible supervisor V such that Lm (V /G) = K if and
only if
(ii) K is controllable (in (7));
(iii) K is Lm (G)-closed (in (5)).
While controllability and Lm (G)-closedness are properties
closed under set union, observability is not; consequently
the supremal sublanguage that satisfies the above three
conditions (or the optimal supervisor) need not exist in
general. This problem motivates us to propose the concept
of relative observability, which is closed under set union.
3.2 Relative Observability of TDES
Fixing a sublanguage C ⊆ Lm (G), we introduce relative
observability which sets C ⊆ L(G) to be the ambient
language in which observability is tested.
Definition 2. Let K ⊆ C ⊆ Lm (G). We say K is relatively
observable with respect to C, G, and P , or simply C-
observable, if for every pair of strings s, s0 ∈ Σ∗ with
P s = P s0 there holds
(∀σ ∈ Σc ) sσ ∈ K, s0 ∈ C, s0 σ ∈ L(G) ⇒ s0 σ ∈ K (10)
˙
where Σc = Σhib ∪{tick}.
Relative observability was first proposed in Cai et al.
(2013) for untimed DES. For TDES here, we extend the
concept by accounting for the event tick which may be
preempted only by a forcible event, in contrast with direct
disablement of prohibitible events.
If C1 ⊆ C2 ⊆ L(G) are two ambient languages, it follows
easily from Definition 2 that C2 -observability implies C1 -
observability. Namely, the smaller the ambient language,
the weaker the relative observability. In the special case
where the ambient C = K, Definition 2 becomes standard
observability (Lin and Wonham (1995)) for the given K.
This immediately implies the following.
Proposition 3. If K ⊆ C is C-observable, then K is also
observable.
The reverse statement need not be true (Cai et al. (2013)).
An important way in which relative observability differs
from observability is the exploitation of a fixed ambient
C ⊆ L(G). Let Ki ⊆ C, i = 1, 2. For (standard)
observability of each Ki , one checks lookalike string pairs
only in Ki , ignoring all candidates permitted by the other
language. Observability of Ki is in this sense ‘myopic’, and
consequently, both Ki being observable need not imply
that their union K1 ∪ K2 is observable. The fixed ambient
language C, by contrast, provides a ‘global reference’:
no matter which Ki one checks for relative observability,
all lookalike string pairs in C must be considered. This
more stringent requirement renders relative observability
algebraically well-behaved: an arbitrary union of relatively
observable languages is again relatively observable.
Proposition 4. Let Ki ⊆ S C, i ∈ I (some index set), be C-
observable. Then K = {Ki | i ∈ I} is also C-observable.
A proof is in Cai et al. (2013) (identical to the untimed
case). Whether or not K ⊆ C is C-observable, write
WODES 2014
Cachan, France. May 14-16, 2014
O(K, C) := {K 0 ⊆ K | K 0 is C-observable} (11)
for the family of C-observable sublanguages of K. Note
that the empty language ∅ is trivially C-observable, thus a
member of O(K, C). By Proposition 4, moreover, O(K, C)
has a unique supremal element supO(K, C) given by
[
supO(K, C) := {K 0 | K 0 ∈ O(K, C)}. (12)
This is the supremal C-observable sublanguage of K. An
algorithm that computes supO(K, C) was presented in Cai
et al. (2013).
Finally, we show that relative observability is weaker than
normality of TDES (Lin and Wonham (1995)), a property
that is also preserved by set unions. A sublanguage K ⊆ C
is (L(G), P )-normal if
K = P −1 P K ∩ L(G). (13)
This implies that no string in K may exit K via an
unobservable transition. Thus normality excludes, when
control is present, the disablement of unobservable, pro-
hibitible events, or the preemption of tick in case tick
is unobservable. Relative observability, by contrast, does
not impose this restriction, i.e. one may exercise disable-
ment/preemption over unobservable events.
Proposition 5. If K ⊆ C is (L(G), P )-normal, then K is
C-observable.
3.3 Algorithm
Let K ⊆ Lm (G) be a nonempty specification lan-
guage. Since K-observability, controllability, and Lm (G)-
closedness are all closed under set union, there exists a
unique supremal sublanguage of K that satisfies these
three properties. Denote this supremal sublanguage by
K ↑ ; by Proposition 3, K ↑ is observable, controllable,
and Lm (G)-closed. Therefore, by Theorem 1, there ex-
ists a nonblocking, admissible supervisor V such that
Lm (V /G) = K ↑ .
Now we present an algorithm to compute K ↑ . Let G and
K be finite-state (trim) TDES (as in (4)) with marked
languages Lm (G) and K, respectively.
Algorithm 1: Input G, K, and P : Σ∗ → Σ∗o .
1. Set K0 = K.
2. For i ≥ 0, apply the algorithm in Wonham and Ramadge
(1987) with inputs G and Ki to obtain Hi such that
Lm (Hi ) is the supremal controllable and Lm (G)-closed
sublanguage of Lm (Ki ).
3. Apply the algorithm in Cai et al. (2013) with inputs G,
Hi , and P : Σ∗ → Σ∗o to obtain Ki+1 such that Lm (Ki+1 )
is the supremal K-observable sublanguage of Lm (Hi ). If
Ki+1 = Ki , then output K↑ = Ki+1 . Otherwise, advance
i to i + 1 and go to Step 2.
Algorithm 1 terminates in finite steps, inasmuch as both
algorithms used in Steps 2 and 3 remove states and/or
transitions from K. The complexity of Algorithm 1 is
exponential in the state size of K, because the algorithm
in Cai et al. (2013) is of this complexity.
For an illustrative example of Algorithm 1, the reader is
referred to Cai et al. (2014), where Algorithm 1 is applied
211
to solve a partial-observation supervisory control problem
of a Guideway example.
4. NEW SUPERVISORY CONTROL OF TDES WITH
PARTIAL OBSERVATION
A new type of supervisory control for TDES was proposed
by Takai and Ushio (2006), and a weak observability
condition derived for the case of partial observation. This
work is first reviewed. Then we introduce relative weak
observability, which is stronger than weak observability
but closed under set union. Computation of the supremal
relatively weakly observable (and controllable, Lm (G)-
closed) sublanguage of a given language will be discussed.
4.1 Takai and Ushio’s Result
Again let Σo ⊆ Σ be a subset of observable events, and
P : Σ∗ → Σ∗o be the natural projection. A supervisor
V under partial observation is any map V : P L(G) →
P wr(Σact ) × P wr(Σf or ) such that for each t ∈ P L(G),
V (t) = (V1 (t), V2 (t)) satisfies the following two conditions:
(i) Σu ⊆ V1 (t);
(ii) V2 (t) ⊆ V1 (t) ∩ Σf or .
Here V1 (t) is the set of events in Σact to be enabled, which
must always include the uncontrollable subset Σu ; V2 (t)
is the set of events in Σf or to be forced to occur, which
must be enabled by V1 . The closed language L(V /G) of the
closed-loop system V /G is defined inductively according to
(i) ² ∈ L(V /G);
(ii) s ∈ L(V /G), σ ∈ Σact ∩ V1 (P s), sσ ∈ L(G) ⇒ sσ ∈
L(V /G);
(iii) s ∈ L(V /G), EL(G) (s) ∩ V2 (P s) = ∅, s.tick ∈ L(G) ⇒
s.tick ∈ L(V /G).
The marked language Lm (V /G) of V /G is defined by
Lm (V /G) := L(V /G) ∩ Lm (G).
A supervisor V is nonblocking if Lm (V /G) = L(V /G).
Let K ⊆ Lm (G). We say K is weakly observable (wrt. G
and P ) if the following two conditions hold:
(1) K is observable with respect to Σhib , i.e. for every pair
of strings s, s0 ∈ Σ∗ with P s = P s0 there holds
(∀σ ∈ Σhib ) sσ ∈ K, s0 ∈ K, s0 σ ∈ L(G) ⇒ s0 σ ∈ K.
(2) For each t ∈ P (K), there exists a subset
 
[
F (t) ⊆  EK (s) ∩ Σf or
s∈K∩P −1 (t)
such that for each s ∈ K ∩ P −1 (t) with tick ∈ EL(G) (s)
there holds
tick ∈ EK (s) ⇔ EL(G) (s) ∩ F (t) = ∅. (14)
Weak observability is identical to observability with re-
spect to Σhib , but exploits choices of forcible events to
address preemption of the event tick. It was shown (Takai
and Ushio (2006)) that if K is observable and controllable,
then it is weakly observable.
WODES 2014
Cachan, France. May 14-16, 2014
Theorem 6. (Takai and Ushio (2006)) Let K ⊆ Lm (G)
be a nonempty language. There exists a nonblocking
supervisor V such that Lm (V /G) = K if and only if
(i) K is weakly observable (in (14));
(ii) K is controllable (in (7));
(iii) K is Lm (G)-closed (in (5)).
Like observability, weak observability is not closed under
set union; consequently the supremal sublanguage that
satisfies the above three conditions (or the optimal su-
pervisor) need not exist in general. This motivates us to
propose relative weak observability below, which is closed
under set union.
Remark 1. The implementation of the supervisor V =
(V1 , V2 ) in Theorem 6 is as follows. After a string s ∈ L(G)
such that s.tick ∈ L(G), V observes the string t = P s ∈
P L(G). Then V enables all events in V1 (t), and forces
all events in V2 (t) = F (t). If one or more events in F (t)
is eligible after s, then tick is preempted; if no event in
F (t) is eligible after s, then tick is enabled. In comparison,
the implementation of the supervisor V in Theorem 1 is
simpler inasmuch as no explicit F (t) is needed for tick
preemption; indeed, V directly decides to enable or disable
tick, and controllability ensures the availability of forcible
events for the disabling/preempting action.
4.2 Relative Weak Observability
Fixing a sublanguage C ⊆ Lm (G), we introduce rela-
tive weak observability which sets C ⊆ L(G) to be the
ambient language (as is done in Definition 2 for relative
observability). The key idea here is to distinguish differ-
ent ‘control patterns’ for tick preemption in each set of
lookalike strings; we do so by imposing on each such set
a special equivalence relation. The equivalence classes of
this equivalence relation have mutually disjoint subsets
of forcible events, so that in each equivalence class tick
preemption may be carried out independently.
Let P : Σ∗ → Σ∗o and s ∈ L(G). Write [s] := {s0 ∈
L(G)|P s0 = P s} for the set of lookalike strings to s in
L(G). Define a binary relation ≡ on [s] as follows: for
all s, s0 ∈ [s], s ≡ s0 if either (i) EL(G) (s) ∩ EL(G) (s0 ) ∩
Σf or 6= ∅ or (ii) there exist s1 , ..., sk ∈ [s], k ≥ 1, such that
EL(G) (s) ∩ EL(G) (s1 ) ∩ Σf or 6= ∅
..
. supWO(K, C) := {K 0 | K 0 ∈ WO(K, C)}.
EL(G) (sk ) ∩ EL(G) (s0 ) ∩ Σf or 6= ∅.
In words, two strings s, s0 ∈ [s] satisfy s ≡ s0 if either (i)
they are followed by some common forcible events that are
eligible in L(G), or (ii) there is a finite chain of strings in
[s] that ‘connects’ s to s0 through some common forcible
events that are eligible in L(G). This implies that s ≡ s0
is false (s, s0 ∈ [s]) if and only if for every s00 ∈ [s] with
s00 ≡ s0 there holds EL(G) (s) ∩ EL(G) (s00 ) ∩ Σf or = ∅.
It is easily verified that ≡ is reflexive, symmetric, and
transitive, and thus an equivalence relation on [s].
Definition 7. Let K ⊆ C ⊆ Lm (G). We say K is relatively
weakly observable with respect to C, G, and P , or simply
weakly C-observable, if the following two conditions hold:
212
(1) K is C-observable with respect to Σhib , i.e. for every
pair of strings s, s0 ∈ Σ∗ with P s = P s0 there holds
(∀σ ∈ Σhib ) sσ ∈ K, s0 ∈ C, s0 σ ∈ L(G) ⇒ s0 σ ∈ K.
(2) For every pair of strings s, s0 ∈ Σ∗ with P s = P s0 there
holds
s.tick ∈ K, s0 ∈ C, s0 .tick ∈ L(G), s ≡ s0 ⇒ s0 .tick ∈ K.
The first condition above is the relative observability of
K with respect to Σhib . The second condition deals with
the event tick: two lookalike strings s, s0 ∈ C which satisfy
s ≡ s0 are required to have identical one-step continuations
of tick, if allowed in L(G), with respect to membership in
K. This is weaker than relative observability with respect
to tick, inasmuch as the requirement is imposed only on
lookalike strings satisfying s ≡ s0 . Therefore the following
result is immediate.
Proposition 8. If K ⊆ C is C-observable, then K is also
weakly C-observable.
As a corollary of Propositions 8 and 5, relative weak
observability is weaker than normality. Next, we show
that relative weak observability is stronger than weak
observability.
Proposition 9. If K ⊆ C is weakly C-observable and
controllable, then K is also weakly observable.
For the proof refer to Cai et al. (2014). As with relative
observability, the fixed ambient language C, as well as the
equivalence relation ≡, renders relative weak observability
algebraically well-behaved: an arbitrary union of relatively
weakly observable languages is again relatively weakly
observable.
Proposition 10. Let Ki ⊆ C, i ∈ IS (some index set), be
weakly C-observable. Then K = {Ki | i ∈ I} is also
weakly C-observable.
For the proof refer to Cai et al. (2014). Whether or not
K ⊆ C is weakly C-observable, write
WO(K, C) := {K 0 ⊆ K | K 0 is weakly C-observable}
(15)
for the family of weakly C-observable sublanguages of
K. Note that the empty language ∅ is trivially weakly
C-observable, thus a member of WO(K, C). By Propo-
sition 10, moreover, WO(K, C) has a unique supremal
element supWO(K, C) given by
[
(16)
This is the supremal weakly C-observable sublanguage of
K. The next subsection presents an algorithm to compute
supWO(K, C).
4.3 Algorithm
As noted immediately above Proposition 8, the only dif-
ference between relative weak observability and relative
observability is the treatment of the event tick: in the
former, essentially, tick may be treated independently for
lookalike strings that do not belong to the same equiva-
lence class of ≡. Thus our idea of computing the supremal
relatively weakly observable sublanguage of a language K
WODES 2014
Cachan, France. May 14-16, 2014
is as follows: (1) identify equivalence classes of ≡, and
relabel tick using distinct event labels tick1 , tick2 , ... for
distinct equivalent classes; (2) apply the algorithm in Cai
et al. (2013) to compute the supremal relatively observable
sublanguage of K; and finally (3) relabel tick1 , tick2 , ...
back to tick.
Let G = (Q, Σ, δ, q0 , Qm ), C, and K be finite-state (trim)
TDES (as in (4)) with marked languages Lm (G), C, and
K, respectively.
Algorithm 2: Input G, C, K, and P : Σ∗ → Σ∗o .
1. For each t ∈ P (L(G)), use the subset construction
technique (Wonham (2013), Section 2.5) to find the subset
Q(t) := {q ∈ Q|(∃s ∈ P −1 (t))δ(q0 , s) = q}.
For each q ∈ Q(t), write EL(G) (q) := {σ ∈ Σ|δ(q, σ)!}.
Then for each pair (q, q 0 ) ∈ Q(t) × Q(t), q ≡ q 0 if either
(i) EL(G) (q) ∩ EL(G) (q 0 ) ∩ Σf or 6= ∅ or (ii) there exist
q1 , ..., qk ∈ Q(t), k ≥ 1, such that
EL(G) (q) ∩ EL(G) (q1 ) ∩ Σf or 6= ∅
..
. observability, weaker than normality, and closed under set
0
EL(G) (qk ) ∩ EL(G) (q ) ∩ Σf or 6= ∅.
Thus for each Q(t) we identify the equivalence classes of
≡, say Q1 (t), Q2 (t), .... For tick defined at some state in
Qi (t), i = 1, 2, ..., relabel it by ticki . Do the corresponding
relabeling in C and K, and denote the relabeled generators
by G0 , C0 , and K0 .
2. Apply the algorithm in Cai et al. (2013) with inputs
G0 , C0 , and K0 , to compute K0↑ , where Lm (K0↑ ) is the
supremal C-observable sublanguage of K.
3. Relabel the events ticki in K0↑ by tick, and denote the
result by K↑ . Lm (K↑ ) is the supremal weakly C-observable
sublanguage of K.
Step 1 of Algorithm 2 above is in the worst case exponen-
tial in the state size of G; this is the overhead complexity
as compared to Algorithm 1. For an illustrative example
of Algorithm 2, the reader is referred to Cai et al. (2014).
Finally, for a nonempty specification language K ⊆
Lm (G), we modify Algorithm 1 by using Algorithm 2 in
Step 3 to compute the supremal weakly K-observability,
controllability, and Lm (G)-closed sublanguage of K. De-
note this supremal sublanguage by K ↑ ; by Proposition 9,
K ↑ is weakly observable, controllable, and Lm (G)-closed.
Therefore, by Theorem 6, there exists a nonblocking su-
pervisor V such that Lm (V /G) = K ↑ .
Remark 2 (tradeoff between timed relative observability
and relative weak observability). We have derived two ob-
servability concepts for timed supervisory control. Timed
relative observability is conceptually simpler (since its
requirement is imposed only on lookalike strings), allows
easier implementation (see Remark 1), but the resulting
tick preemption behavior is generally more restrictive. On
the other hand, timed relative weak observability requires
extra information about the equivalence relation ≡ on
lookalike strings, to trade for generally more permissive
tick preemption behavior. The decision as to which ob-
servability concept to use therefore depends on how much
extra information is needed to achieve the correspond-
213
ing behavior improvement; in practice the latter will be
case-dependent. Nevertheless, since we have designed algo-
rithms for both observability concepts, our suggestion is as
follows. First compute the supremal relatively observable
sublanguage K1 (of a given specification language K);
if the tick preemption behavior of K1 is ‘satisfactory’,
then use K1 . Otherwise, compute the supremal relatively
weakly observable sublanguage K2 of K; comparing K2
with K1 , if the improvement of tick preemption behavior
is ‘significant’, then use K2 .
5. CONCLUSIONS
We have studied two new observability concepts in su-
pervisory control of timed DES under partial observa-
tion. First, we have introduced timed relative observabil-
ity, which is stronger than timed observability, weaker
than normality, and closed under set union. An algorithm
has been presented for computing the supremal relatively
observable, controllable, and Lm (G)-closed sublanguage
of a given language. Second, we have identified timed
relative weak observability, which is stronger than weak
union. An algorithm has been presented for computing
the supremal relatively weakly observable sublanguage of
a given language.
REFERENCES
Brandin, B. and Wonham, W. (1994). Supervisory control
of timed discrete-event systems. IEEE Trans. Autom.
Control, 39(2), 329–342.
Cai, K., Zhang, R., and Wonham, W.M. (2013). On
relative observability of discrete-event systems. In Proc.
52nd IEEE Conf. on Decision and Control, 7285–7290.
Florence, Italy.
Cai, K., Zhang, R., and Wonham, W.M. (2014).
On relative observability of timed discrete-event
systems. Technical report, Systems Control
Group, ECE Dept, University of Toronto. Online
at https://sites.google.com/site/kaikai627/publication.
Cieslak, R., Desclaux, C., Fawaz, A.S., and Varaiya, P.
(1988). Supervisory control of discrete-event processes
with partial observations. IEEE Trans. Autom. Control,
33(3), 249–260.
Lin, F. and Wonham, W.M. (1988). On observability of
discrete-event systems. Inform. Sci., 44, 173–198.
Lin, F. and Wonham, W.M. (1995). Supervisory control of
timed discrete-event systems under partial observation.
IEEE Trans. Autom. Control, 40(3), 558–562.
Takai, S. and Ushio, T. (2006). A new class of supervisors
for timed discrete event systems under partial observa-
tion. Discrete Event Dynamic Systems, 16(2), 257–278.
Wonham, W.M. (2013). Supervisory Control of Discrete-
Event Systems. Systems Control Group, ECE Dept,
University of Toronto, updated July 1, 2013. Available
online at http://www.control.toronto.edu/DES.
Wonham, W.M. and Ramadge, P.J. (1987). On the
supremal controllable sublanguage of a given language.
SIAM J. of Control and Optimization, 25(3), 637–659.