Modelling and Hierarchical Control of CBTC
Modelling and Hierarchical Control of CBTC
Abstract—Over the last decade, the railway industry has a great on two results from this category, one of them structural
evolution about signaling system and there is more orientation decentralized control and the other is hierarchical control.
from the standard railway signaling system to the communication- The decentralized control architecture suggested by Lee and
based signaling system day to day. Communications-based train Wong (2002). They handled plant models that are composed of
control (CBTC) is a very flexible and useful approach to check
a number of subsystems that are coupled with shared events.
train activity and track operation. This system basically build
upon radio communication to transfer in time and correct train Specifications are given for each subsystem individually, and
control information. the task is to synthesize individual supervisors for these
In this paper, we focus on model the all necessary CBTC elements subsystems. As the all subsystems are coupled each other,
with finite state automata and build CBTC control architecture synthesis will in general need to refer to the synchronous
with decentralized DES and support the existing control product of all subsystems. Conditions under which such product
architecture with a three-level hierarchy. For the overall system, can be avoided are given in Lee and Wong (2002).
we show hierarchical consistency and that the closed-loop In hierarchical architectures (Zhong and Wonham, 1990; da
behavior is non-blocking. This paper gives an overview of the Cunha et al., 2002; Hubbard and Caines, 2002), controller
modelling a discrete event system about CBTC and gives control
synthesis is based on a plant abstraction which called high-level
of CBTC.
Keywords—Discrete Event Systems, Hierarchical Control, Finite model. This is supposed to be less complex than the original
State Automata, CBTC, Supervisory Control Theory plant model which called low-level model. Technically,
abstractions can be defined as language projections. While from
I. INTRODUCTION worst situation scenarios projections are known to be of
In today's world, the usage area of intelligent systems is exponential computational complexity, application relevant
increasing day by day. With the development of technology, situations with polynomial complexity are defined by Wong
applications on intelligent systems that will facilitate human life (1997). At this point very significant question arises how to
have also gained momentum. Intelligent systems spread over a derive the plant abstraction, such that a high-level controller can
lot of field. One of them is definitely transportation. Especially be implemented by available low-level control activity which
at big cities, the number of people is increasing day by day. This means hierarchical consistency. A characterization of this
cause increase urbanization, population growth and the feature is given by Zhong and Wonham (1990) [10].
overcrowding. This is why time is very important for people. In this paper we show in Section 1 that basic notations and
The most important thing in the field of transportation is to definitions of supervisory control theory. Section 2 we
transport people in an accurate and timely manner. For demonstrate and prove hierarchical control system architecture
example, for a subway, the maximum efficiency of trains and non-blocking control for the architecture. In Section 3 we
depends on the condition of the railway, the condition and the demonstrate decentralized control and then combined with
timing of the trains. Various technologies have been used to hierarchical control architecture. Section 4 we explain some
achieve this, and we can rank them in fixed block and moving basis information about CBTC then apply decentralized
block. In today's world, there is an intensive study about hierarchical control architecture to the CBTC systems and
moving block systems. One of them is Communications-based synthesize control for CBTC.
train control shortly CBTC. On the other hand, there are various
II. PRELIMINLARIES
approaches for controlling such discrete event systems like
CBTC. One of them is the supervisory control. The supervisory Now it will be demonstrated basic term about supervisory
control theory for discrete event systems (DES) was first put control theory.
forward by Ramadge and Wonham [1]. There are several Σ represent finite alphabet and Σ ∗ represent the set of all finite
different studies to reduce the complexity of synthesis strings which called Kleene Closure. , Σ ∗ are two strings
∗
algorithms for the supervisory control of discrete event systems and Σ concatenation of these two string. It is written as
Rudie and Wonham, 1992[2], Jiang et al., 2001[3] Yoo and = and is prefix of when ≤ and , , Σ ∗ with
Lafortune, 2000[4], Hubbard and Caines 2002[5], da Cunha et = . The empty string ε is the identified element of
al., 2002[6], Zhong and Wonham, 1990[7]; Lee and Wong, concatenation: = = for any string . Let is a
2002[8]; Wong and Wonham, 1996[9]. Our framework builds language over Σ. is a subset of Σ∗ ( L ⊆ Σ ∗ ). language is the
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.
2018 6th International Conference on Control Engineering & Information Technology (CEIT), 25-27 October 2018, Istanbul, Turkey
prefix closure of L and consists of all the prefixes of all the / represent controlled system. The language generated by
strings in L. ≔{ Σ∗ | ∃ . . ( )}. is prefix / is ( / ). Defined by (i) ( / ) and (ii) ( / )
closed if = . The other important operation is natural if and only if ( / ), ( ) and ( ). The
projection, or simply projection. Projection means takes a string language marked by S/G is ( / )≔ ( / )∩ ( ).A
formed from the larger event set (Σ l=large) and erases events closed-loop feedback system is non-blocking if ( / )=
in it that do not belong to the smaller event set (Σ s=small). ( / ) and blocking if ( / ) ≠ ( / ).
Projection defined as : Σ ∗ → Σ ∗ . From this point (i) ( ) ≔ is said to be a controllable language according ( ) provided
(ii) ( ) ≔ Σ otherwise that there is a supervisor so that = ( / ). The set of all
( )≔ Σ \ Σ (iii) ( ) ≔ ( ) ( ) for Σ ∗ , Σ ∗ . controllable language is represented by ( ( )). From here
The other operation is inverse projection. Given a string of ( ) = { ⊆ ( )| ∃ ℎ ℎ = ( / )}.
events in the smaller event set (Σ ), the inverse projection P
Therefore, for all specification language there exists a
returns the set of all strings from the larger event set (Σ ) that
supremal controllable sublanguage of D with respect to ( ).
project, with P, to the given string. Inverse projection of t
returns the set of strings that are projected on t. This is defined as ( )( ) ≔ ∪ ( ) ⊆ }.
P ( ) ≔ { Σ ∗ : ( ) = }. The other operation parallel Supervisor S cause close-loop behavior ( ) ( ) is said to be
composition or sometimes called synchronous composition maximal permissive. This is realized a maximal permissive
|| ⊆ Σ∗ . and are two languages. supervisor using with ( ) ( ). [11],[12]
|| = P ( ) ∩ P ( ) ⊆ Σ ∗ . A specification language D is -closed if ∩ = .
On the other hand an automaton is defined over its formal ( )-closed languages set’s is shown with Ӻ ( ) .
description. An automaton is a construct made of states If a specification D element of Ӻ ( ) , then the plant generated
designed to determine if the input should be accepted or language ( ) is non-blocking under maximal permissive
rejected. An finite automaton is a six tuple supervision.[11]
= (X, Σ, δ, , , ) where X is the finite states, Σ is the
finite alphabet of events, δ is transition function, is control
pattern is initial state , is the set of marked states III. HIERARCHICAL CONTROL SYSTEM
. Δ: × → is the transition function δ( , ) =
means that there is a transition with event from state to
state ; generally , δ is a partial function on its domain. The
generated language ( ) and the marked language ( ) are
defined as ( ) ≔ { Σ ∗ | δ( , ) }
( ) ≔ { ( ) | δ( , ) } such that ( ) ⊆ ( ).
For definition parallel composition or synchronous composition
of two automata and ,
|| ∶= × ,Σ ⋃ Σ , , || ,( , ), ×
then ( | ) = ( || ( ).
In a supervisory control theory, we can write Σ = Σ ∪ Σ ,
Σ ∩ Σ = ∅ , Σ is controllable evetns and Σ is
uncontrollable events. A control pattern represent with γ
Fig. 2. Hierarchical control architecture
symbol and Σ ⊆ γ ⊆ Σ and is the set of all control pattern
⊆2 . We take a close look at hierarchical control. At above figure
is low-level plant model and is the low-level supervisor.
and is a closed-loop system which interacts each other with
and . means control action at low-level and
means feedback information at low level. On the other
hand we can write is the high-level plant model which
abstracted and is the high-level supervisor. and also
closed-loop system. Low-level and high-level are connected
each other with and . brings to
Fig. 1. The feedback loop of supervisory control, is the high-level control action on and operates the
uncontrolled system and is the supervisor abstract plant according to low-level model which is
detailed model. and express command and
Supervisors is a function from the language created by to the control respectively and and express report and
power set of Σ: : ( ) → 2 or : ( ) → advice. [12]
( ) is the set of enabled events so this means can only Abstraction is keyword for hierarchical control structure. We
execute event if ( ) include this event. Supervisor cannot can obtain high level plant and high level language with
disable uncontrollable event. abstraction.
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.
2018 6th International Conference on Control Engineering & Information Technology (CEIT), 25-27 October 2018, Istanbul, Turkey
As we mentioned before G is finite automaton and we can write over the alphabet ≔ Σ ∪ Σ ∪ … … ∪ Σ or ≔∪ Σ .
= (X, Σ, δ, , , ). is set of high level events and Also we can write the controllable and uncontrollable events
⊆ . We can use projection which called reporter map is a , ≔ ∩ and , ≔ ∩ where = ∪
( = ) ∶ ∗ → ( ∗) such that (i) ( )= and and ∩ = ∅. The subsystems like , are synchronized
(ii) ( ) = ( ) or ( ) = ( ) , Σ and . by shared events if ∩ ≠ ∅.
is the high-level language which is characterized with ≔ Local supervisor is also defined as mentioned before. Local
( ( )). is high-level marked language and ⊆ . low-level supervisor : → ( control pattern). Low-level
must be regular. closed-loop languages ≔ ( / ) , ≔ ∩ , , ∶=
Therefore we can write ( ) = and ( )= . || , ≔ || , = ∩ .
From this point we can write also high-level uncontrollable Σ = ( ), = ( ) where is generator. The high-
and high-level controllable Σ events.Hence =Σ ∪Σ , level supervisor : → and high-level closed-loop
and Σ ∩ Σ = ∅. At above it is defined hierarchical control language ( / ). A valid low-level supervisor : : →
system architecture. A hierarchical control system include required to fulfill ( ( / )) ⊆ ( / ).
, , , . Supervisor and provides below status: Finally for the hierarchical abstraction the reporter map or
: → and is high-level control patterns we can natural projection : ∗ → ( )∗ and ≔∪ , , ( ∩ )
define as ≔ {γ | Σ ⊆ γ ⊆ }. and the high level language ≔ ( ) and ≔
: ( ) → with ( ( / )) ⊆ ( / ). { . . ( )∩ ≠ ∅ and such that =
( )
We can obtain such that ( / ) = ( ) with a non- At this point we can combine decentralized control system and
blocking and high-level closed-loop. is high-level hierarchical control architecture. This is hierarchical and
specification. decentralized control system. [8]
Given ( , , ) we find supervisor which low-level
supervisor such that ( /G) is non-blocking. On the other
hand, we cannot obtain always low-level supervisor which
provides this equation ( ( / )) ⊆ ( / ). If this
equations is satisfied then we have hierarchical consistency
The other important features are local non-blocking and marked
state consistency.
( , , ) is a hierarchical abstraction and is locally
non-blocking if for all ( ) with ( ) = and
∀ ( ), ∃ ( − )∗ such that ( ).
Hierarchical abstraction ( , , ) is locally non-blocking if
is locally non-blocking.
Let explain marked state acceptance, let ( , , ) be a Fig. 3. Hierarchical and decentralized control architecture
hierarchical abstraction and identify
V. EXAMPLE SYSTEM
, ≔{ ( )| ( )
= ∧ ∃ . . ( )} ⊆ ∗ Communications-based train control (CBTC) uses radio
As the set of exit strings of . communication to transfer timely and correct train control
The string is marked state accepting if for all information. CBTC is the choice of mass-transit railway
operators today, with over a hundred systems currently installed
, ∃ ≤ ℎ ( )= and means
worldwide. The safety related, time-critical applications such as
exit string. train control impose stringent reliability and availability
( , , ) is marked string accepting providing that is requirements on the radio communication technology used.
marked string accepting for all . The aim of train control system is to prevent trains from crash
At above definition means that every string match with a and avoid derailing [14], [15], [16].
marked high-level string has a marked predecessor string in the The fixed block include color light signals track circuits which
low-level.[10] give information presence of train and axle counters. However
this technology is getting old. Such systems responsible for
most of the delays. This is why the fixed block systems are
IV. HIERARCHICAL CONTROL FOR DECENTRALIZED
rapidly being altered by modern signaling
SYSTEMS
systems,[17],[18],[19].
In decentralized control system the systems are separated to the In modern, communication-based railway signaling or moving
subsystem. So decentralized control system include more block, different tools of telecommunication systems are used to
subsystems. These subsystems have own supervisor. We can transmit train control information between the train and the
be modeled finite state automata , = 1,2 … . . , and all wayside. CBTC is a modern, radio-communication-based
subsystems have own alphabet Σ , = 1,2 … . . , . The overall signaling system. CBTC using radio communication, it enables
systems is defined as ≔ || || … . . || or ≔ || high resolution and real-time train control information. Hence
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.
2018 6th International Conference on Control Engineering & Information Technology (CEIT), 25-27 October 2018, Istanbul, Turkey
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.
2018 6th International Conference on Control Engineering & Information Technology (CEIT), 25-27 October 2018, Istanbul, Turkey
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.
2018 6th International Conference on Control Engineering & Information Technology (CEIT), 25-27 October 2018, Istanbul, Turkey
Fig. 12. Decentralized and hierarchical control architecture [7] H. Zhong and W.M. Wonham. On the consistency of hierarchical
supervision in discrete-event systems. IEEE TAC, 1990.
[8] S-H. Lee and K.C. Wong. Stuctural decentralised control of concurrent
For the rest of system same computations are made and find all discrete-event systems. EJC, 2002.
high-level projected subsystems then we combined subsystems [9] K.C. Wong and W.M. Wonham. Hierarchical control of discrete-event
which have interaction between each other. At the end in order systems. Discrete Event Dynamic Systems, 1996.
to find top of the hierarchy in other word all systems behavior, [10] K. Schmidt, J. Reger, T. Moor. Hierarchical control for structural
we combined all projected subsystems. Thus we found top of decentralized des .
the hierarchy. [11] C.G Cassandras and S. Lafortune. Introduction to discrete event systems.
As a result in this paper complexity of system is reduced using Second edition. 2008.
with decentralized and hierarchical control theory. Thanks to [12] W.M Wonham Supervisory Control of discrete event systems ECE
1636f/1637s 2014-15 pp.203-204.
this theory state space exploding is prevented.
[13] S. Morar, “Evolution of communication based train control worldwide,”
VI. CONCLUSIONS in Proc. IET Professional Develop. Course Railway Signalling Control
Syst. (RSCS), London, U.K., Jun. 2010, pp. 281–289.
In this paper basic knowledge is presented about
[14] “Sector overview and competitiveness survey of the railway supply
communication based control train (CBTC) systems and the industry,” European Comm., Rotterdam, The Netherlands, 2012.
supervisory control for discrete event systems. We made [Online].Available:http://ec.europa.eu/DocsRoom/documents/3950/
computations on states spaces which grow exponentially with attachments/1/translations/en/renditions/native.
the number of system component also decentralized structure [15] Frost & Sullivan. (Jul. 2013). Strategic Analysis ofCommunication Based
of composed systems in combination with a hierarchical Train Control Systems in the Western European Urban Rail Market.
[Online].Available:http://www.frost.com/sublib/displayreport.do?id=M9
abstraction is showed. The decentralized structure has been 2D-01-00-00-00.
used for reduce complexity of the supervisor implementation. [16] The Global Rail Market Now to 2016: UNIFE Study Key Findings
On the other hand non-blocking behavior and hierarchical &Future Outlook, UNIFE, Brussels, Belgium, 2008. [Online].
consistency of the controlled system is proved. To the end Available:http://www.rolandberger.at/media/pdf/rb_press/Roland_Bergr
theory is applied a real world example which name is CBTC. _Studie_UNIFE_20080924.pdf
[17] “Worldwide rail market study—Status quo and outlook 2016,” UNIFE,
REFERENCES Brussels, Belgium, Tech. Rep., 2008.
[18] I. Sener, O. T. Kaymakci, U. Ustoglu, G. Cansever. Specification and
formal verification of safety properties in a point automation system.
[1] Ramadge, Peter J.; Wonham, Walter M. (January 1987). "Supervisory Turkish Journal of Electrical Engineering & Computer Sciences
Control of a Class of Discrete Event Processes". SIAM Journal on Control doi:10.3906/elk-1311-27
and Optimization (SICON). 25 (1): 206–230. doi:10.1137/0325013
[19] M.A. Oz, O. T. Kaymakci. An Automatic Formal Model Generation and
[2] K. Rudie and W.M. Wonham. Think globally, act locally: decentralized Verification Method for Railway Interlocking Systems.Gazi University
supervisory control. IEEE TAC, 1992 Journal of Science. GU J Sci 30(2): 133-147 (2017)
[3] S. Jiang, V. Chandra, and R. Kumar. Decentralized control of discrete [20] Wikipedia. Communications-Based Train Control—Wikipedia, The Free
event systems with multiple localspecializations. Proc. ACC, 2001 Encyclopedia.Accessed on May 30, 2016 [Online].
[4] T. Yoo and S. Lafortune. A generalized framework for decentralized Available:https://en.wikipedia.org/w/index.php?title=Communications-
supervisory control of discrete event systems. WODES, 2000. based_train_control&oldid=740770099
[5] P. Hubbard and P.E Caines. Dynamical consistency in hierarchical [21] “Benefits and barriers to CBTC and ETCS convergence,” in Proc.
supervisory control. IEEE TAC,2002. MetroRail Conf., London, U.K., Mar. 2012.
[6] A.E.C. da Cunha, J.E.R. Cury, and B.H. Krogh. An assume guarantee J. Farooq, Radio Communication for communications-based train
reasoning for hierarchical coordination of discrete event systems. control(CBTC): A tutorial and survey. IEEE communıcatıons surveys &
WODES, 2002. tutorıals, vol. 19, no. 3, third quarter 20
Authorized licensed use limited to: ULAKBIM UASL ISTANBUL TEKNIK UNIV. Downloaded on December 31,2024 at 23:38:40 UTC from IEEE Xplore. Restrictions apply.