Embedded Systems As Discrete Event Systems

1
Embedded Systems As Discrete Event Systems

• M. Rokonuzzaman

2
Outline

• Background and Motivation
• Related Work
• Overview of Discrete Event Dynamic System and
  Petri Net
• Modeling of IAD as DES
• Implementation of IADs (Running Length Meter)
  DES Model
• Summary and Recommendations for Future Work

3
Keyword

• embedded computing,
• industrial automation devices,
• reactive and real-time system,
• discrete event system (DES),
• uni-capacity Petri net,
• finite state machine,
• software implementation as DES,
• and formal development methodology.

4
Industrial Automation and Embedded Systems

• Modern industrial automation devices are usually
  implemented as embedded computing applications.
• In these applications, software components
  running on microcontroller like limited computing
  capacity platform performs core functions such as
  perception, reasoning, and deciding about taking
  actions.

5
Challenges

• Event driven reactivity, real-time, limited
  resources, concurrence, and reliability are major
  development challenges .
• The nature of applications also limit the freedom
  of detecting and fixing bugs by following
  conventional software development approach of
  repeatedly running the application in actual
  environment.
• Moreover, the complexity of modern intelligent
  automation devices is continuously growing as
  well.

6
Limitation of Conventional Approach

• The conventional intuition based development
  approach of subsystem level functional
  decomposition, generating code for implementing
  each subsystem and integrating them as a whole
  is facing the limit of ensuring error free
  understanding of requirement and translating the
  requirement into software code .

7
Related Development

• It has been reported that discrete event system
  formalism is suitable to model reactive and
  real-time characteristics of embedded devices.
• Being graphical as well as mathematical discrete
  event system tool, Petri net has been found to be
  suitable in modeling such systems .

8
Limitation of Current Development

• It appears that the application of Petri net
  based DES formalism has not progressed beyond
  modeling, analysis, design, and verification .
• Still to date, the implementation of DES model
  in software code largely depends on traditional
  software development approach .
• Largely intuition and trial-error based
  traditional software development approach does
  not appear to be adequate to deal with the need
  of handling growing complexity, ensuring
  reliability, and meeting market pressure.

9
Our Approach

• Therefore, the work presented in this paper has
  made an attempt to make progress in applying the
  same state equation, which originates from
  analysis and design of the device as DES to
  govern the interactions between software
  components or subroutines implementing the
  device.

10
Concept of DES and Petri Net

The state equation of Petri net model of DES is
shown below
11

• The state equation for this transition is shown
  below

12
Modeling of IAD as DES
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Implementation
18
(No Transcript)
19
(No Transcript)
20

• It has been observed that the implementation of
  the DES model on microcontroller requires that
  the computational complexity of the state
  equation manipulation should be kept very low.
• One of the ways to address this requirement is to
  represent marking vector, incidence matrix, and
  transition control vector in certain way so that
  their bit wise manipulation can compute state
  transition.

21

• Based on this understanding, this work proposes
  that Petri net based DES model should have finite
  capacity 1.
• Such proposition also implies that all of its arc
  weights are 1s.
• Moreover, the Petri net should be pure that is
  for a pair of place p and transition t, p cannot
  be both input and output place for t.
• It is proposed that a token in a place only
  indicates conditions requiring only one bit for
  representation presence (1) or absence (0). As a
  result, both incidence matrix and marking vector
  can be manipulated using efficient bit wise
  operation.

22
Summary

• This methodology has shown that the same state
  equation, which describes behaviors of the
  modeled system, could be used to control the
  execution of transition associated code modules
  during operation.
• As a result, the gap between the modeled system
  and the implementation of the actual system has
  been reduced.