Outline
Introduction
Problem Definition
System Dynamics Basics
CDEEP : Current State of the art
CDEEP Model
Features of a System Dynamics Model
Conclusion & Future Work
References
Introduction
India produced 401,791 engineers in 2003-04 and in 2004-05, the number of engineering graduates increased to 464,743 [1].
But only 25 per cent of them are employable[1].
Main Reason : lack of well qualified teachers
Possible Solution :
Make IIT education accessible through Distance Education
Cost Effective
Global Reach
Introduction continued …
IITB has been running the distance education program since last 10 years.
Currently in the form of CDEEP to provide:
Good quality courses taught by IIT Bombay faculty
To everyone
At any place
Both synchronous and asynchronous modes
But number of students benefiting from CDEEP live courses has not increased as expected
Problem Definition
To model and analyse CDEEP system
To find out :
If there is any bottleneck resource
What–If analysis
If any policy changes needed
using System Dynamics
Why???
System Dynamics Basics
Computer
simulation
modeling for studying and managing complex feedback systems, such as business, engineering, and social systems Think in terms of cause-and-effect
Focus on Feedback Loops
situation when output from an event will influence the same event in the future
Study Grade
s
Parents’
Expectation s
More More
More
SD Modeling: Standard approach [2]
Identify the problem
Develop a dynamic hypothesis
Create a basic causal loop diagram
Convert the causal diagram to a Stock flow diagram
Write the equations
Estimate the parameters and initial conditions.
using statistical methods, expert opinion, market research data or other relevant sources.
Simulate the model and analyze results
Causal Loop Diagram
shows how one variable affects another.
nodes represent variables and arrows (called causal links) represent relationship
difficult to infer the behavior of a system only from its casual-loop representation
+ Feedbac
k Loop Node
Causal
Link time
Populatio n
Stock and Flow Diagram
Distinguishes between different types of variables
Consists of three different types of elements:
stocks, flows, and information
Stoc k
Flo Informatio w
n
Stock and Flow Diagram cntd….
SFD allows to represent relations among variables in terms of equations.
For Example
It becomes infeasible to solve as stocks and flows increase
Use computer simulators
Many simulators are available, (none is open source )
We used Vensim PLE by Ventana Systems, Inc. [4]
Simulation result is time-history of variables
in terms of Graph/Table
Population = Initial(Population)+ (birth- death)dt
CDEEP : Current State of the art
Distance Education through
Live Webcast and Satellite Transmission
Dynamic System with Feedback Loops
4 studios for live webcast (only 1 for satellite )
Live Webcast through Internet at 100 kbps for each connection
Not many students participating in this program
Our Work
Modeled Webcast and EDUSAT parts independently
Applied iterative approach to develop the model Variable Units Initial value/
assumption Total number of
Students
Students 20 Number of courses courses 20 Quality of video dimensionle
ss
between 0 and 1
Total available bandwidth
Kbps 8 Mbps
Bandwidth per connection
Kbps/
student
Ideally 100 kbps
Student satisfaction dimensionle ss
between 0 and 1
Server Performance dimensionle ss
between 0 and 1
Initial Webcast Model
Causal Loop Diagram
Initial Webcast Model
Stock and Flow Diagram
joining new students
Students Leaving
Simulation Results
Consistently 1
Equilibrium
Bottlenec k
Observations
Number of Students becomes constant (=200 student) after 24 months
Increasing number of courses doesn’t help
Bandwidth is the only bottleneck
Server is always underutilized
Limitations
MHRD grants can be used to bring in more resources, e.g. Bandwidth
Student feedback does matter
Marketing issues can not be ignored
Modified Webcast Model
Causal Loop Diagram
Modified Webcast Model
Stock and Flow Diagram
Becomes 1 Gbps after 24
months
From 4 to 6
Simulation Results
One more here
Server Overload Huge ed
increment due to increase in
BW
Observations
Grants can be spent for different resources
Bandwidth increase much needed
If bandwidth is increased, server will become overloaded after 3 semesters
No. of courses limited by no. of studios
Marketing issues are very important
Feedback from students will influence no. of courses
Similarity of syllabus with other universities affects inflow
EDUSAT Model
Transmission through EDUSAT satellite
Dedicated 1 Mbps uplink and 500 kbps downlink
Student Interactive Terminals (SIT) for reception
Currently 72 Remote Centre (RCs), mostly engineering colleges
RC coordinators and Instructors to ensure proper functioning
EDUSAT Model
Causal Loop Diagram
EDUSAT Model
Stock and Flow Diagram
Results
Effect of relevance of courses
Results
Optimal: 20 courses and 0.7 marketing will reach 373
Number of courses vs. number of students Marketing vs. number of students after 18 months
Results
Effect of Distribution of Incoming Grants
Optimal mix : 20 courses and 0.7 on marketing
Grants Enter Here
Observations
Attention needs to be paid on publicizing
CDEEP programs and encouraging student to join CDEEP
Effect of grants visible after 12 months
Optimal mix : 20 courses and 0.7 marketing efforts
SDModel : Features
Current simulators are all proprietary applications
very limited collaboration among them
No truly successful open source System
Dynamics model builder currently available.
Studied SystemDynamics Simulator[8]
Huge code without proper documentation
Prepared a higher level flowchart of a model and its constituent model components.
Referred an initiative SD Info Model[9]
System Dynamics Model
Dark line shows containment
Dotted line shows information
flow
Conclusion
System Dynamics proved to be an important tool for modeling CDEEP system
Models were verified by CDEEP staff
Results obtained may help in improvement of existing system
Future work
Recommendations made may be validated by implementing them over the actual CDEEP system
Publication
Poster titled “Using System Dynamics to Model and Analyze a Distance Education Program” accepted in International
Conference on Information and
Communication Technologies and Development (ICTD) 2010.
References
[1] McKansey Global Institute. Report on Emerging global labour market,2005.
[2] John Morecroft, Strategic modeling and business dynamics: a feedback systems approach , Page no. 106
[3] Deepak B. Phatak Kannan M. Moudgalya and R. K. Shevgaonkar.
Engineering education for everyone: A distance education experiment at IIT Bombay. Frontiers in Education, 2008.
[4] System Dynamics Modelling, A Practical Approach, Chapman &
Hall, 1996.
[5] http://www.cdeep.iitb.ac.in/
[6] http://www.vensim.com/
[7] http://www.public.asu.edu/~kirkwood/sysdyn
[8] http://sourceforge.net/projects/system-dynamics [9] http://sourceforge.net/projects/sdinfomodel
Thank You
Appendix1 : EDUSAT Model
Variable Equation/Initial Value
Number of students INTEG (inflow-outflow, 100)
Inflow DELAY FIXED( (Average student satisfaction
*Number of RCs*Number of transmitted
courses*Quality of Transmitted Video) /15, 2, 50 )
Outflow (1-Average student satisfaction)*(1-Quality of Transmitted Video)*Number of students/10 Marketing about
CDEEP programme
IF THEN ELSE(Grants from MHRD>0, 0.7 , 0.2 )
Relevance with other university syllabus
0.8 Quality of
Transmitted Video
Equipment Condition at RC
Appendix2 : EDUSAT Model
Variable Equation/Initial Value
Number of RCs DELAY FIXED( RC Instructor's
Motivation*Marketing about CDEEP programme*600,12, 25 )
Equipment
Condition at RC
IF THEN ELSE(Support staff for Equipment Maintenance>10, 0.8, 0.5 )
Incentives for RC Instructor
0.5
Grants from MHRD STEP(1e+08, 24 ) Number of
transmitted courses
IF THEN ELSE(Number of
studios*11>10+Feedback from
students/5+2*RC Instructor's Motivation, INTEGER (Number of studios*6+Feedback from students/10) , Number of studios *11)