Analysis of Agile and Multi-Agent Based Process Scheduling Model

International Refereed Journal of Engineering and Science (IRJES)
ISSN (Online) 2319-183X, (Print) 2319-1821
Volume 4, Issue 8 (August 2015), PP.23-31
www.irjes.com 23 | Page
Analysis of Agile and Multi-Agent Based Process
Scheduling Model
Y. M. Malgwi1.*
, N. B. Shelleng2.
, K. J. Danjuma3.
1
Computer Science Department, Modibbo Adama University of Technology, Yola, - Nigeria
2
Mathematical Sciences Department, Adamawa State University, Mubi, -Nigeria.
3
Computer Science Department, Modibbo Adama University of Technology, Yola, - Nigeria.
Abstract:- As an answer of long growing frustration of waterfall Software development life cycle concepts,
agile software development concept was evolved in 90’s. The most popular agile methodologies is the Extreme
Programming (XP). Most software companies nowadays aim to produce efficient, flexible and valuable
Software in short time period with minimal costs, and within unstable, changing environments. This complex
problem can be modeled as a multi-agent based system, where agents negotiate resources. Agents can be used to
represent projects and resources. Crucial for the multi-agent based system in project scheduling model, is the
availability of an effective algorithm for prioritizing and scheduling of task. To evaluate the models, simulations
were carried out with real life and several generated data sets. The developed model (Multi-agent based System)
provides an optimized and flexible agile process scheduling and reduces overheads in the software process as it
responds quickly to changing requirements without excessive work in project scheduling.
Keywords:- Agile Software development process methodologies, Extreme programming, Agents and Multi-
agents.
I. INTRODUCTION
Agile software development is a group of software development methods based on iterative and
incremental development, where requirements and solutions evolve through collaboration between self-
organizing, cross-functional teams (Gonzalez & Pilar, 2009). It promotes adaptive planning, evolutionary
development and delivery. (Hock, 2009).
The most popular definition of Agent was proposed by (Wooldridge & Jennings, 1995): An Agent is
essentially a special software and hardware component that has the characteristics of autonomy, sociality, and
reaction and pro-action. It is a computer system that is capable of independent (autonomous) action on behalf of
its users or owner. It can provide different interaction interface for the outsiders, and even have the
characteristics such as knowledge, belief, and intention and so on. (Pour et al. 2004).
Just as its name implies, Multi-agent System (MAS) is a system that consist of numbers of agents
within an environment which interact with one another. It can be used to solve problems that are difficult or
impossible for an individual Agent (Jennings, Sycara & Wooldridge, 1998).
In the most general case, agents can be acting on behalf of users with different goals and motivation.
To successfully interact, they will require the ability to cooperate. Coordinate, and negotiate with each other,
much as people do. Indeed, multi-agent systems enhance overall system performance, in particular along such
dimensions as computational efficiency, reliability, extensibility, responsiveness, reuse, maintainability, and
flexibility (Michael, 2002) & (Fabio et al. 2000).
Lack of penetrations of the modern agile planning tools during software development process usually
provides a ‘quick and dirty’ solution which is informally managed. Typically, informal managed planning
factors are:
i) Scheduling of tasks and resources
ii) Relationship or Communication between entities.
Although the principles of agile development rely on communication instead of rigorous planning, this
fact can be explained by the lack of easily applicable algorithm solutions. Informal approaches work well in
smaller projects but not sufficient in larger projects. As the size and complexity increase, scheduling becomes a
very complex process and advocate tool support.

Analysis of Agile and Multi-Agent Based Process Scheduling Model
As a consequence, optimized and flexible project plans are crucial issues from the economic
considerations of both customer and developer’s side. These critics underline the importance of providing a
more established model for agile process scheduling. In this article, the aim is to diminish these barriers and
implement a model that schedules the agile process.
The aim of this research is to develop a multi-agent based agile process scheduling model that is more
efficient, flexible and has limited resource management capabilities.
II. METHODOLOGY
Two mathematical models (Agile and Multi-agent based process scheduling) will be design to
mathematically represent the XP process. Also, two algorithms that will implement the mathematical model will
also be developed and simulated with real life and generated data sets to actualize their effectiveness. Visual
C++ will be used as a tool in order to ensure smooth implementation. The models created will follow the
principles of branch and bound optimization technique in order to generate an optimized schedule for the XP
processes stated earlier.
Problem-solving Framework
The framework will be based on agile scheduling problem solving which is modeled by a 6-phase
closed-loop XP process shown in Figure 3.1. The process contains the following functions: Selection of user
stories, breaking down of tasks, planning the release, Executing/developing the plan, Releasing the software and
Evaluating the system by the user/ customer. The arrows represent the flow of data between each of these
functions, which forms a continuous feedback mechanism.
Figure 3.1: XP Process-cycle framework (Ian, 2009)
Each of the functions in Figure 3.1 can be further decomposed to address the more specific problem of multi-
agent based agile planning and scheduling.
The analysis phase
The analysis phase aimed to clarify the problem without any (or minimal) concerns about the solution.
The analysis phase is carried out through a number of steps, described.
Use Case Diagram
Use cases been an effective way to capture the potential functional requirements of a new system. The
use case was used in representing the XP process scenario that demonstrates how the system interacts with the
external environment to achieve a specific goal. The use case diagram is produced as shown in Figure 3.2
below.

Figure 3.3: Use case diagram for XP processing (Fabio et al. 2004)
Initial Agent Types Identification
This identifies the main agent types and subsequent formation of a first draft of the agent diagram. The
following rules are being applied;
i.) Adding one type of agent per actors.
ii.) Adding one type of agent per resource. By applying the above rules to the agile process case study, the
initial diagram shown in Figure 3.3 is obtained.
Figure 3.4: Agent Diagram for XP (Fabio et al. 2004)

With reference to Figure 3.3 above, the agent diagram includes four types of elements:
1. Agent types: the actual agent types, represented by circles.
2. Humans: people that must interact with the system under development, Represented by the
UML actor symbol.
3. Resources: external systems that must interact with the system under development, represented by
rectangles.
4. Acquaintances: represented by an arrow linking instances of the above elements, specifying that the
linked elements will have to interact in some way while the system is in operation.
Responsibilities Identification
In this step, for each identified agent type, an initial list is made of its main responsibilities in an
informal and intuitive way. The artifact resulting from this process is the responsibility table.
The following rules are applied in this step:
i) The initial set of responsibilities was derived from the use cases identified in figure 3.3 above.
ii) The agents’ responsibilities were considered.
By applying the above rules to the agile process scheduling case study, the consideration of the XP
Process agent is initiated and Table 1 is produced
Responsibility table for Agents in XP Process
Agent Type Responsibilities
User agent 1. Initiates the project
2. Provides requirements inform of user stories
3. Evaluates the system at each iteration
Manager Agent 1. Gets requirements from user agent inform of stories
2. Breaks down user stories
3. Retrieves the relevant Scheduling agent
4. Tracks the activities of the Team Members
5. Releases software
6. Terminates the project
Scheduling Agent 1. Retrieves tasks and resources from manager agent.
2. Schedules tasks to resources.
Programmer Agent 1. Gets tasks from Scheduling agent.
2. Develops individual units
3. Retrieves the relevant Unit Test agent
4. Provides status to manager agent
5. Integrates units to the whole system
6. Tests the whole system
Unit tests Agents 1. Receives completed units from programmer agent for testing.
2. Carries out unit testing
Table 1: Responsibility identification table.
Acquaintances Identification
In this step, the focus was on who needs to interact with whom and the agent diagram (Figure 3.2) is
updated by adding proper acquaintance relations connecting agents that need to have one or more interactions.
An obvious acquaintance relation in the XP process case study is required between different XP process agents.
Figure 3.5: Agent diagram for XP processes depicting (Fabio et al. 2004) 2004) Acquaintance.

Multi-agent planning and scheduling process
As shown in Figure 3.5, these steps consist of:
i.) Formulating or receiving from another source (users/customers).
ii.) Structuring this objective in a form where it can be easily decomposed into a partially- ordered set of sub-
problems or jobs.
iii.) Surveying the environment for available agents and services that may be used to complete these jobs.
iv.) Mapping jobs to available services or sets of services that is capable of completing them.
v.) Determining the allocation of jobs to agents, such that the resulting schedule is optimized according to user-
defined parameters.
vi.) forwarding this solution to the appropriate agents for Evaluation
Figure 3.5: Multi-agent planning and scheduling process in the context of the
closed-loop architecture (Mark 2003).
It is important to note that the small arrows between the specific tasks in Figure 3.5 represent
precedence relations, while the large arrows between the high-level functions represent continuous data flow
and feedback between these phases. It should be noted, however, that the feedback provided by the data flow
could represent precedence relations at a higher level itself in the context of this architecture, as one phase may
not be allowed to begin its next cycle until it has received data from another phase.
Problem formulation and decomposition
In the multi-agent framework that we are considering, there are sets of agent in XP process, each with a
different set of available abilities and services. A single agent is given an objective to complete, possibly from
another agent, and it wishes to take advantage of the resources provided by these other agents in the XP process
to complete the objective more efficiently. The planning agent’s first step is to decompose its objective into a
number of tasks that can be allocated to other agents in the XP process and completed in parallel.
However, there are often many possible problem formulations for a given objective and choosing the
best way to decompose the objective which may depend on the structure of the agent organization and the
number of different service types provided by these agents. In the scenario that this research addresses, the
agent’s main goal is to choose the job precedence that produces the schedule with the most possible user stories
completion within a fixed – time.
Figure 3.5: Multi-agent planning and scheduling process in the context of the closed-loop architecture
(Mark 2003).

III. MATHEMATICAL MODEL FOR AGILE PROCESS SCHEDULING
Problem Variables
Given a number of user stories j(𝑗 ∈ 𝑊 ∶ 𝑊 = 𝑛), with resource demand Wj and iterations k (𝑘 ∈
𝑗: 𝑗 = 𝑜)with different iteration velocities Ck within a release. Because we use fixed –time scheduling 𝑜 ≜
𝐹𝑇| 𝑙 𝑘
𝑙
≤ 𝑛 Where FT is the fixed-time duration of the release and 𝑙 𝑘
𝑙
is the length of iteration k.
Lets assign each user story into one iteration so that the total required effort in iteration k does not exceed Ck
and the number of iterations used as a minimum-while both precedence (matrix) 𝑃𝑗𝑗 ′ ∈ {0,1} (where 𝑃𝑗𝑗 ′ = 1 if j
precedes j′ otherwise 𝑃𝑗𝑗 ′ = 0) and coupling relations (matrix) 𝐶𝑗𝑗 ′ ∈ {0,1} (Where 𝐶𝑗𝑗 ′ = 1 if j is coupled with j’,
otherwise Cjj′ = 0) hold.
A possible mathematical model can be in the form formulated below.
Maximize z = 𝑃𝑗 𝑋 𝑘𝑗
𝑛
𝑗=1
𝑚
𝑘=1 ………………………………... (1)
Subject to:
𝑊𝑗 𝑋 𝑘𝑗 ≤ 𝐶𝑘
𝑛
𝑗=1 𝑌𝑘 ………………………………………... (2)
𝐾′
− 𝐾 ≥ 𝑃𝐽′ 𝐽 : 𝑋 𝑘′ 𝑗 ′ = 𝑋 𝑘𝑗 = 1 …………………........ (3)
1/ 𝐾′
− 𝐾 + 1 = 𝐶𝐽′ 𝐽 : 𝑋 𝑘′ 𝑗 ′ = 𝑋 𝑘𝑗 = 1 ………….…. (4)
𝑋 𝑘𝑗
𝑚
𝑘=1 ≤ ∀ 𝑗 ∃! 𝑘 …………………………………….. (5)
Where: Yk = 0 or 1, and Xkj = 0 or 1.
Xkj = {
Yk = {
Ck is a positive integer
Wj ≤ Ck for ∀ k, j
IV. MATHEMATICAL MODEL FOR THE MULTI-AGENT BASED SYSTEM
Problem Variables
For easily formulating a mathematical model for the multi-agent based system representing our agile
software development process (more precisely extreme programming). It is inimical we identify our constraints
decision variables and then formulate our objective function.
Input Parameters
j is the index of all set of available agents A. Each agent is indexed numerically
denoted by i.
P- Immediate job precedence matrix
Where:
Pj′j = {
Q- Full user story priority matrix
Qj′j = {
B- Agent ability matrix
BjA = {
A possible mathematical model for the multi-agent based agile software development process scheduling can be
in the form formulated below.
Maximize z = 𝐵𝑗′𝑎 . 𝑋𝑗
𝑛
𝑎=1
𝑚
𝑗 =1 ′𝑎
Subject to:
Bj′a = 1 ∀ j ∈ W, a ∈ A
Xj′ a = 1 ∀ j ∈ W
Xj′ a + Xj′ a ≤ 1 ∀ (j′j) ∈ W, a ∈ A
Where:
Xj′ a = 0 or 1, Xj′ a ∈ {0, 1}
1 if user story j directly precedes j
0, otherwise
1, if user story j′ comes anytime before j
0, otherwise
1, if agent A provides the services needed to complete story j
0, otherwise
1, if j is assigned to iteration k0, otherwise
1, if iteration k is used0, otherwise

Bj′a = 0 or 1, Bj′ a ∈ {0, 1}
Algorithm
A suitable algorithm for implementing the mathematical model formulated above is provided below:
Require: (IR
∈ N)
a ∈ A, Bj′ a ∈ {0, 1}
j ∈ W, Qj′j ∈ {0, 1}, P j′j ∈ {0, 1}
Ensure: ∀ j ∃! W and ∀ a ∃! A (Xj′ a ∈ {0, 1})
1: repeat
2: a ∈ A ⊆ A*
3: VR
← 𝐵𝑗 j′a: a ∈ A
4: K, l′k
5: Ck ←VR
* l′k
6: X ← Schedule (Q, A, B)
7: until X is satisfying
8: return X.
V. RESULTS AND DISCUSSIONS
Results
The model was tested using different data sets to determine its efficiency and flexibility. The data set
used consist of seven actual deals that were collected, these includes Collateral evaluation (RA), Risk
assumption (RB), Ukrainian deal flow I (RC) & II (RD), Romanian deal flow I (RE), II(RF), and III (RG).
These are indexed RA to RG respectively (gotten from the back log of IRIS application developed by
multi logic Ltd) (Akos, 2011). All releases had same project members (16 programmers) iteration length (2
weeks), iteration velocity (30 story point), domain, customer and development methodology (XP) but were
differentiated by user stories, iteration counts, length of each iteration and release duration.
In order to create more flexibility, the user of system input the different services provided by the
available agents (programmers, managers, users, test agent and so forth). The model creates an optimal
combination of all the different services.
The agile process schedule as shown in table 4.1 provides an optimal iteration and release schedules for
the combination of an already specified constraint. The user does not have the privileged to modify the number
or type of participants that can take part in the project. This however does not mean that the schedule provided is
not optimal, but provides a maximized priority schedule as shown in table 4.1 below.
Table 4.1 Optimized agile Iteration and Release Schedule values
Keys Max = DV - Maximum Deliverable Values
IC – Iteration Count RD– Release Duration
NP – Number of Programmers NT – Number of Task
IL – Iteration Length NUS – Number of User story
For the multi-agent based agile process scheduling, we take the same data sets as that of the agile
process scheduling. In order to create more flexibility, we allow the user of system to input the different services
provided by the available agents (programmers, managers, users, test agent and so forth). The model creates an
optimal combination of all the different services. Here, we propose that the multi-agent based agile process
scheduling will create a more optimal schedule thereby having a higher Max Value as dully shown in table 4.2
below.
IC NP IL RD NT NUS Max=DV
(Z)
Release
Values
RA 20 10 10 7 10 33 400 12,800
RB 24 10 11 9 12 25 480 11,520
RC 22 10 12 10 11 27 440 11,440
RD 23 10 10 8 10 27 460 11,960
RE 21 10 9 8 12 43 420 17,640
RF 20 10 8 9 10 26 400 10,000
RG 24 10 10 11 11 43 480 20,160

IC NA IL RD NT SPA NUS Max=DV
(Z)
Release
Values
RA 20 5 10 7 10 12 33 3,000 2,880.000
RB 24 7 11 9 12 15 25 3,300 2,376.000
RC 22 8 12 10 11 13 27 3,600 2,808.000
RD 23 7 10 8 10 14 27 3,000 2,340.000
RE 21 8 9 8 12 12 43 2,700 3,402.000
RF 20 8 8 9 10 13 26 2,400 1,800.000
RG 24 6 10 11 11 10 43 3,000 3,780.000
Table 4.2 Optimized Multi-agent based agile Iteration and Release Schedule values
Keys
Max = DV - Maximum Deliverable Values
IC – Iteration Count RD – Release Duration
NA - Number of Agents PA - Priorities of service provided by agents
IL – Iteration Length NUS - Number of User story
VI. DISCUSSION OF RESULTS
The main concept of agile process scheduling is based on multiple knapsack optimization technique.
The proposed multi-agent based system covers a wide-ranging release scheduling (fixed time).
The multi-agent based system made it possible to adapt an efficient global optimization algorithm for
more flexibility and smooth iterations. The algorithms strives to prevent resources overload-which often yields
increasing delivery risks, and prevent resource underload - which captivates economically, and badly utililized
iterations.
The results reveal that the outcome of the research is an extension of readily available scheduling tools
which helps collected the process scheduling data (user stories, required effort, team velocity, etc) Therefore,
with this extension, it is believed that one can produce a flexible and efficient process scheduling system easily
based on the collected data.
The method also indicates that it requires no much time because of the good communications or
relationship between entities. It expresses dependencies between deliverables features, as it produces optimal
schedule within seconds. However, the actual major difference between the two systems (agile process
scheduling and Multi-agent based process scheduling) is the higher quality schedule realized (avoiding
underload and overload), as the multi-agent based system produces a better resources utilization and make it
possible to re-schedule the process anytime within seconds in order to support the what-if-analysis.
VII. CONCLUSION
The proposed model gives the main parameters of the typical agile process scheduling space (such as
objectives and constraints) and presents an optimization model that can be realized by optimization tools or by
implementing the suggested custom-made algorithm.
The goal of this research has been to implement and analyze the agile and the multi-agent based
process scheduling models for efficiency and flexibility. Additionally, the approach provides a more informed
and established decisions with application of what-if analysis (rescheduling the release by altering its
parameters). The findings of this result reveal that:
i) An agent provides an interoperable interface to an arbitrary system and/or behaves like a human agent
working for some clients in pursuit of its own agenda.
ii.) The multi-agent systems can model complex systems and introduce the possibility of agents having
common or conflicting goals. These agents may interacts with each other both directly (clicking on the
environment) or directly (via communication).
iii.) The system is optimal and flexible with limited resources management capabilities.
Based on the findings of this research, the Multi-agent based systems may be recommended for
instance in the field of telecommunication systems where large distributed networks of interconnected
components which need to be monitored and managed in real time.

REFERENCES
[1]. B. Fabio, C. Giovanni & G. Dominic, Developing Multi-agent system with JADE, John Wiley & sons
Ltd, 2004.
[2]. G. Pour, F. Yao & C. Yu, A Mobile Agent-Based Architecture for Mobile Systems Supporting
Distributed Software Project Management. IEEE SoftCOM, 2004
[3]. M. Hock, Review of agile methodologies in software development. International journal of research
and reviews in applied sciences, volume 1, issue 1, 2009
[4]. M. Wooldridge, An Introduction to Multi-agent System, John Willey & Sons Ltd, 2004.
[5]. F. T. Mark, (2003). Optimization techniques for task allocation and scheduling distributed multi-agent
operations. A Master thesis submitted to department of electrical and computer science Massachusetts
institute.
[6]. N. Jennings, K. Sycara & M. Wooldridge, a Roadmap Agent Research and Development. Int. Journal
of Autonomous Agents and Multi-Agent Systems. 16 (1998) 62–87.
[7]. R.Gonzalez & P. Pilar Some findings concerning requirements in agile methodologies. Product-
focused software process improvement, 32 (4), (2009) 171-184pp.
[8]. S. Akos, Conceptual scheduling model and optimized release schedule for agile environment,
information and software technology.53 (2011)574-591.
[9]. S. Ian, Software Engineering, ninth edition, Pearson Education, Inc., publishing, Addison –Wesley:
New York.
[10]. W. Michael & N.Jennings, (1995). Intelligent Agents: Theory and Practice, Knowledge Engineering
Review, 10(1995) 115-152.

Analysis of Agile and Multi-Agent Based Process Scheduling Model

More Related Content

What's hot (19)

Viewers also liked (15)

Similar to Analysis of Agile and Multi-Agent Based Process Scheduling Model (20)

More from irjes (20)

Recently uploaded (20)

Analysis of Agile and Multi-Agent Based Process Scheduling Model