1 / 20

Ant Colony Optimization for Software Project Scheduling and Staffing with an Event-Based Scheduler

Ant Colony Optimization for Software Project Scheduling and Staffing with an Event-Based Scheduler. Presented by: Qui Nguyen Ying Jiang. submitted to: Professor Sherwin Shirmohammadi in partial fulfillment of the requirements for the course ELG 5100. Agenda. Introduction

sonora
Download Presentation

Ant Colony Optimization for Software Project Scheduling and Staffing with an Event-Based Scheduler

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Ant Colony Optimization forSoftware Project Scheduling and Staffingwith an Event-Based Scheduler Presented by: Qui Nguyen Ying Jiang submitted to: Professor Sherwin Shirmohammadi in partial fulfillment of the requirements for the course ELG 5100

  2. Agenda • Introduction • Model Description • Related Models • Event-Based Scheduler(EBS) • Ant Colony Optimization(ACO) • Experimental Studies • Conclusion and Future Works • Q&A

  3. Introduction(1) • Software project planning challenges • Workload, cost estimation. • Project scheduling, resource allocation. • Solutions: • Program evaluation and review techniques(PERT). • Critical path method(CPM). • Resource-constraints project scheduling problem(RCPSP) • Problems? • PERT, CPM: not consider resource allocation. • RCPSP: not consider resource allocation with various skills. • Task scheduling and resource allocation separated! • ...inefficient resource allocation, poor mgmnt. performance

  4. Introduction(2) • Other modern approaches: • Event-Based Scheduler(EBS) • Task list + planned employee allocation matrix. • Solve task scheduling and resource allocation issues • Adjust workload, d/w resource conflicts efficiently. • More practical, flexible with task preemption. • Reducesearch space size, increase search process. • Ant Colony Optimization(ACO) • Schedule critical tasks as early as possible. • Assign tasks to suitable employees with required skills • Converge fast and perform well.

  5. Model Description(1) • Description of Employees • People-intensive activity. • Wages, skills, working constraints. • Objective: assign employees to suitable tasks • Salary= • bsi-basic salary/month • hsi-normal work salary/hr • ohsi-overtime salary/hr • nh-legal working hr/month • maxhi-max. working hr/month • [joini,leavei]-time window • {S1i,S2i,... Sφi}-skill list • φ: # of skills • Sjiϵ[0,5]

  6. Model Description(2) • Description of Tasks • Precedence graph(TPG). • Proficiency of ith employee for task tj. • Total fitness of employee for tj on tth month • V=8-round(Ftj*7+0.5) • whtij# working hr of ithempl. for tj at tth month • Achievement for tj on tth month • Object function:

  7. Related Models • RCPSP: • Schedule tasks to satisfy precedence constraints • Determine order of tasks • ...empl. allocation issue? • Employee Allocation Model: • Multiskill Scheduling Model: • RCPSP extention: RCPSP + multiskillempl. modeling. • Address task scheduling, empl. allocation issues. • Task preemption?...less efficiency! • Time-Line-Based Model: • Solve RCPSP prob...but • whij-working hr of ithempl. for tj • Task scheduling, resource conflict? • Desultorytask • Large scale!

  8. Event-Based Scheduler(1) • pwhij-orig. planned working hr of ithempl. for tj • Enable task preemption. • How to adjust pwhij to whij ? • Event: t=1,join/leave,finished task.

  9. Event-Based Scheduler(2) • Advantages: • Compared to RCPSP: • Solve task scheduling and empl. allocation. • Task preemption, resource conflict addressed. • Compared to Time-Line-Based Model: • Reduce search space: O(m.n.end) to O(m.n) • Moderate granularity to tasks • More flexible. • Sufficient info for modeling.

  10. Ant Colony Optimization(1) • Algorithm solves planning problem. • Idea: • Ants search food based on special chemical(pheromone) • ACO types: • Ant colony system (ACS): chosen • Fast convergence speed based on past search experience. • Pheromone update rules • Global: best-so-far solution • Local: reduce pheromone on path to more diversity • Max-min ant system (MMAS). • How to use this idea to construct • Task list. • Employee allocation matrix

  11. Ant Colony Optimization(2) • Construct task list. • absolute position model: • Find order of tasks • Define pheromone of putting tj to kth position. • eligibleSet: tasks satisfied precedence graph constraint • 2 steps: • Step a: Put eligible tasks into eligibleSet. • Step b: k=1:n do • {- select a task from eligibleSet, put to kth of task list. • - Update eligibleSet by removing selected task, add eligible tasks • }

  12. Ant Colony Optimization(3) • Construct employee allocation matrix. • Pheromones and heuristic value • :choose ith employee for task tj • :assign k (k=25%nh,50%nh,...maxhi) of ith employee’s working hrs to tj . • : choose lower salary, higher proficiency employee.

  13. Ant Colony Optimization(4) • 2 steps construction • step a: Set all values=0 • step b: Each task, assign workload for tjby • step b1: evaluate for all employees • step b2: select available employee from eligibleSet • step b3: Assign working hrs of selected employee for tj • step b4:if (# of selected employees for tj)>= maxheadj • then: done workload assignment for tj • otherwise: go to step b2 • (maxheadj: max # of employees for task tj)

  14. Experimental Studies(1) • Experiment settings: • 80 randomly generated instances + 3 real instances • Algorithms comparison • ACO • ACO-L: ACO with local mutation • 3dGA: Time-Line-based GA using 3D with task preemption • KGA: Knowledge-based genetic for multiskill scheduling • KGA-p: task preemption from EBS implemented. • TS: tabu search (no task preemption).

  15. Experimental Studies(2) • Costs: • ACO, ACO-L: better others, ACO-L best due to local mutation • KGA-p better KGA: task premptionfrom EBS is effective • ACObetter KGA-p all cases: ACO algorithm is effective.

  16. Experimental Studies(2) • Convergence: • ACO, ACO-L: find good solutions at very beginning of search process • ACO-L further improve the performance with local mutation.

  17. Experimental Studies(3) • Plan structures: • Workload assignment stable • Resource released once task is done • Workload assignment unstable • Impractical: same task, dif. group • Delay staff #1 done to work for task 2

  18. Conclusion and Future Works • New Method Proposed • Event-Based scheduler • ACO to solve complicated planning problems • Experiments show • Representation scheme with EBS is efficient. • ACO yields better plans with lower costs, fast convergence • Future works • Employee experience implemented. • Training model implemented.

  19. References • [1] Wei-Neng Chen, and Jun Zhang, “Ant Colony Optimization for Software Project Scheduling and Staffing with an Event-Based Scheduler,” IEEE transactions on software engineering, vol39, January, 2013 • [2] C.K.Chang, H.Jiang, Y.Di, D.Zhu, and Y.Ge, “Time-Line Based Model for Software Project Scheduling with Genetic Algorithms,” Information and Software Technology, vol.50, PP.1142-1154,2008 • [3]C.K.Chang, M.J.Christensen, and T.Zhang, “Genetic Algorithms for Project Management,” Annals of Software Eng., vol15,no.10,pp.107-139,2001 • [4] M.Dorigo and T.Stutzle, Ant Colony Optimization. MIT Press, 2004 • [5] M.Dorigo and L.M.Gambardella, “Ant Colony System: A Cooperative Learning Approach to TSP,” IEEE trans. Evolutionary Computation, vol.1,no.1,pp53-66, 1997 • [6] K.Praditwong, M.Harman, and X.Yao, “Software Module Clustering as a Multi-Objective and Search Problem,” IEEE Trans. Software Eng., vol.37, no.2,pp.264-282, Mar 2011

  20. Q&A

More Related