Shifts Solver

Finds a solution to fitting employees with preferences in a schedule for shifts-based operation using heuristical backtracking.

Shifts Solver is a Google AppEngine application. To try Shifts Solver install AppEngine SDK and/or an AppEngine IDE which allows you to run Shifts Solver localy or to upload it to Google infrastructure (you may use free AppEngine tier to run it). I implemented Shift Solver using Eclipse IDE for Java with AppEngine plug-in. Simply follow how-tos to set up the development environment and then use Eclipse project that can be found in my Git repository to import Shifts Solver project to Eclipse. Alternatively you may use your favorite tool/IDE to build and deploy the project.

Overview

Home:

[](http://mindforger.com/project/s2/s2home.png)

Employees management:

[](http://mindforger.com/project/s2/s2employees.png)

Month preferences management:

[](http://mindforger.com/project/s2/s2preferences.png)

Solution management:

[](http://mindforger.com/project/s2/s2solution.png)

[](http://mindforger.com/project/s2/s2dlouhan.png)

[](http://mindforger.com/project/s2/s2allocation.png)

[](http://mindforger.com/project/s2/s2solutionEdit.png)

Functional Specification

Employee roles:

• editor
• staffer
• morning sportak
• sportak

Employee jobs:

• fulltime
• part time

Shift types:

• workdays
  • morning shift
    • 1x editor
    • 1x 6AM staffer
    • 1x 7AM staffer
    • 1x 8AM staffer
    • 1x sportak (can be morning sportak)
  • afternoon shift
    • 1x editor
    • 4x staffer
    • 1x sportak
  • night shift
    • 1x staffer (MUST: staffer, editor, morning sportak i.e. not sportak)
• weekend
  • morning shift
    • 1x editor (MUST: same as Friday afternoon)
    • 1x 6AM staffer
    • 1x sportak
  • afternoon shift
    • 1x editor (MUST: same as morning)
    • 1x staffer
    • 1x sportak
  • night shift
    • 1x staffer (MUST: staffer, editor, morning sportak i.e. not sportak)

Schedule:

• For one month
• Dependencies to previous month schedule (e.g. editor Friday to Sunday continuity)

Employee preferences:

• preference type:
  • YES: I want this shift/day
  • NO: I cannot make this shift/day
  • WHATEVER: I don’t care
• preference target:
  • (whole) day
  • any shift type (see above)

Rules:

• MUST: shift counts for one job (except editor continuity below)
• MUST: any employee role may serve on Monday to Saturday nights
• MUST: part time employee to serve on Friday night
• MUST: part time employee to serve on Saturday night
• MUST: fulltime employee to serve on Sunday night
• SHOULD: editor to serve on Sunday night (TODO)
• MUST: employee birthday is marked as NO preference
• MUST: national holidays are solved in the same way as Sunday (PARTIAL)
• Editor Friday to Sunday continuity:
  • MUST: particular editor to serve Friday afternoon + Saturday morning and afternoon + Sunday morning and afternoon shifts
  • MUST: editor continuity counts for 1 + 3 (Friday + weekend) = 4 shifts in job (note that it would be 5 otherwise) (TODO)
  • SHOULD: editor to have same staffer and sportak in continuity (TODO)
• MUST: employee cannot have more than one shift in a day
• MUST: there must be at least 8 hours between two shifts served by the same employee
  • MUST: if employee serves afternoon shift, then first shift it may serve is morning 8AM (6AM and 7AM CANNOT be served by this employee)
  • MUST: if employee serves night shift, then it CANNOT serve morning shift next day
  • MUST: if employee serves on Saturday and Sunday then it CANNOT serve on Monday (TODO)
  • SHOULD: if employee has any shift in 5 consecutive days, then it SHOULD get one day FREE
• Fulltime employee jobs:
  • MUST: job shifts = workdays * 8 / 7.5
  • MUST: fulltime employee must get max shifts +/-1 shift
  • MUST: 2 night shifts in month at most
  • MUST: if employee has holidays on workday, then it adds 1 to her/his job shifts
  • SHOULD: morning and afternoon shifts to be balanced 50%/50%
• Part time employee jobs:
  • MUST: max job shifts = fulltime job shifts / 2

Technical Architecture

Frontend:

• Google Web Toolkit

Backend:

• Google AppEngine

Plan

The implementation plan:

• solver
  • strategy:
    • try to build schedule w/o need to backtrack
    • smart allocation and capacity anticipation
    • jobs for month - do I have enough?
    • calculate capacity: editors, nights, sports, …
    • backtracking to be used only in situations when you cannot allocate wisely
    • allocate user and ensure it wont fail (failure is last option causing backtrack) e.g. make sure editor has 5 shifts capacity when assigned on Friday afternoon
  • incorporate POSITIVE preferences in solver (do 2 cycles of employees and remember it in solution: want employees for day and THEN don’t care employees) —