Distributed planning in a distribution centre

1
Distributed planning in a distribution centre

• Patrick De Causmaecker, Peter Demeester, Greet
  Vanden Berghe, Bart Verbeke

2
overview

• problem
• scheduling of departments personnel, with many
  strong and soft constraints
• complexity of this problem
• solution
• local scheduling with TS
• distribution with agents and negotiation
• results
• distribution centre

3
Distribution centre DistriMedia

• Books and cards come in from the printer
• Orders come in from booksellers
• Boxes with books and cards go out

4
(No Transcript)
5
qualification table of distribution centre
6
complexity of the problem

• scheduling of
• 19 employees
• 12 tasks
• 58 hours (one week)
• number of possible solutions
• 219212240 271
• 1.000.000.000.000.000.000.000

7
The way to go

• initial assignment of personnel to tasks
• based on history
• scheduling per task with tabu search
• (random search, with a memory)
• distribution of personnel between tasks
• (agent technology/negotiations)

8
Initial Assignment of personnel
Task 1 Tabu Search
Task 3 Tabu Search
Task 2 Tabu Search
First Solution
First Solution
First Solution
Negotiation
Final solution
Final solution
Final solution
9
History record
10
work load for each task/day

• receipt 25 hours -gt 3 employees (? 1 hour)
• picking 55 hours -gt 7 employees (! 1 hour)
• put on 9 hours -gt 1 employee (? 1hour)
• reject2 hours -gt 0 employees (? 2 hours)
• end station 8 hours -gt 1 employee (ok)
• wrapping 2 hours -gt 0 employees (? 2 hours)
• relocation 8 hours -gt 1 employee (ok)
• transport 14 hours -gt 2 employees (! 2 hours)
• returns 6 hours -gt 1 employee (! 2 hours)
• administration 9 hours -gt 1 employee (? 1 hour)
• shopping 1 hour -gt 0 employees (? 1 hour)

11
scheduling per task with tabu search

• random initial solution (in our case, this is
  easy)
• moves to solutions within a neighbourhood guided
  by a memory
• (some moves become forbidden- tabu)
• constraints are checked with numberings
• good method for scheduling

12
distribution of personnel between tasks or
departments
TabuSearch
Take most expensive
Forward
Evaluate
Restart
Add Cost coverage
Take cheapest
Search new problem
13
Results assignment

• receipt EL,MA, MV
• picking CM, CV, EV, KC, LD, RV, SV
• put on RD
• reject nobody
• end station JD
• wrapping nobody
• relocations JC
• transport IM, TD
• returns KG
• administration VV
• shopping nobody

14
work load for each task/day

• receipt 25 hours -gt 3 employees (? 1 hour)
• picking 55 hours -gt 7 employees (! 1 hour)
• put on 9 hours -gt 1 employee (? 1hour)
• reject2 hours -gt 0 employees (? 2 hours)
• end station 8 hours -gt 1 employee (ok)
• wrapping 2 hours -gt 0 employees (? 2 hours)
• relocation 8 hours -gt 1 employee (ok)
• transport 14 hours -gt 2 employees (! 2 hours)
• returns 6 hours -gt 1 employee (! 2 hours)
• administration 9 hours -gt 1 employee (? 1 hour)
• shopping 1 hour -gt 0 employees (? 1 hour)

15
Results distribution
16
Future work

• Reassignment of personnel
• Other negotiation models....
• Other (larger) companies problems

17

• Questions...?