This paper presents a hybrid AI approach for a class of overconstrained Nurse Rostering Problems. Our approach comes in two phases. The first phase solves a relaxed version of problem which only includes hard rules and part of nurses' requests for shifts. This involves using a forward checking algorithm with non-binary constraint propagation, variable ordering, random value ordering and compulsory backjumping. In the second phase, adjustments with descend local search and tabu search are applied to improved the solution. This is to satisfy the preference rules as far as possible. Experiments show that our approach is able to solve this class of problems well.
[1]
Vipin Kumar,et al.
Algorithms for Constraint-Satisfaction Problems: A Survey
,
1992,
AI Mag..
[2]
Peter van Beek,et al.
On the conversion between non-binary constraint satisfaction problems
,
1998,
AAAI 1998.
[3]
Rina Dechter,et al.
Tree Clustering for Constraint Networks
,
1989,
Artif. Intell..
[4]
Eugene C. Freuder,et al.
Partial Constraint Satisfaction
,
1989,
IJCAI.
[5]
Fahiem Bacchus,et al.
Looking Forward in Constraint Satisfaction Algorithms
,
1999
.
[6]
Alan Borning,et al.
Constraint hierarchies
,
1992
.