Chapter 10 - Optimization in CSPs

Abstract

This chapter discusses constraint satisfaction optimization problems (CSOP), partial constraint satisfaction problems (PCSP), techniques for tackling the CSOP, and minimal violation problem (MVP). The CSOP and the PCSP are motivated by real life problems such as scheduling. In the CSP research community, research in CSOP and PCSP is not as abundant as research in the standard CSP. Most CSP techniques that are applicable to finding all solutions are relevant to solving CSOPs. The examples of such techniques are solution synthesis, problem reduction, and the fail first principle. Such techniques are more effective when the problem is tightly constrained. The most general tool for solving CSOPs is branch and bound (B&B). However, as CSPs are NP-hard in general, complete search algorithms may not be able to solve very large CSOPs. Preliminary research suggests that genetic algorithms (GAs) might be able to tackle large and loosely constrained CSOPs where near-optimal solutions are acceptable.