Post-Optimizing Individual Activity Plans through Local Search

Abstract

Post-optimization through local search is known to be a powerful approach for complex optimization problems. In this paper we tackle the problem of optimizing individual activity plans, i.e., plans that concern activities that one person has to accomplish independently of others, taking into account complex constraints and preferences. Recently, this problem has been addressed adequately using an adaptation of the Squeaky Wheel Optimization Framework (SWO). In this paper we demonstrate that further improvement can be achieved in the quality of the resulting plans, by coupling SWO with a post-optimization phase based on local search techniques. Particularly, we present a bundle of transformation methods to explore the neighborhood using either hill climbing or simulated annealing. We present several experiments that demonstrate an improvement on the utility of the produced plans, with respect to the seed solutions produced by SWO, of more than 6% on average, which in particular cases exceeds 20%. Of course, this improvement comes at the cost of extra time.