[2E1-OS-13a] AIと制約プログラミング(1/3)

[2E1-OS-13a-01] A Study on ASP-based Integration of Systematic and Stochastic Local Search

〇Kazuya Kuwahara¹, Naoyuki Tamura², Mutsunori Banbara¹ (1. Nagoya University, 2. Kobe University)

Keywords:Answer Set Programming, Large Neighborhood Search, Timetabling

In this paper, we describe an approach to integrating systematic search and stochastic local search based on Answer Set Programming (ASP). We propose a heuristic method suitable for ASP, called Large Neighborhood Prioritized Search (LNPS). In LNPS, an initial solution is gradually improved by destroying and re-searching the solution one after another. The resulting system reads and combines a problem instance of ASP facts and first-order encoding for problem solving, which is subsequently solved by LNPS algorithm implemented using Python interface of ASP solver clingo. We establish the competitiveness of our approach by empirically contrasting clingo and more dedicated implementations, through curriculum-based course timetabling.