A symbolic-interval inference approach for constraint satisfaction: implementation on kinematics applications

Abstract

In this study, a general solution methodology is presented for continuous constraint satisfaction problems. The performance of the method is demonstrated on challenging kinematics problems where the goal is to identify all real solutions. The developed approach assumes a collaborative methodology that integrates interval partitioning (IP), symbolic-interval inference (SII, a novel subdivision direction selection method) and local search. The resulting methodology aims at reducing the search space and discarding infeasible subspaces effectively and reliably during the early stages of the search so that exact solutions can be identified faster by local methods. SII conducts parallel multi-variable bisection with the aim of achieving immediate reduction in total box infeasibility of the majority of sibling boxes generated in every partitioning iteration. Empirical results are obtained on seven kinematics applications and compared with results obtained by other symbolic-interval methods.