A cooperative based Guided Local Search (GLS) framework has been proposed to deal with difficult combinatorial optimization problem in [1], [2]. This framework is called Population Based Guided Local Search (P-GLS). In P-GLS, we study how GLS performance can be further enhanced thorough designing a cooperative mechanism based on the search principle Proximate Optimality Principle (POP) [3]. The experimental results on the traveling salesman problem (TSP) shows the effectiveness of P-GLS compared to original GLS. The aim of this paper is to test the PGLS on another combinatorial optimization problem, namely Multidimensional Knapsack Problem (MKP). The experimental results also confirm the effectiveness of P-GLS compared to the original GLS and the parallel GLS algorithm without collaboration.