This paper proposes a new heuristic approach based on the Particle Swarm Optimization (PSO) for the Multidimensional Knapsack Problem (MKP). Instead of the penalty function technique usually used to deal with the constrained problem, a heuristic repair operator utilizing problem-specific knowledge is incorporated into the modified algorithm. Computational results show that the new PSO based algorithm is capable of quickly obtaining high-quality solutions for problems of various characteristics.