The container loading problem, a real hard problem, is usually difficult to obtain even a suboptimal solution because of not only multiple complicated restrictions but also of multiple objectives. In this paper, a heuristic algorithm is proposed for solving the container loading problem in the real-world. The algorithm is based on Drum-Buffer-Rope presented in the Theory of Constraints and the multi-agent cooperative negotiation strategy. A particular attention is focused on improving the constrained agent by striving for the trade-off of restrictions and cooperative negotiations, so that the final solution can arrive its biggest profit. Since many real-world problems are restricted by many complicated restriction that are difficult to be satisfied simultaneously, a method used by human experts called restriction relaxation is embedded in the proposed algorithm, which makes the algorithm have a high degree of flexibility. In this paper, the proposed algorithm is also compared with other two classical optimization algorithms based on Local Search and Tabu Search.