In traortation systems, delivery of goods from source to destination is usually carried out by intermediaries. These intermediaries carry out activities such as storage and re-distribution of goods. These types of systems are known as multi-echelon systems. Each echelon of these systems is part of the distribution of the system's products. Multi-echelon systems are usually driven by local managers to Achievement of objectives such traffic law and environmental regulations. On the other hand, companies whose scope of activities includes these areas They need to adapt themselves to these rules. The limitation imposed by these rules prevents large vehicles from entering the cities. Therefore, shipping companies should use authorized vehicles to continue their activities within cities. These changes in the rules have led companies to use a new system to maintain their activity and keep costs down. One of the multi-echelon systems which, while eliminating the constraints of the rules, have a fair level of cost, are two-echelon distribution systems. In this system, companies create middle depot around cities. Then, large vehicles deliver the goods from the main depot to the middle depot. After that small vehicles that are compatible with urban laws The goods are traorted from mid-depot to customers. On the other hand, distributors always face limited loads. Failure to take into account these limitations may lead to a change in the company's planning at the last minute And in the form of increased costs or even the infisibility to run some programs. Therefore, in this thesis, an integrated model of the two-echelon vehicle routing problem and three-dimensional loading constrait of goods are presented. Subsequently, for model validation, the problem is measured by different samples in small dimensions And for large examples, there are three adaptive large neighborhood search algorithms, adaptive large neighborhood search algorithms with sinus selection, adaptive large neighborhood search population base with sinus selection algorithms Has been used .The results show that the adaptive large neighborhood search population base with sinus selection algorithms yields better results than the other two algorithms.