<p>The Vehicle routing problem (VRP) holds a preeminent status within the domain of combinatorial optimization, representing a classical and foundational conundrum in the context of freight transportation. Its significance transcends its immediate applications, as it finds widespread utility in diverse transportation logistics and distribution systems. The particulars of the task, along with its associated constraints, give rise to various VRP variants, including the capacity vehicle routing problem (CVRP), Time-Window, Pickup and Delivery, Multiple Depot, Stochastic, and Split Delivery. These variants encompass multiple objectives, aiming to optimize factors such as time, costs, and the reduction of carbon emissions. Notably, the problem CVRP, involving vehicles of different sizes, stands out as a distinctive variant that necessitates a dynamic interaction between the exploration and exploitation phases due to its specific characteristics. This modification aims to achieve realistic delivery volumes, economize on toll booth expenses, reduce carbon emissions, and optimize travel distances when coordinating the routing of vehicles with varying sizes. The paper presents "mCO," an improved version of the CO algorithm, designed to solve CVRP variants with different vehicle capacities by integrating CO with OBL and RWS mechanism. To validate the robustness of mCO, our study conducted performance evaluations in three distinct CVRP scenarios. Initially, we assigned 30 customers to assess its resilience. Subsequently, we applied mCO to real-world tests involving the distribution of cement using vehicles of limited capabilities to serve 70 customers in Vietnam. Our experimental results consistently demonstrate that mCO outperforms other specialized techniques designed for limited-capacity CVRP. This positions mCO as a decision support tool for suppliers, especially in transporting cement using limited-capacity vehicles.</p>