Abstract-In this paper, we propose an improved memetic algorithm by combining an evolutionary algorithm (EA) with Monte Carlo simulation (MCS) to identify the robust number of vehicles in environments with shuttle transport tasks (ESTTs). ESTTs are very common settings where vehicles shuttle between pickup and delivery points to transport goods. Examples of ESTTs are military bases, warehouses, manufacturing floors, and container terminals. In this study, MCS works as a local search to take into account risks of disruptions and guide the EA towards more reliable solutions. The disruptions arise from changes in the travel time of vehicles which can be caused by any breakdown, collision, accident and deadlock. Identifying the robust number of vehicles can improve sustainability and reliability of ESTTs against possible changes. This paper improves our previous attempts in which we studied a combination of an EA and MCS. Results of those studies showed the process of sampling in MCS is very time consuming. This prevents the EA from having an accurate estimation of the robust solutions within reasonable time. This paper improves the performance of the EA to make it possible to reach high quality solutions in reasonable time to make it practical for the realworld applications. Firstly, it proposes a new sampling technique to generate samples that reflect the worst-case scenarios better. This helps the EA to find better robust solutions using the fewer number of samples. Secondly, it proposes a new adaptive sampling technique to adjust the number of samples during evolution. To evaluate the algorithm we tested it in one of the most common ESTTs: real-world container terminals. Experimental results show that by such improvements the performance of the EA is improved significantly, making the proposed algorithm perfectly usable for its real-world case studies.