The major international airport, Amsterdam Airport Schiphol, is redesigning its security checkpoints for 2024, for which a novel type of security checkpoint configuration is considered. In this redesign, one seeks to find the optimal checkpoint configuration; however, this remains non-trivial. The checkpoint configuration problem requires a bottom-up simulation approach to realistically evaluate checkpoint performance and an efficient optimisation approach to optimise computationally intensive simulation models. This research aims to address the checkpoint configuration problem by developing and evaluating a novel simulation optimisation (SO) method for optimal configuration of airport security checkpoints. This is achieved by combining an agent-based simulation model with an improved Tabu Search (ITS) optimisation technique. The ITS is extended from the conventional TS through three stages. Firstly, a neighbourhood decomposition strategy (NDS) limits the number of neighbours that require simulation, suppressing the computational effort. Moreover, local optima and the repetition of solutions are prevented with Frequency Memory (FM) and a Reactive Tabu list (RTL). Finally, aspiration criteria prevent significant deteriorating moves in the search process. It was shown that agent-based simulation models can realistically represent security checkpoints. The main conclusions show that the ITS converges faster than the classical TS and that it can find a better solution within the iterative domain. The NDS is crucial for reducing the computational time of agent-based, computationally intensive simulation models. Furthermore, this research shows that FM and RTL prevent solutions from repeating in the search; however, FM and RTL have no impact on the search procedure of this specific problem as the probabilistic properties of the NDS counteracted the mechanisms.