To study traffic flow characteristics at freeways, traffic model s are used. One of these traffic models is the micro-dynamic traffic model Getram/AIMSUN2 (in this study referred to as AIMSUN2) developed at the Barcelona University of Technology (UPC). DHV Environment and Infrastructure is using this software package for traffic modelling in the Netherlands. To achieve sufficient accuracy between simulation outcomes and observations traffic models need to be calibrated. At this moment, a procedure to perform the calibration process for AIMSUN2 does not exist. The objective of this study is to design such as procedure. This study focuses on the driver behaviour on freeways. The objective of a calibration process is to estimate the unknown model parameters and to optimise the overall model performance. By definition, a simulation model generates output on multiple aspects such as flows, density and travel time. Each of these aspects is associated to it is own performance indicator. The overall performance of a model can be defined as a weighted sum of these indicators. The weights depend on the priorities of model experts, but are seldom made explicit. In this study a procedure is applied to determine these weights from revealed preferences of model experts. First, a problem analysis is performed to distinguish typical aspects of the calibration process. Possibilities to estimate unknown model parameters and optimisation processes to compute the distance between simulation outcomes and observations are defined. The optimisation processes is divided into a one dimensional and a multi-criteria analysis. During the one dimensional optimisation single criterion variables are observed while during the multi-criteria analysis the total performance (all examined criterion variables) of a model is examined. Various estimation techniques are distinguished to determine the model parameter values which are used within the mentioned optimisation processes. Next, a description of the micro-dynamic traffic model AIMSUN2 is given. The computation of the driving speed for vehicles at sections is discussed. Also, the sub models defining the individual driver behaviour are described and examined. A qualitative analysis is performed to define the influences of model parameters on criterion variables. Parameters are classified into fixed parameters, model dependent parameters and location dependent parameters. Basic network configurations for simulations are proposed. A stretch, a fork and a join are distinguished as basic configurations By means of this classification, specific driver characterist ics are related to network configurations. An objective function is proposed to compute a distance measure between observations and examined criterion variables. The following criterion variables are included in this objective function: flow, density, speed, travel time and location/ length traffic-jams. It is proposed to compute the mentioned distance measure by a mean absolute error proportional (MAEP) function. Experiences of model experts are used to define the weights of examined criterion variables A sensitivity analysis (quantitative analysis) is performed to define the sensitiveness of model parameter values on criterion variables. A calibration procedure is proposed based on experiences obtained during this calibration study. In later versions of AIMSUN2, it should be possible to combine the findings of this study into a network wide applicable calibration model. An important aspect of this study has been the study of the categories to which model parameters are assigned. In the AIMSUN2 (version 3.2 which is not publicly available and used for this study) various model parameters used in the model implementation are assigned to a category which is less suitable for calibration. This classification implies that there are only limited possibilities for tun i ng the model during the various phases of the calibration process.