Until today, the driving behaviour of the surrounding traffic used in the simulation technology was according to the traffic laws, i.e. the simulator driving environment is always smooth and the driver focuses only on his/her own driving and the specific tasks that are requested by the respective program/exercise that he/she follows. Therefore, when training a driver trainee with the simulator, there is a need to present a realistic traffic environment, i.e. among the traffic participants, also include those that do not behave as expected, either because they brake the law or because they are not able to estimate a given situation and react properly. Within the TRAIN-ALL project (co-funded by the EC, 6th FP), a number of innovative simulator modules were developed for motorcycle riding, passenger car (both for novices and emergency drivers) and truck driving. The new tools include also VR-based immersive simulation tools, as well as a common architecture and a modular simulator design process for multi-user, group training. Among the developed tools is the Ambient Intelligence (AI) module constituting an important novelty in terms of its concept and functionality. It extracts the profile of actual drivers in simulator scenarios and transforms it to surrounding vehicle behaviour profile (thus creating natural traffic participants in simulator scenarios). The inclusion in the simulation of other traffic participants, that interact naturally (not always legally though) with the trainee, allows to "fully immerse" the driving simulator into a realistic traffic environment. Thus, it is different from just introducing artificially illegal or varying traffic behaviour to surrounding vehicles, as in this case, the introduced behaviour is a natural and random one. By mixing "true" traffic simulation and multi-user driving simulator, a new generation of computer-based training tools may emerge. Thus, the aim here is to have a more natural traffic behaviour in the driving simulation-based training process, so that the trainees can be trained on the real - and sometimes unexpected - traffic behaviour of other road users.