To achieve realism in force-on-force battle training, it is important that the major factors of the battlefield are simulated in a realistic way. We describe an architecture for battle training and evaluation which provides a framework for integrating multiple sensors, simulators and registration equipment together with tools for analysis and presentation. This architecture is the basis for the MIND system, which is used in realistic battle training and for advanced after-action review. MIND stores the information recorded in a database which is the basis for subsequent analysis of training methods and improvement of tactics and military equipment. Data collected during battle training can support both modelling of Distributed Interactive Simulation (DIS) objects and the presented Time- delayed DIS (TDIS) approach. TDIS facilitates the training of staffs and commanders on high levels under realistic circumstances without the need of trainees and trainers on the lower unit levels. Systematic evaluation and assessment of the MIND system and its influence on realistic battle training can provide information about how to maximise the effect of the conducted battle training and how to best support other applications that use information from the system.