NEURAL NETWORKS FOR ONLINE TRAINING EVALUATION IN VIRTUAL REALITY SIMULATORS
Santos Machado, Ronei Marcos de Moraes · 2004
Abstract ⎯ In several research areas, virtual reality environments have been constructed for training objectives. The goal is to immerge the user into a virtual world to provide realistic training and realistic interactions. However, it is important to know the quality of training to provide a status of the user performance. An online evaluation system allows the user to improve his/her learning because it can identify, immediately after the training, where he/she committed mistakes or presented low efficiency. To evaluate the user’s performance it is necessary collect data from training. In some applications, data collected from user’s interaction cannot be adequate to classical statistical distributions. To solve that problem we propose the use of Neural Networks to online evaluation of procedures in virtual reality simulators. This approach is very simple and it can be used for Web-based simulation evaluation, using plug-ins or agents to collect information about the different variables of user’s simulations.