A design proposal of a game-based professional training system for highly dangerous professions

Abstract

Recently, modern society frequently faces local wars, terrorism, earthquakes, fire accidents, epidemics and coal mining accidents. Members of highly dangerous professions must obtain rigorous training so that they could bear the great historical mission. Generally speaking, these professions include armed forces, special police force, fire department, astronauts and mine disaster rescue troop. The game-based professional training systems for highly dangerous professions have their own distinct requirements. The aim of game-based learning systems is not only the study of declarative knowledge, but also entraining procedural knowledge through repeated practice until it becomes an automatic skill. The result of highly dangerous professional training is extremely important, since if a trainee dose not master the basic knowledge and skill, they could be in grave danger; the trainee's mental qualities should be continuously prompted by the training system so that they could be act intuitively under the most execrable circumstance. Based on requirements analysis and taking the case of mining rescue into account, we divide the whole training system into three parts: machine learning subsystem, brain information subsystem and credit-assignment subsystem. The machine learning subsystem (as know as serious game subsystem), contains the audio-visual coherency analysis, semantic annotation of a scene based on association memory, cooperating management of audio-visual cross-modal signals, personalization rendering of a scene. The brain information subsystem includes functions for receiving, storing and analyzing trainee's trial data based on visual and auditory signals from EEG, sEMG and psychological tests. The credit-assignment subsystem involves trainee's profiles and effect evaluation which are sent from brain information subsystem to machine learning subsystem, while the plan of knowledge learning, the result of skills training and consequence of the desensitization trial are sent as the feedback to brain information subsystem. Therefore, the whole framework works as a reinforcement learning system. The kernel of this system is the cooperating learning schema of audio-visual cross-modal signals. Furthermore, in this system the main visual signals contain scene textures, 3D character animation, 3D scene animation, while the main auditory signals contain the realistic sound, the on-the-spot orders, the on-the-spot yells and background music. In the light of cognitive principles, the following factors should be considered when a game-based leaning system is designed: (1)The working memory including phonological loop and visuo-spatial sketchpad act as two slave systems, play the role of dual sensory channels so that semantic coherency of the visual and the auditory data could be combined with the prior knowledge to be formed as long-term memory; (2) A goal of cooperative learning for audio-visual cross-modal signals is to create an approach which can process verbal information(like the realistic sound and the on-the-spot orders) and non-verbal information(such as 3D character animation as well as 3D scene animation) from the two separate subsystems; (3) Schema acquisition (based on Theory of Cognitive Load -TCL) should be a primary means of learning, and the automation of cognitive process (including declarative knowledge procedural knowledge) will be used to reduce working memory load.