On autonomous agent modelling for virtual offshore environments

Abstract

This study explores how simulation based training for offshore emergency situations can embrace software agents who exhibit human-like behaviour when exposed to hazardous situations such as fire aboard, smoke or explosions. Simulation based training uses a virtual environment to expose a participant to scenarios related to mustering events on offshore oil and gas platforms. These scenarios are also relevant to a number of other industrial applications, as they help rehearse for emergency situations such as installation fires. The agent model proposed here exploits the concepts of similarity-matching and frequency gambling as the primary knowledge retrieval methods and uses the agent’s reliability based selection of appropriate knowledge-units to make a decision in the event of a hazard. The agent’s reliability is a probability that it acts rationally, and is estimated as a function of the agent’s mental modalities: stress, panic, fear, overconfidence and distraction. The effects of these modalities during simulated harsh weather conditions and hazardous events are presented in the form of computer simulations. These simulations show that the use of the agent-model in a training software would enhance the scope of learning by exposing the human participant to more natural human-like behavior during a simulated hazardous event.