On Transitions Functions Model for Decision-Making in Offshore Operations

Offshore operations occur in highly critical environments where safety and trust in the systems are highly relevant. These systems are complex, extremely slender, subject to long transient periods. They need to deal with sparse data measurements and very often unexpected situations occur. Higher level of automation has been proven to be beneficial in tackling these aspects and contribute to a safer and more efficient operation. The research presented in this paper aims to advance the use of autonomous decision-making in critical operations by presenting a framework for automated learning of transition functions model by use of physics-based models of the processes. Transition functions model is a bottleneck in adopting in practice algorithmic techniques such as Markov Decision Problem and Reinforcement Learning. The research presented in the paper addresses challenges in modelling transition functions when few learning data sets are available.