Team training in complex domains often requires a substantial number of resources, e.g. vehicles, machines, and role-players. For this reason, it may be difficult to realise efficient and effective training scenarios in a real-world setting. Instead, part of the training can be conducted in synthetic, computer-generated environments.
Reinforcement learning for improved utility of simulation-based training
In his doctoral thesis, Johan Källström has studied how the machine learning method reinforcement learning can be used to construct synthetic agents in support of simulation-based pilot training. The focus has been on methods that can find a balance between multiple objectives, for instance different training objectives, to optimise the utility of the user. The thesis presents agents that can provide support to instructors by efficient learning of a set of Pareto optimal policies, and by efficient adaptation to user needs in operational training systems.
To the thesis:
Johan Källström, to the left, and Frans Oliehoek, University of Technology, The Netherlands, Chair of examining committee. Photo credit David Bergström
