Updated model incorporates future scenarios

PhD researcher Robbert-Jan Merk (NLR) recently received a best paper award in Dalian (China) for his paper ‘Action selection using Theory of Mind: a case study in the domain of fighter pilot training’.

Publication  The subject of Merk’s paper is a cognitive computer model that he himself developed. This model simulates the human capacity for estimating the thoughts and motivations of other people, also known as Theory of Mind.

What is new about this model is that it incorporates uncertainty in the discourse, while also including reasoning about future scenarios, two aspects not previously incorporated in Theory of Mind models. Merk has already tested his new model on a human opponent in the NLR’s simulation environment. For this, the NLR simulation was linked to the NLR’s Fighter 4-Ship at Volkel Air Base. Mission trainings in formation are simulated and studied in the Fighter 4-Ship simulator. Merk works for the NLR’s Training, Simulation & Operator Performance section.

PhD research  

With his PhD research, Merk aims to improve the realism and intelligence of computerised opponents for flight trainings conducted in simulators. This will allow pilots who train in simulators to receive better trainings. The PhD project is affiliated with a larger NLR project, called the ‘Smart Bandits’ project. In addition to this new Theory of Mind model, Merk has also developed models about situational awareness (the ability to develop a mental image of your surroundings), a decision making model and a surprise model.

Smart Bandits project  

In the framework of the Smart Bandits project, NLR is collaborating with VU University Amsterdam (where Merk’s PhD supervisor is employed) on a ‘computer-generated enemy’. In this project, NLR provides these virtual enemies with greater ‘intelligence’, which in turn makes the fighter pilot trainings more realistic. Merk conducts research of the human behaviour in ‘agents’. These agents are independent software components that are autonomous and intelligent and can be equipped with the model that Merk has developed. The aim is for these Smart Bandits’ to operate in future as fully developed adversaries in various training scenarios.