Title: Simultaneously recovering running cost and Hamiltonian in Mean Field Games system

Abstract: We propose and study several inverse problems for the mean field games (MFG) system in a bounded domain. Our focus is on simultaneously recovering the running cost and the Hamiltonian within the MFG system by the associated boundary observation. There are several technical novelties that make our study highly intriguing and challenging. First, the MFG system couples two nonlinear parabolic PDEs within one moving forward and the other one moving backward in time. Second, there is a probability measure constraint on the population distribution of the agents. Third, the simultaneous recovery of two coupling factors within the MFG system is far from being trivial. We present two different techniques that can ensure the probability constraint as well as effectively tackle the inverse problems, which are respectively termed as high-order variation and successive linearisation. In particular, the high-order variation method is new to the literature, which demonstrates a novel concept to examine the inverse problems by positive inputs only. Both cases when the running cost depends on the measure locally and non-locally are investigated.