Trajectory intent prediction of autonomous systems using dynamic mode decomposition

Proliferation of autonomous systems have increased the threat space and the economic risk in several national infrastructures, e.g., at airports. Therefore, reliable detection of their intention is paramount to ensure smooth operation of national services and societal safety. This article reports a data-driven trajectory intent prediction algorithm which is based on a linear model structure of the autonomous system dynamics obtained from a dynamic mode decomposition algorithm. The model computation is enhanced by two sources of physics informed knowledge associated to the energy functional. Two different prediction algorithms that consider fixed or time-varying references are designed in terms of the availability of control input measurements. Rigorous theoretical results are provided to support the approach using matrix decomposition and optimization techniques. Simulation and experimental studies are carried out to verify the effectiveness of the proposal.