Gradient-based parameter calibration of an anisotropic interaction model for pedestrian dynamics

Abstract We propose an extension of the anisotropic interaction model which allows for collision avoidance in pairwise interactions by a rotation forces (Totzeck (2020) Kinet. Relat. Models 13 (6), 1219–1242.) including agents’ body size. The influence size on self-organisation agents channel and crossing scenarios as well fundamental diagram is studied. Since stated coupled system ordinary differential equations, we are able to give rigorous well-posedness analysis. Then state parameter calibration problem that involves data from real experiments. prove existence minimiser derive corresponding first-order optimality conditions. With help these conditions, gradient descent algorithm based mini-batches set. employ proposed fit strength parameters given results underpin feasibility method.