Applying neural fields to the stability problem of an inverted pendulum as a simple biped walking model

This paper proposes a control architecture based on neural fields for a relatively complex and unstable dynamical system. The neural field model is capable of addressing goal-based planning problems and has properties, like embedding in an Euclidean space and linear stability, that potentially make it well-fitted for dynamic control tasks. The neural field control architecture is tested with the inverted pendulum problem. The cart-and-pole inverted pendulum is used as a simple biped walking model, where the cart models the center of pressure and the pole models the center of mass. The parameterized (i.e. non-evolved) neural field control architecture is compared against an evolved recurrent neural field controller applied to the same control task. The non-evolved neural field controller performs, in the simulation, better than the evolved recurrent neural network controller. Furthermore, the neural field has a spatial representation which allows an easy visualization of its field potentials.