Dynamical Systems Theory of Embodied Perceptual Decision-Making

Abstract: Decision-making is a fundamental building block of behavior, both in natural settings such as animal movement and in engineered settings such as autonomous robotics. In both of these contexts, decisions are often made on the move through an interplay of continuously changing sensory information and goal-oriented action selection. In this talk, we present a dynamical systems framework for understanding how discrete choices emerge through feedback between neural activity and movement through space. Starting from a stochastic ring-attractor model motivated by biological neural circuits involved in perceptual evidence integration, we will show how reduced representations such as competing neural accumulators and local drift-diffusion descriptions of choice arise at different scales of the same process. Bifurcation analysis of the reduced decision dynamics then reveals how movement reshapes competition among alternatives, and how asymmetries in target geometry and quality bias transitions from compromise to commitment captured through movement. These results link canonical models of perceptual choice and reveal how environmental geometry shapes embodied decisions. We will illustrate how insights from this biologically inspired modeling and analysis translate to novel design principles for autonomous behavior in mobile robots.

astrophysicscomputational biologycomputational engineering, finance, and sciencegeneral mathematicsnonlinear sciencescomputational physicsfluid dynamicsgeneral physics

Audience: researchers in the topic

Northwestern Applied Mathematics Seminar