When
Thematic Session 3: Modeling and Analysis (Tuesday, 09:40)
Abstract
Dynamic Attending Theory posits that we entrain to time-structured events in a similar way to synchronizing oscillators. Much research in this field attempts to identify oscillating circuits in the brain that may be involved in human beat perception. Oscillating neural circuits in vertebrates, however, are well understood and well adapted for rhythmic behaviour. These circuits, known as central pattern generators (CPGs), organize movement patterns between limbs during locomotion, and are able to adapt to a wide range of tasks, environments and perturbations. Here, we present virtual quadruped robots with CPGs evolved to spontaneously entrain to a rhythmic stimuli over a range of frequencies. In contrast to most current robotic approaches, the gait frequency is an input-dependent emergent property of the network, due to a heavily bio-inspired neuron model and network design. Through self-organization, the CPG network is thus able to successfully entrain to rhythms, with feedback only required for stabilization of the body. Using multi-objective evolution, we are able to automatically generate a range of controllers for the same hardware, emphasising different combinations of abilities. We study the entrainment and beat-tracking ability of these agents for real musical stimulus as a function of properties such as pulse clarity, shedding light on how the task of beat tracking may be distributed between the body and brain, and discuss the extension of our approach to multi-agent systems.
Bio
Alex Szorkovszky is an applied mathematician whose main research interests are collective behaviour, complex systems, cultural evolution and artificial life, using the tools of dynamical systems, statistical and agent-based modelling. He is currently undertaking a Marie Sk?odowska-Curie Actions fellowship (2021-2023) in which he is developing adaptive robotic agents in order to uncover potential mechanisms behind entrainment in humans.