A Winner-Takes-All Mechanism for Event Generation

TL;DR

This work addresses the need for robust, adaptable central pattern generators by introducing a rebound winner-takes-all mechanism that combines intrinsic rebound dynamics with competitive inhibition. The authors develop a ring-structured oscillator framework, demonstrate endogenous rhythmic generation, and show how weak external inputs can entrain phase while global inputs modulate frequency. The approach preserves the simplicity of traditional half-center oscillators while enabling richer control over phase and timing, suggesting strong potential for hardware implementations in neuromorphic robotics. The study lays the groundwork for quantitative analysis, scaling to larger networks, and practical hardware validation in future work.

Abstract

We present a novel framework for central pattern generator design that leverages the intrinsic rebound excitability of neurons in combination with winner-takes-all computation. Our approach unifies decision-making and rhythmic pattern generation within a simple yet powerful network architecture that employs all-to-all inhibitory connections enhanced by designable excitatory interactions. This design offers significant advantages regarding ease of implementation, adaptability, and robustness. We demonstrate its efficacy through a ring oscillator model, which exhibits adaptive phase and frequency modulation, making the framework particularly promising for applications in neuromorphic systems and robotics.

Figures (5)

• Figure 1: Rebound spiking in HH model.
• Figure 2: HCO. The parameters for inhibitory synapses are [$g_\text{syn}=10,\tau=1,V_\text{th}=-65,\alpha=1.5$].
• Figure 3: Representations of an all-to-all inhibitory network
• Figure 4: On the left is the notation for the rebound winner-takes-all ring oscillator. On the right are two raster plots of membrane voltages for two ring oscillators, each composed of different types of neurons. For the ring of Hodgkin-Huxley neurons the parameters for inhibitory synapses are [$g_\text{syn}=-10,\tau=1,V_\text{th}=-65,\alpha=1.5$] and the parameters for excitatory synapses [$g_\text{syn}=0.5,\tau=5,V_\text{th}=-65,\alpha=1.5$]. For the ring of Ribar-Sepulchre neurons the parameters for inhibitory synapses are [$g_\text{syn}=-5,\tau=0.1,V_\text{th}=-4,\alpha=2$] and the parameters for excitatory synapses [$g_\text{syn}=0.3,\tau=60,V_\text{th}=-4,\alpha=2$].
• Figure 5: Synchronizing a ring oscillator composed of 5 Hodgkin–Huxley neurons to an external rhythmic input. Parameters for inhibitory synapses [$g_\text{syn}=-15,\tau=0.1,V_\text{th}=-65,\alpha=1.5$]. Parameters for excitatory synapses [$g_\text{syn}=10,\tau=0.1,V_\text{th}=10,\alpha=1.5$]. a) Endogenous rhythm of the ring. b) The ring being entrained by the external input. c) Adaptive controller is applied together with the external rhythmic input.

Theorems & Definitions (1)

• Definition 1: Hodgkin-Huxley rebound neuron