Seminar Details
2026-04-07 (13:00) : Dealing with Variability and Mismatch in Closed‑Loop Neuromorphic Control: From Limitations to Design Opportunities
At Shannon room
Organized by Institute of Information and Communication Technologies, Electronics and Applied Mathematics
Speaker :
Alessio Franci (Liège)
Abstract :
Neural systems exhibit substantial variability in their underlying parameters, yet they reliably generate robust and adaptive closed-loop behaviors in sensorimotor tasks. Neuromorphic systems display an analogous form of variability, primarily due to transistor mismatch, raising the question of how such mismatch can be exploited rather than suppressed. Understanding how to harness this inherent device-level variability to design robust and adaptive neuromorphic controllers is an active and promising research direction.
In this talk, I will first introduce the mechanisms and functional roles of variability in biological neural systems and relate them to parameter mismatch in neuromorphic hardware, with a particular focus on closed-loop control scenarios. Building on this background, I will present two recent lines of work: (i) variability in human gait control, and (ii) a quantitative analysis of mismatch effects in fully analog neuromorphic neurons. I will conclude with an overview of emerging neuro-inspired strategies leveraging transistor mismatch to achieve robustness and adaptability in neuromorphic control systems.