2026-09-22T12:46:59Zhttps://api.repository.cam.ac.uk/server/oai/request

oai:www.repository.cam.ac.uk:1810/3812422025-03-14T14:34:00Zcom_1810_213729com_1810_256065col_1810_219485

Control of neuronal circuits: from biology to robotics https://doi.org/10.17863/CAM.116524 Schmetterling, Raphael 0000000267587219 Sepulchre, Rodolphe Rhythmic robotic controllers often take inspiration from neuroscience. Due to the complexity of biophysical neurons, neuro-inspired controllers are frequently founded on abstract models such as non-linear oscillators. Such controllers retain some of the desirable properties of neuronal circuits, but lack the physical embodiment that is characteristic of biological systems and that is key to their effectiveness. In this thesis we take a step towards reconciling biophysics with bio-inspired control, using the language of control theory to do so. The proposed controller is event-based, and has the physical realisation of an analogue neuromorphic circuit. The event-based behaviour is rooted in the presence of mixed feedback. The behaviour is regulated using output feedback and also an adaptive control that tunes the gains of the positive and negative feedback loops. Adaptive control is aligned with neuromodulation, which is central to the adaptation and robustness of animal nervous systems. We illustrate the potential of the event-based neuromorphic approach on the simple mechanical model of a pendulum. In addition to the pendulum controller, we also propose a methodology for the control of biological or neuromorphic neuronal circuits. In particular, we explore the classical paradigm of indirect adaptive control to design neuromodulatory controllers in biophysical neuronal models. This provides a methodology that aligns with impedance control in robotics. The method relies on parameter estimates obtained with a recently-proposed adaptive observer that implements a centralized recursive least squares algorithm. Inspired by biology, we show that decentralization and redundancy help recover the performance of this algorithm in the presence of uncertainty and mismatch on the internal dynamics of the model. University of Cambridge 2024-10-05 Thesis Doctoral Doctor of Philosophy (PhD) eng https://www.repository.cam.ac.uk/handle/1810/381242 https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/070ae320-941a-4201-869c-3ce11bed00e2/download 47fa3c2c1809bb675fd062eb89985549 https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/62577906-96e4-41e6-b141-846a0c42837b/download 87eda9de84448d1f82354d60eee3eb5f https://creativecommons.org/licenses/by-sa/4.0/ adaptive control bio-inspired robotics central pattern generators event-based control neuromodulation neuromorphics neuroscience robotics