Answers are grounded in this paper and cite their sources.
This analysis estimates the optimal number of spiking neurons for linear quadratic regulator control of cartpole systems, suggesting foundational methods for neuromorphic computing.
The emerging field of neuromorphic computing for edge control applications poses the need to quantitatively estimate and limit the number of spiking neurons, to reduce network complexity and optimize the number of neurons per core in an application-specific neuromorphic hardware. The near-linear behaviour of a Leaky-Integrate-and-Fire neuron can be exploited to achieve the Linear Quadratic Regulator (LQR) control of a cartpole system. This has been shown in simulation, followed by a demonstration on a single neuron hardware, known as Lu.i. The optimal number of neurons required for the population encoding of a signal for such control is quantitatively estimated through Fourier analysis in order to improve the precision of signal representation, and then such a population of neurons has been simulated in the Nengo Neural Engineering Framework (NEF) to show the control performance on a cartpole across varying bob mass, length and number of links. This work lays the foundation of non-heuristic approaches to estimate the optimal spiking neural network complexity in an NEF, based on the target application, and enables further research on quantifying neural network topologies.
No takes yet. Share an insight, caveat, or question.
Banerjee et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: