Theoretical foundations for layered architectures and speed-accuracy tradeoffs in sensorimotor control

Abstract

Nervous systems sense, communicate, compute, and actuate movement using distributed components with trade-offs in speed, accuracy, sparsity, noise, and saturation. Nevertheless, the resulting control can achieve remarkably fast, accurate, and robust performance due to a highly effective layered control architecture. However, there is no theory explaining the effectiveness of layered control architectures that connects speed-accuracy trade-offs (SATs) in neurophysiology to the resulting SATs in sensorimotor control. In this paper, we introduce a theoretical framework that provides a synthetic perspective to explain why there exists extreme diversity across layers and within levels. This framework characterizes how the sensorimotor control SATs are constrained by the hardware SATs of neurons communicating with spikes and their sensory and muscle endpoints, in both stochastic and deterministic models. The theoretical predictions of the model are experimentally confirmed using driving experiments in which the time delays and accuracy of the control input from the wheel are varied. These results show that the appropriate diversity in the properties of neurons and muscles across layers and within levels help create systems that are both fast and accurate despite being built from components that are individually slow or inaccurate. This novel concept, which we call "diversity-enabled sweet spots" (DESSs), explains the ubiquity of heterogeneity in the sizes of axons within a nerve as well the resulting superior performance of sensorimotor control.

The unreasonable effectiveness of deep learning in artificial intelligence

“…Brains have additional constraints due to the limited bandwidth of sensory and motor nerves, but these can be overcome in layered control systems with components having a diversity of speed-accuracy tradeoffs (31). A similar diversity is also present in engineered systems, allowing fast and accurate control despite having imperfect components (32).…”

Experimental and educational platforms for studying architecture and tradeoffs in human sensorimotor control

“…The impact of the SATs in nerve signaling on the SATs in sensorimotor control performance has been characterized in our previous paper [17] and is reviewed in our companion paper [14]. In this section, we validate the theoretical results using a mountain-bike game in the experimental platform.…”

“…A companion paper [14] develops a theoretical framework to connect the system-level and compoenet-level SATs. The results suggest theoretically optimal layering creates "diversity sweet spots (DSSs) to effectively layer sensorimotor control with appropriate diversity in neurons/muscles to achieve systems that are both fast and accurate despite being built from components that individually are not.…”

“…1). The details for the theory is reviewed in a companion paper [14]. Because it is impossible to noninvasively manipulate the internal delay and data rate in neural signalling, we find an alternative to add the external delay and quantization in actuator which generates an identical effect to simulate the hardware SATs in nerves.…”

Antithetic integral feedback for the robust control of monostable and oscillatory biomolecular circuits

“…Feedback regulation is ubiquitous in biology, playing a crucial role in high-level phenomena, such as sensory perception in animals (Wiener (1948); Nakahira et al (2019)), all the way down to the most basic processes in life, like the regulation of amino acid biosynthesis (Monod (1971)). As we refined our understanding of molecular biology, it became abundantly clear that life not only relies heavily of feedback control, but that this control is often extremely precise and robust (Barkai and Leibler (1997)).…”