A model of neural population dynamics for flexible sensorimotor control

Modern large-scale recordings have revealed that motor cortex activity during reaching follows low-dimensional dynamics, thought to reflect sensorimotor computations underlying muscle activation. However, the origin of these low dimensional patterns, and how they flexibly reorganize across different tasks remain unclear. Here we demonstrate that the key features of neural activity can naturally emerge in a linear model combining a random network with a biomechanical system. Remarkably, this model shows explicitly how a fixed network can achieve flexible control of multiple behaviours through optimal mapping of sensory feedback onto the network. Finally, analytical decomposition of the controller reveals that low-dimensional network dynamics follow directly from the propagation of low-dimensional feedback signals from the biomechanical plant through the network. This formalism provides a computational framework to interpret flexible motor control in the nervous system, which directly links neural population dynamics to sensorimotor behavior.