Volume: 31 • Pages: 1-27

Compensating for sensorimotor noise and for temporal delays has been identified as major functions of the nervous system. However, the aspects have often been described separately in the frameworks of optimal cue combination or motor prediction during movement planning. But control theoretic models suggest that these two operations are performed simultaneously, and mounting evidence supports that motor commands are based on sensory predictions rather than sensory states. In this paper we study the benefit of state estimation for predictive sensorimotor control. More precisely, we combine explicit compensation for sensorimotor delays and optimal estimation derived in the context of Kalman filtering. We show, based on simulations of human-­‐inspired eye and arm movements, that filtering sensory predictions improve the stability margin of the system against prediction errors due to low-­‐dimensional predictions, or to errors in the delay estimate. These simulations also highlight that prediction errors qualitatively account for a broad variety of movement disorders typically associated with cerebellar dysfunctions. We suggest that adaptive filtering in cerebellum, instead of often-­‐assumed feed-­‐forward predictions, may achieve simple compensation for sensorimotor delays and support stable closed-­‐loop control of movements.