Intermittent movement control emerges from information-based planning

Abstract

Mammalian motor control is inherently discrete, with movement corrections occurring at rates determined by task demands and the quality of sensory information. While several models have been proposed to explain this discreteness, it remains unclear when a new movement should be initiated and how long it should last. To address this gap, we introduce the Information Predictive Control (IPC) framework, which combines model predictive control with information theory. IPC triggers corrections only when unexpected deviations arise and when corrective actions are likely to succeed. By quantifying "surprise" relative to predicted internal and external states, IPC generates successful movements while robustly integrating sensorimotor noise, task constraints, and target variability. Simulations show that IPC reproduces human-like behavior in discrete reaching, continuous target tracking, and adaptive planning under uncertainty, while dynamically adjusting the planning horizon in complex, unpredictable environments.