Schematic Model.

a. Diagram of an adaptive state-feedback control model of the Central Nervous System (CNS). The controller (top, blue) is decomposed into (bottom): a state estimator (Est.), which estimates the current state, an adaptive estimator (Adapt.) which evaluates the model parameters (here the defining coefficient of the force field), and a parametric controller (Control) mapping estimated state into motor commands (see text for details). b. Trial-by-trial adaptation model where the estimated parameter changes offline between two trials. Black lines indicate the estimated parameter over time during the trials indexed by k. Observe the estimate of the force field parameter is bounded by our experimental results (red line), since anticipatory compensation for the force field was partial. c. Trial-by-trial (offline) and online adaptation schemes, here the estimated force field parameter is allowed to change within a trial, possibly beyond the estimate used prior to movement onset.