Fig. 9: Alternative to energy minimization: Sensory recalibration via proprioceptive realignment.

a As an alternative to energy minimization, we considered sensory recalibration via proprioceptive realignment as a hypothesis for adaptation. This recalibration realigns the proprioception from the two legs to conform with the expectation that both legs usually are on the same surface with a common speed, rather than surfaces with different speeds. The proprioceptive recalibrator takes in the sensory information from proprioceptors of the two legs and vision and computes a recalibrated version of the proprioceptive information. b The recalibrated sensory information is used by the stabilizing controller, so the direct feedback controller of Fig. 1 is replaced by a more general sensorimotor transformation. c Proprioceptive correction for each leg shown as a function of time for a particular split-belt protocol. This correction is subtracted from the initial proprioceptive estimate to obtain the recalibrated proprioception. The correction for the fast leg is positive and that for the slow leg is negative. 100% recalibration corresponds to completed proprioceptive realignment and 50% recalibration is close to that observed in experiment⁴². d Using the recalibrated proprioception as feedback in the stabilizing controller results in the reduction of step length asymmetry, but even 100% recalibration does not result in symmetry or positive step length asymmetry. Thus, the model predicts that proprioceptive realignment cannot be fully responsible for split-belt adaptation. e Proprioceptive realignment also reduces energy coincidentally even without energy being an explicit objective in this situation. Light blue shaded region in panels c-e is the period of split-belt adaption.