Jun Kobayashi

In simulation, under reduced accuracy or temporally correlated errors, recalibration or fixed-weight learning captured most gains; a disagreement-dependent rule beat fixed weights mainly under occasional large errors (4.05% lower MSE).

This paper develops a mathematical model to explain how sensory reliability affects postural control during quiet standing, showing that precision parameters determine sensory reweighting and stability.

The paper shows that adapting only the readout of a frozen continuous-time neural network can reproduce force-field adaptation and after-effects in a simulated motor control task.

This study evaluates fixed controllers for musculoskeletal arm posture control and finds none can achieve stable near-task holds, suggesting the need for learned control mechanisms.

The paper investigates how a predictive state observer adapts its sensory prediction-error correction gain in a muscle-driven reaching task using available signals, finding delay-dependent optimal gains and limitations of adaptive correction.