Confidence-modulated shared control for reliable intracortical BCI navigation in dynamic environments

Brain-computer interfaces enable people with paralysis to control neuroprosthetic devices, but reliable operation in dynamic environments remains limited by noise in decoded neural commands. We developed a confidence-modulated probabilistic shared-control framework that adaptively integrates the decoded neural commands with a probabilistic temporal prior to stabilize execution while preserving the directional structure of the decoded command. In two male macaques performing closed-loop tasks in complex environments, shared control nearly eliminated execution-level failures, while maintaining the directional structure of neural commands. Abrupt target changes revealed a limitation: temporal-history dependence transiently slowed responsiveness when recent movement history became inconsistent with the updated command, although this delay was not obligatory. Resetting the temporal prior removed the residual shared-control delay and restored performance, demonstrating a specific contribution of temporal history to the impairment. Here, we show that shared control improves the reliability of continuous intracortical brain-computer interfaces while revealing a trade-off between stabilization and responsiveness.