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

Shared AI-brain control nearly eliminates failures in intracortical BCI navigation during dynamic tasks, reducing obstacle collisions and overshoot in macaques.

• Why it matters: Reliable BCI operation in real-world, dynamic environments is limited by neural command fluctuations, hindering safe and effective neuroprosthetic use for people with paralysis.
• What they did: A confidence-modulated AI-brain shared-control system was developed, integrating decoded neural commands with a probabilistic temporal prior, tested in two macaques performing virtual navigation.
• The result: The approach significantly improved stability and safety, with offline analysis revealing that algorithmic adjustments can further enhance responsiveness, guiding safer neuroprosthetic designs.