Dataset for: Stabilization–Responsiveness Trade-offs in Continuous Shared-Control for Invasive Brain–Computer Interfaces.

This dataset contains neural and behavioral recordings from two rhesus macaques performing intracortical brain–computer interface (BCI) navigation tasks in a virtual environment. Neural activity was recorded from three 96-channel Utah arrays implanted in ventral premotor cortex (PMv), dorsal premotor cortex (PMd), and primary motor cortex (M1), while animals controlled a virtual sphere using decoded neural activity in closed loop.

The dataset includes trial-level behavioral trajectories, decoded velocity commands, and neural spike-rate features used to train and evaluate a preferential subspace identification (PSID) neural decoder and a probabilistic shared-control controller (RT-V2). Experiments were conducted across multiple sessions and tasks designed to probe stabilization–responsiveness trade-offs in neural control, including Fixed Obstacle, Appearing Obstacle, and Respawn navigation tasks.

Each session contains time-aligned neural activity (50 ms bins), decoded velocity signals, executed trajectories, task event timestamps, and trial metadata. These data enable reproduction of the behavioral analyses and figures reported in the associated manuscript and provide a benchmark dataset for studying shared autonomy and adaptive control in high-bandwidth invasive BCIs.