However, the stability and robustness of immersive virtual reality interaction remain precarious under realistic conditions where tracking noise, occlusion, and user movement variability can cause unintended activations and instability. This paper presents an deployable design for immersive human-computer interaction (HCI), where the concept of interaction is viewed as a closed-loop process consisting of probabilistic intent estimation, stability-gated commitment, constraint execution, and comfort-gated multimodal feedback. The design extends a unified estimator for targeted selection, direct manipulation, and spatial user interface operation by delaying irreversible commitment until a hysteresis condition is met for improved stability without causing noticeable lag. The design also includes a comfort risk estimator for adapting commitment and feedback thresholds for optimal comfort without sacrificing usability. The study employed a controlled within-subjects design for motion-stressed selection, constraint execution, and attention-switching spatial user interface operation, measuring objective performance, errors, system stability, and subjective measures of workload and discomfort. The results showed a significant decrease in unintended commitments and correction costs, improved task success rates with manipulation constraints, and a decrease in discomfort growth rates for spatial user interface operation with high motion variability.
