Understanding User Adaptation to Network Impairments and Training: Evidence from Simulated VR Teleoperation

Network impairments such as delay, jitter, and packet loss remain a major challenge for streaming and remote-rendered Virtual Reality (VR) systems. While prior work primarily focuses on system-level mitigation, less is known about how users behaviorally adapt their control strategies under degraded network conditions. We investigate how users behaviorally adapt their control strategies in an immersive VR simulation of a teleoperation task under controlled communication perturbations. Thirteen participants performed the same pick-trace-place manipulation task across three training days under three impairment conditions-delay, jitter, and packet loss-each with six severity levels. From trial-level trajectory features, we infer latent control strategies using Gaussian mixture models and categorize them post hoc as Continuous, Move-wait, and Predictive. Our results reveal that impairment severity shifts strategy composition in condition-specific ways, while repeated practice consistently drives users toward Continuous control. Within packet-loss conditions, we further identify interpretable sub-strategies within the Predictive strategy and observe systematic reweighting across training days. These findings provide a behavior-centered account of user adaptation to network impairments in VR and inform strategy-aware interaction design.