KJCKorea
Journal Central

In quantum computing, the output probability distribution may vary with each execution of the same circuit due to its sensitivity to noise. In this study, 2-qubit quantum gate combinations that are expected to produce identical output distributions under ideal conditions are designed as metamorphic test cases, and Depolarizing, Amplitude Damping, and Phase Damping noise models are applied to analyze the effect of each noise type on the output probability distribution. Experimental results show that in all noise models, when the noise probability reached 0.01, the p-value dropped below the significance level, causing the collapse of the defined metamorphic relation. It was also quantitatively confirmed through KL-Divergence and Cosine Similarity that the divergence from the expected output distribution accumulates with increasing noise strength. When applying metamorphic testing to quantum gate combinations, quantum gate testing can be performed by verifying the feasibility of metamorphic relation judgment and quantitatively tracking the process using the three proposed metrics. The results of this study can be applied to quantum computing reliability evaluation and noise-tolerant circuit design, and present the potential as a versatile quantum software testing methodology.

Large-scale services widely use distributed pipelines for real-time log processing. However, default buffering policies, although intended to protect system resources, can create bottlenecks that degrade real-time performance. This study constructed a large-scale distributed load environment and tracked end-to-end latency from log generation to final data warehouse loading at the millisecond level. Experiments were conducted across a five-stage optimization scenario by adjusting buffer sizes, wait times, and scan intervals. The results show that a low-latency configuration can reduce explicit buffering delay, but may increase packet-level overhead and degrade throughput. In contrast, the proposed hybrid configuration does not aim for absolute optimality in a single metric; instead, it applies cross-layer tuning to mitigate ingestion-layer traffic variability while minimizing downstream transmission delays. Under the evaluated conditions, the hybrid configuration achieved the lowest P95 latency, prevented pipeline collapse in micro-batch environments, and maintained throughput exceeding approximately 300 events per second.

The rapid development of autonomous driving has emphasized the role of swivel seats in maximizing spatial utility and passenger convenience, significantly raising the importance of developing advanced actuation systems. Conventional trapezoidal velocity profiles induce infinite jerk and mechanical shocks due to discontinuities in acceleration. Meanwhile, complex polynomial-based S-curve techniques impose a high computational burden on embedded systems. To overcome these limitations, this paper proposes a single cosine function-based velocity trajectory utilizing a time normalization technique that maps the physical time domain to a phase variable within the interval 0≤t≤π/2. The proposed method ensures the mathematical continuity of both velocity and acceleration, thereby constraining jerk to a finite value. Furthermore, by ensuring that acceleration starts from zero, it fundamentally eliminates initial operational impulses. In particular, because it is formulated as a single mathematical equation, we demonstrate that precise real-time control is achievable even in low-specification Micro Controller Unit (MCU) environments. By suppressing abrupt inertial changes, the proposed method is expected not only to provide passengers with superior ride comfort and psychological stability but also to yield industrial benefits, such as enhanced durability through the reduction of mechanical vibrations.