In the rapidly evolving landscape of cyber-physical systems (CPS), the demand for high-quality datasets to validate models and algorithms is more pressing than ever. Yet, the high costs, time, and expertise required to collect real-world data have long been significant barriers. Enter Yaa Acquaah, a researcher from the Department of Computer Science at North Carolina Agricultural & Technical State University, who has been tackling this challenge head-on. Her recent study, published in *Discover Applied Sciences* (translated as *Discover Practical Sciences*), explores the use of synthetic data generation to bridge this gap, with promising implications for industries like energy, healthcare, and smart cities.
Cyber-physical systems integrate physical and digital systems through networks of sensors, actuators, and controllers. These systems are critical in sectors such as healthcare, smart cities, transportation, energy management, and autonomous vehicles. However, the complexity and cost of collecting real-world data have limited the availability of genuine CPS datasets. Acquaah’s research leverages DoppelGANger, a Generative Adversarial Network (GAN)-based model, to generate synthetic CPS data using real datasets from four different domains: the Water Distribution Testbed (WDT), Hardware in the Loop Industrial Control System (HAI), Gas Pipeline, and Power System datasets.
The study compares synthetic datasets to real-world data through statistical analyses, visualization methods, and anomaly detection approaches. The results are intriguing. The difference in Silhouette scores across all CPS datasets was under 0.53, with the smallest difference of 0.071 in the HAI dataset and the largest difference of 0.525 in the Power System dataset. The Power System dataset also exhibited the smallest difference in validation Mean Absolute Error (MAE), with values of 0.0184 for the real data and 0.0196 for the synthetic data. However, correlation analysis revealed differences in feature relationships between real and synthetic datasets, and t-SNE visualizations showed inconsistent alignment for the HAI, Gas Pipeline, and Power System datasets as training epochs increased.
“While the synthetic data shows promise, there’s still room for improvement,” Acquaah notes. “The fidelity and reliability of synthetic datasets need to be enhanced to ensure they can be used confidently in critical applications.”
The energy sector, in particular, stands to benefit from these advancements. Power systems and gas pipelines are complex CPS that require robust datasets for testing and validation. Synthetic data generation could significantly reduce the time and cost associated with data collection, accelerating innovation and improving system reliability. “This research opens the door to more efficient and cost-effective ways of developing and validating CPS models,” Acquaah explains. “It’s a step toward making these systems more resilient and adaptable.”
As the field of CPS continues to evolve, the need for high-quality, reliable datasets will only grow. Acquaah’s work highlights the potential of synthetic data generation to meet this need, paving the way for more advanced and reliable CPS models. The study, published in *Discover Practical Sciences*, offers a glimpse into a future where synthetic data plays a pivotal role in shaping the next generation of cyber-physical systems.