The potential of digital product twins is determined significantly by the availability and the usability of product data from both, the digital and the physical product twin. Expanding the data availability and usability could enable better analysis and optimization of physical entities in various scenarios, enabling new business models based on accurate predictions. Data stored in a digital twin at a reduced resolution can be approximated using numerical or machine learning approaches. This paper investigates the application of AI models to expand the availability and usability of digital twin data, allowing more precise and resilient predictions. Using the example of a remote-controlled miniature racing car, the main objective is to develop a robust virtual and AI-driven model that can accurately estimate the real-time displacement of the physical miniature car on a racetrack. This is achieved by estimating car positions between discrete measuring points transmitted by the track’s onboard systems using linear regression, cubic regression, and an LSTM neural network. The estimated results are then compared to each other and validated using original data.