In this chapter, we present the design and development of a comprehensive smartphone sensor-based HAR dataset, collected explicitly in an uncontrolled environment, to improve real-world applicability. Leveraging the omnipresence and sensing capabilities of modern smartphones, we capture raw sensor data from accelerometers and gyroscopes embedded in commercially available devices. The dataset consists of diverse activities performed naturally by 20 participants across varied conditions, addressing limitations observed in existing datasets that predominantly rely on controlled settings. We detail the data acquisition methodology, including sensor placement, sampling rates, and activity taxonomy, ensuring a robust representation of real-life movements. Additionally, we highlight the challenges of uncontrolled data collection, such as class imbalance and variability due to differences in participant physiology and behaviours. To facilitate reproducibility and future research, we provide an overview of the dataset structure, annotation process, and pre-processing techniques aimed at enhancing data quality. This dataset serves as a valuable resource for developing adaptive, privacy-preserving, and generalizable HAR models, particularly in healthcare and assistive applications.