Mobile Devices' Low Charge, High CPU Load, Number of Sensors, and OS Features Delay Sensor Data Readings

Sensor data reliability plays a critical role in applications that collect and use this data. The types of effects that can invalidate the data could be physical data drift, bias, missing data or outliers, to name a few. These can arise from several factors, where power management and operating systems, that are built to conserve energy and optimize application usage, are of the most concern. Thus, this study assessed an impact of the battery charge, functional state of the device, and the number of sensors employed. We used variability in the timestamps of the sensor data readings to measure the quality of the data acquired. Our experiments revealed that lower battery charge and high CPU loads have significantly impacted the mean interval between the sensor data readings. However while this happened at a very small scale (1-2ms) it had different impact for different devices. We also noticed that the number of sensors and device movement significantly changed the variability but not consistently across devices. Our most impactful finding is that sensor measurements can be significantly delayed, for up to 60ms. These large gaps in sensor readings can cause aliasing distortions affecting mobile apps. Moreover, our experiments show that the timing of sensor data can be inconsistent and highly dependent on the device's conditions. Detecting and compensating for sensor data loss and resampling the data, though at the cost of accuracy, may be a potential solution.