Rapid clock synchronisation for ubiquitous sensing services involving multiple smartphones

This paper investigates the precision of rapid clock synchronisation for ubiquitous sensing services which consist of multiple smartphones. Specifically, we consider scenarios where multiple smartphones are used to sense physical phenomena, and subsequently the sensor data from multiple distributed devices is aggregated. We observe that the accumulated clock drift for smartphones can be more than 150ms per day in the worst case. We show that solutions using the public Network Time Protocol (NTP) can be noisy with errors up to 1800ms in one request. We describe a rapid clock synchronisation technique that reduces drift to 10ms on average (measured by linear regression) and achieves pair-wise synchronisation between smartphones with an average of 27ms (measured by accelerometer), following a Gaussian-like distribution. Our results provide a lower bound for rapid clock synchronisation as a guide when developing ubiquitous sensing services using multiple smartphones.