Multi-hypothesis GPS and electronic fence data fusion for safety-critical positioning in railway worksites

Safety-critical applications often use position information as a mean of assessing the safety level of people. For this reason, such information is required to be precise in terms of accuracy and timeliness. This paper regards position mechanisms for personalized warning systems for railway workers. Position accuracy for safety assessment purposes is defined as the precise identification whether the worker is located in a dangerous or safe zone within a certain worksite. This paper extends a previous publication from the same authors to a scenario with multiple workers, while analyzing the combination of wearable GPS receivers and electronic fences strategically placed at the worksite. The proposed data fusion algorithm comprises a Kalman Filter (KF) for filtering GPS observations and a Hidden Markov Model (HMM) for fusion with fence data. A Multiple-Hypothesis Tracking (MHT) mechanism is used to handle multiple workers within the worksite as a mean to compensate the inability of the fence to distinguish the workers. The proposed solution is analyzed under experimental setups. The obtained results outperformed a GPS-only solution and the previously proposed solution by reducing or even removing false alarm and safety-related missed detection events.