Train Localisation using Wireless Sensor Networks
Safety and reliability have always been concerns for railway transportation. Knowing the exact location of a train enables the railway system to react to an unusual situation for the safety of human lives and properties. Generally, the accuracy of localisation systems is related with their deployment and maintenance costs, which can be on the order of millions of dollars a year. Despite a lot of research efforts, existing localisation systems based on different technologies are still limited because most of them either require expensive infrastructure (ultrasound and laser), have high database maintenance, computational costs or accumulate errors (vision), offer limited coverage (GPS-dark regions, Wi-Fi, RFID) or provide low accuracy (audible sound). On the other hand, wireless sensor networks (WSNs) offer the potential for a cheap, reliable and accurate solutions for the train localisation system. This thesis proposes a WSN-based train localisation system, in which train location is estimated based on the information gathered through the communication between the anchor sensors deployed along the track and the gateway sensor installed on the train, such as anchor sensors' geographic coordinates and the Received Signal Strength Indicator (RSSI). In the proposed system, timely anchor-gateway communication implies accurate localisation. How to guarantee effective communication between anchor sensors along the track and the gateway sensor on the train is a challenging problem for WSN-based train localisation. I propose a beacon driven sensors wake-up scheme (BWS) to address this problem. BWS allows each anchor sensor to run an asynchronous duty-cycling protocol to conserve energy and establishes an upper bound on the sleep time in one duty cycle to guarantee their timely wake-up once a train approaches. Simulation results show that the BWS scheme can timely wake up the anchor sensors at a very low energy consumption cost. To design an accurate scheme for train localisation, I conducted on-site experiments in an open field, a railway station and a tunnel, and the results show that RSSI can be used as an estimator for train localisation and its applicability increases with the incorporation of another type of data such as location information of anchor sensors. By combining the advantages of RSSI-based distance estimation and Particle Filtering techniques, I designed a Particle-Filter-based train localisation scheme and propose a novel Weighted RSSI Likelihood Function (WRLF) for particle update. The proposed localisation scheme is evaluated through extensive simulations using the data obtained from the on-site measurements. Simulation results demonstrate that the proposed scheme can achieve significant accuracy, where average localisation error stays under 30 cm at the train speed of 40 m=s, 40% anchor sensors failure rate and sparse deployment. In addition, the proposed train localisation scheme is robust to changes in train speed, the deployment density and reliability of anchor sensors. Anchor sensors are prone to hardware and software deterioration such as battery outage and dislocation. Therefore, in order to reduce the negative impacts of these problems, I designed a novel Consensus-based Anchor sensor Management Scheme (CAMS), in which each anchor sensor performs a self-diagnostics and reports the detected faults in the neighbourhood. CAMS can assist the gateway sensor to exclude the input from the faulty anchor sensors. In CAMS, anchor sensors update each other about their opinions on other neighbours and develops consensus to mark faulty sensors. In addition, CAMS also reports the system information such as signal path loss ratio and allows anchor sensors to re-calibrate and verify their geographic coordinates. CAMS is evaluated through extensive simulations based on real data collected from field experiments. This evaluation also incorporated the simulated node failure model in simulations. Though there are no existing WSN-based train localisation systems available to directly compare our results with, the proposed schemes are evaluated with real datasets, theoretical models and existing work wherever it was possible. Overall, the WSN-based train localisation system enables the use of RSSI, with combination of location coordinates of anchor sensors, as location estimator. Due to low cost of sensor devices, the cost of overall system remains low. Further, with duty-cycling operation, energy of the sensor nodes and system is conserved.
Advisor: Huang, Zhiyi; Zhang, Haibo; Deng, Jeremiah D.
Degree Name: Doctor of Philosophy
Degree Discipline: Department of Computer Sciences
Publisher: University of Otago
Keywords: Localisation; WSN; Trains; Particle Filtering
Research Type: Thesis