SubTrack: Enabling Real-Time Tracking of Subway Riding on Mobile Devices
Summary (6 min read)
Introduction
- Traditional approaches using timetables need continuous attentions from the subway riders and are limited to the poor accuracy of estimating the travel time.
- Recent approaches using mobile devices rely on GSM and WiFi, which are not always available underground.
- As a result, people could easily miss their stops unless they are fully attentive during their subway rides.
- Based on the fact of the prevalence usage of mobile phones these days, the authors seek a solution that can enable real-time tracking of subway riding on mobile devices, providing intelligent information of the passenger’s location underground and predicting the arrival time.
- In order to accurately track the motion states of the subway train, including both velocity and position, the authors need to calibrate the accumulated velocity and distance which have large drift errors due to biased motion sensor readings on mobile phones.
II. CHALLENGES & SYSTEM OVERVIEW
- Smartphone-based localization has been widely studied in many different scenarios [4, 8].
- But as far as the authors know, few work has addressed the problem of localization in underground public transportation systems, where GPS signal and wireless infrastructure are not always available.
- Existing work mainly relies on the built-in inertial sensors to allow smartphones to determine their location substantially [4, 5].
- But some strong assumptions of these approaches, such as fixed smartphone postures and reliable sensor readings, prevent them from practical use.
- The authors first point out several key challenges of the system design, and then introduce the proposed system work flow.
A. Challenges
- To ensure real-time and accurate subway rider tracking, the following key challenges should be dealt with in the system.
- Particularly, if the smartphone is not placed in parallel with the moving direction of the train, it is difficult to correctly derive the motion state of subway train based on the raw sensor readings, not even to mention the smartphone of changing postures.
- It is intuitive to apply existing GPS-based and wireless infrastructure-based localization methods to underground transportation system.
- Due to the inherent flicker noise in the electronics and in other components susceptible to random flickering [6], inertial sensors, including both accelerometer and gyroscope, will produce biased measurements over time, which lead to inaccurate velocity estimation results.
- It is not realistic to remove such bias inside the sensor themselves.
B. System Overview
- Given the above challenges, the authors propose a real time fully-automatic subway riding tracking system, SubTrack.
- It aims to track the motion states of the subway train, including both velocity and position, through accumulated accelerometer readings over time, and then predict the arrival time to next train stop given the current motion state and subway map.
- In Passenger Motion State Detection & Data Reconstruction, the smartphone of a particular passenger periodically scan nearby cell-tower IDs (Cell-IDs) until it finds one matched record in the Subway Station Inner Cell-ID Database, which is built based on historical Cell-ID collections from all subway stations.
- Inside the core component Train Motion State & Arrival Time Estimation, SubTrack first derives the real-time velocity of subway train through accumulating the accelerometer readings over time.
- Integrating both the stationary reference points and specific historical velocity from Train State Estimation Database, the velocity drift error will be mitigated based on the linear model between the drift error and time.
III. TRAIN MOTION STATE TRACKING AND ARRIVAL TIME PREDICATION
- The authors focus on detecting whether the train stops, tracking train’s position and predicting arrival time using a mobile device (e.g., smartphone).
- In order to support the above modules, there are many building blocks including Passenger Station Entrance Detection, Train Departure Detection and Coordinate Alignment which are discussed in Section IV and Section V.
A. Train Stopping Detection
- Train stopping detection aims to detect when the train has stopped and how long it stays at the train stops (i.e., stopping periods).
- It is easy to find that the STE of the accelerometer readings on z-axis is much lower when the train has stopped.
- Lk is the number of sensor samples between the kth pair of reference points.
- Based on the velocity drift gradient ∆vel, the calibrated velocity velcali(t) that is drift error-free could be obtained in real time: velcali(t) = vel(t)− driftA − (t− tA)×∆vel, (3) where vel(t) is the estimated velocity at time t, driftA is the velocity drift at reference point A (i.e., the previous reference point before t) at time tA.
- Figure 4 demonstrates an example of the velocity estimation for five-stop subway ride with both their proposed real-time velocity estimation method (i.e., online) and the method in [7] (i.e., off-line).
C. Traveling Distance Estimation
- Given the calibrated velocities, the authors could derive the current position of the train through the traveling distance estimation and thereby predict the arrival time to next train stop.
- Specifically, the traveling distance of a subway train can be obtained based on the integration on the calibrated velocity over time.
- (4) Figure 5 gives an example of the estimated traveling distance over 5 train stops.
- Note that there is no accumulated error in their distance estimation since the distance of each stop-to-stop segment is estimated independently.
D. Arrival Time Prediction
- The authors next perform arrival time predication to the next train stop.
- Accordingly, the proposed system will keep updating the arrival time to the final destination in real-time, and notify the riders to prepare to get off near arrival.
- In their empirical study in the two cities of China and USA, the authors observe that the train usually experiences three motion phases between two adjacent train stops as shown in Figure 6: acceleration (s = 1), approximate uniform motion (s = 0) and deceleration (s = −1).
- According to their experimental observations, the authors assume that the absolute accelerometer readings during acceleration and deceleration are a constant value, acc. velcali(t) and dis(t) are the calibrated velocity and distance, veluni is the uniform velocity which can be measured when s = 0 or obtained from historical velocity when s =.
- The authors can further predict the arrival time to the final destination by integrating historical remaining time information to the rest of the stations.
IV. DETECTION OF PASSENGER MOTION STATE BEFORE BOARDING
- The authors focus on the passenger’s motion state detection.
- Specifically, two main topics are included: 1) Passenger’s station entrance detection.
- It aims to determine when the passenger enters the subway station and which station it is; 2) Train departure detection.
- The authors will detect whether the train departs by differentiating the accelerometer readings caused by the moving train from those corresponding to the walking passenger.
A. Passenger Station Entrance Detection
- In the proposed SubTrack system, it is critical to pinpoint the departure time and departure station as the starting point of the trip via subway.
- The authors can find that when the passengers are at the station entrance yet not getting in, their smartphones could associate with multiple Cell-IDs, and switch among some of them frequently.
- Note that each color corresponds to one particular subway station.
- The authors could infer that those Cell-IDs must be specifically deployed inside the stations.
- Inspired by the above observations, the authors can determine whether the passenger enters a subway station by periodically examining the duration (i.e., λ seconds) that his/her smartphone has been in association with a unique Cell-ID inside the subway station.
B. Train Departure Detection
- In order to pinpoint the departure time of the train, the authors first need to determine whether the train that the passengers board is on the go.
- Therefore, it could help to determine when the train departs by differentiating the accelerometer readings.
- Figure 8 shows the acceleration magnitude of the acceleration readings after gravity removal before and during the subway train moves.
- Figure 9 depicts the average and variance of accelerometer readings in sliding windows.
- The authors can find that when the passenger walks, both the average and variance of acceleration maintain at high level.
V. COORDINATE ALIGNMENT
- Coordinate alignment works after the passenger boards the train.
- The authors first sense and align the current posture of the smartphone by using quaternion [12], which is obtained from the API of CMAttitude [13].
- Figure 10(c) plots the gravity-aligned acceleration (i.e., â1) while the train passes two adjacent stations, in which the acceleration on z axis is aligned with zt axis with gravity filtered.
- Utilizing the variant trigonometric function described in [14], the authors can derive the deviation angle θ that the acceleration in â1x&â1y relative to the known acceleration that is parallel with yt.
A. Experimental Methodology
- The experiments are carried out in two cities, one in China (subway lines No. 2 & No. 4) and the other in USA (subway lines No. A & No. E), with 8 participants over a four-months time period.
- Figure 11 shows the four subway lines where the experiments are conducted.
- In the experiments, the authors ask every participant to carry the smartphone installed with SubTrack and take the subway with both random departure and destination stations for the four lines in two cities.
- There is no special requirement for the participants on their movement and holding styles of the smartphones while taking the train.
- All the system outcomes, including passenger station entrance, arrival time prediction, distance estimation and train stopping detection, will be stored in the memory for statistical analysis while the participants are conducting experiments.
B. Metrics
- The authors use the following five metrics to evaluate the performance of SubTrack.
- The distance estimation error is defined as the difference of the estimated train traveling distance and the actual distance the train travels.
- Precision is the fraction of the retrieved instances (i.e., the experiments that passengers are detected entering in the station) that are relevant, and recall is the fraction of the relevant instances (i.e., all experiments that passengers indeed enter in the station) that are retrieved.
- The stopping detection accuracy is the fraction of all train stops that are detected.
C. Distance Estimation Accuracy
- The authors first study the distance estimation accuracy of SubTrack and compare the performance with SubwayPS [5] that relies on the train schedule on timetable.
- The overall performance of SubTrack significantly outperforms SubwayPS.
- In contrast, according to the timetable of the subway line No.2 in the City-1 [4], the mean distance estimation error of SubwayPS is as high as 483m (35% of average distance 1380m between two adjacent stations), which is over seven times larger than that for SubTrack.
- It is also worthwhile to point out that the distance estimation error can be over 100m in few cases for SubTrack.
- This is because the subway train takes sharp turns at several places, resulting in severe acceleration fluctuation and thereby incorrect accelerometer readings.
D. Arrival Time Estimation Accuracy
- The authors next examine the arrival time estimation accuracy at different motion phases of the train.
- The authors first present overall arrival time prediction error shown as the black curves in Figures 13(a) and 13(b).
- The authors observe that median prediction errors are less than 6.4s and 7.8s, while 90% prediction error are less than 12s and 11s in City-1 and City2, respectively.
- Therefore, the arrival time prediction varies in wide range and the mean prediction error is the largest in all three phases.
- In the uniform motion phase, the average prediction error is less than that in the acceleration phase, since the train usually stays at a stable motion state velocity during this phase.
E. Cell-ID based Station Entering Point Detection
- Figure 15 depicts the performance of their station entering point detection method with different duration thresholds in two cities.
- The false triggering will also happen more frequently when the passenger walks by stations.
- The decreased recall rate indicates that SubTrack can not determine whether the passenger is already inside the station even after the passenger has entered.
- This is because the connection to cell-tower may be lost for seconds due to the poor signal quality at some spots inside the stations.
- Given the updated threshold, most of the false triggering are eliminated while the entering point can also be detected with the average precision at 0.975.
F. Train Stopping Detection
- Table II illustrates the train stopping detection accuracy with different sampling rates on motion sensors.
- The authors observe that the detection accuracy increases as the sampling rate increases.
- In particular, the detection accuracy reaches 0.955 when the sampling rate is higher than 60 Hz.
- Note that if the passengers prefer to save the energy for a longer battery lifespan, the detection accuracy can still maintain high level by reducing the sampling rate a little bit (e.g., 40 Hz).
- Comparing to the existing work [2, 3, 17], their system provides much better performance on the train stopping detection.
VIII. CONCLUSION
- The authors propose SubTrack aiming to achieve realtime automatic tracking of subway riding leveraging the builtin sensors on mobile devices.
- A realtime velocity calibration method is developed to mitigate the velocity drift error resulted from the inherent bias in the motion sensors.
- The authors algorithm takes the advantages of the accumulated acceleration readings and the typical moving patterns of the train to estimate the train’s velocity and position, and further predict the arrival time of the train in real time.
- Extensive experiments conducted in two cities, one in China and the other in USA, validate the effectiveness and efficiency of the proposed SubTrack system.
- The results demonstrate that SubTrack can accurately track the position of subway riders, predict the arrival time and push the arrival notification in a timely manner.
Did you find this useful? Give us your feedback
Citations
References
465 citations
"SubTrack: Enabling Real-Time Tracki..." refers background in this paper
...Smartphone-based localization has been widely studied in many different scenarios [4, 8]....
[...]
420 citations
"SubTrack: Enabling Real-Time Tracki..." refers methods in this paper
...In particular, velocity estimation of vehicles using GPS module in mobile devices has been developed [19, 20]....
[...]
...Existing studies [19, 20] could estimate vehicle velocity by using GPS, GSM signal strength or motion sensors of mobile devices....
[...]
300 citations
"SubTrack: Enabling Real-Time Tracki..." refers background or methods in this paper
...2 in the City-1 (China) [4], the mean distance estimation error of SubwayPS is as high as 483m (35% of average distance 1380m between two adjacent stations), which is over seven times larger than that for SubTrack....
[...]
...Existing work mainly relies on the built-in inertial sensors to allow smartphones to determine their location substantially [4, 5]....
[...]
...New approaches explore the possibility of using inertial sensors embedded in mobile devices to detect the train dynamics [4, 5] together with fixed timetables....
[...]
...Smartphone-based localization has been widely studied in many different scenarios [4, 8]....
[...]
...A few studies [4, 5] propose to track the train’s position using the timetable of subway....
[...]
109 citations
"SubTrack: Enabling Real-Time Tracki..." refers background in this paper
..., less than 2 Hz [10], whereas the train on the go produces a much higher frequency sensor readings....
[...]
89 citations
"SubTrack: Enabling Real-Time Tracki..." refers background or methods in this paper
...Existing studies [6, 7] show that the accumulated drift error in inertial navigation system increases linearly over the time, which suggests that such a linear relationship could be determined if some actual information (e....
[...]
..., real-time) velocity calibration scheme based on the proposition that the accumulated drift error increases linearly over time, which has been verified by many existing studies [6, 7]....
[...]
...An alternative way is to use built-in motion sensors in the mobile devices [7]....
[...]