How Long to Wait?: Predicting Bus Arrival Time with Mobile Phone based Participatory Sensing Pengfei Zhou, Yuanqing Zheng, Mo Li -twohsien 2012.9.3 Outline • • • • • Introduction System design Evaluation Limitations Conclusion Introduction • Why travelers do not like to travel by bus? – Excessively long waiting time • Existing methods to predict arrival time – Timetable ( operating hours, time intervals, etc.) – Special location tracking devices on buses Objective • Crowd-participated approach – Sharing users – Querying users Mobile Phone – Backend server • Energy friendly – Microphone, accelerometer Main idea • Map the bus routes to a space featured by sequences of nearby cellular towers Challenges • Bus Detection • Bus Classification • Information Assembling System Design Pre-processing Celltower Data Top-3 strongest cell towers 300 meters apart Example Bus Detection • Audio detection : short beep audio response Peak at 1 kHz and 3kHz Bus Detection • Sliding window, size: 32 samples • Empirical threshold: three standard deviation Bus Detection • Accelerometer detection – Bus v.s. Rapid train Bus Detection • Threshold – Small: trains will be misdetected as buses – Big: miss detection of actual buses Bus Classification • Cell tower sequence matching – Smith-Waterman algorithm • If ui = Cw ∈ Sj , ui and Sj are matching with each other, and mismatching otherwise Bus Classification • = 0.5−1 w: rank of signal strenth penalty cost for mismatches : -0.5 Overlapped route • Survey 50 bus route Range of cell tower: 300-900 meters threshold of celltower sequence length : 7 Cell tower Sequence Concatenation Arrival Time Prediction EVALUATION Experimental Methodology • Mobile phones – Samsung Galaxy S2 i9100 – HTC Desire • Experiment environment – 4 campus shuttle bus routes – 2 SBS transit bus route 179 and 241 Bus Detection Performance Bus vs. MRT Train False detection: Driving along straight routes late during night time Bus Classification Performance Arrival Time Prediction Arrival Time Prediction System Overhead • Battery lifetime Limitation and On-going Work • • • • Alternative reference points Number of passengers First few bus stops Overlapped routes Conclusion • Present a crowd-participated bus arrival time prediction system using commodity mobile phones. • Evaluate the system through a prototype system deployed on the Android platform with two types of mobile phones.