James Biagioni

Cooperative transit tracking using smart-phones

Real-time transit tracking is gaining popularity as a means for transit agencies to improve the rider experience. However, many transit agencies lack either the funding or initiative to provide such tracking services. In this paper, we describe a crowd-sourced alternative to official transit tracking, which we call cooperative transit tracking. Participating users install an application on their smart-phone. With the help of built-in sensors, such as GPS, WiFi, and accelerometer, the application automatically detects when the user is riding in a transit vehicle. On these occasions (and only these), it sends periodic, anonymized, location updates to a central tracking server. Our technical contributions include (a) an accelerometer-based activity classification algorithm for determining whether or not the user is riding in a vehicle, (b) a memory and time-efficient route matching algorithm for determining whether the user is in a bus vs. another vehicle, (c) a method for tracking underground vehicles, and an evaluation of the above on real-world data. By simulating the Chicago transit network, we find that the proposed system would shorten expected wait times by 2 minutes with only 5% of transit riders using the system. At a 20% penetration level, the mean wait time is reduced from 9 to 3 minutes.

Map inference in the face of noise and disparity

This paper describes a process for automatically inferring maps from large collections of opportunistically collected GPS traces. In this type of dataset, there is often a great disparity in terms of coverage. For example, a freeway may be represented by thousands of trips, whereas a residential road may only have a handful of observations. Additionally, while modern GPS receivers typically produce high-quality location estimates, errors over 100 meters are not uncommon, especially near tall buildings or under dense tree coverage. Combined, GPS trace disparity and error present a formidable challenge for the current state of the art in map inference. By tuning the parameters of existing algorithms, a user may choose to remove spurious roads created by GPS noise, or admit less-frequently traveled roads, but not both. In this paper, we present an extensible map inference pipeline, designed to mitigate GPS error, admit less-frequently traveled roads, and scale to large datasets. We demonstrate and compare the performance of our proposed pipeline against existing methods, both qualitatively and quantitatively, using a real-world dataset that includes both high disparity and noise. Our results show significant improvements over the current state of the art.

Mining large-scale, sparse GPS traces for map inference: comparison of approaches

We address the problem of inferring road maps from large-scale GPS traces that have relatively low resolution and sampling frequency. Unlike past published work that requires high-resolution traces with dense sampling, we focus on situations with coarse granularity data, such as that obtained from thousands of taxis in Shanghai, which transmit their location as seldom as once per minute. Such data sources can be made available inexpensively as byproducts of existing processes, rather than having to drive every road with high-quality GPS instrumentation just for map building - and having to re-drive roads for periodic updates. Although the challenges in using opportunistic probe data are significant, successful mining algorithms could potentially enable the creation of continuously updated maps at very low cost. In this paper, we compare representative algorithms from two approaches: working with individual reported locations vs. segments between consecutive locations. We assess their trade-offs and effectiveness in both qualitative and quantitative comparisons for regions of Shanghai and Chicago.

Inferring Road Maps from Global Positioning System Traces

As a result of the availability of Global Positioning System (GPS) sensors in a variety of everyday devices, GPS trace data are becoming increasingly abundant. One potential use of this wealth of data is to infer and update the geometry and connectivity of road maps through the use of what are known as map generation or map inference algorithms. These algorithms offer a tremendous advantage when no existing road map data are present. Instead of the expense of a complete road survey, GPS trace data can be used to generate entirely new sections of the road map at a fraction of the cost. In cases of existing maps, road map inference may not only help to increase the accuracy of available road maps but may also help to detect new road construction and to make dynamic adaptions to road closures—useful features for in-car navigation with digital road maps. In past research, proposed algorithms had been evaluated qualitatively with little or no comparison with prior work. This lack of quantitative and comparative evaluation is addressed in this paper with the following contributions: (a) a comprehensive survey of the current literature on map generation; (b) a description of the first method for the automatic evaluation of generated maps; (c) a qualitative, quantitative, and comparative evaluation of three reference algorithms; and (d) an open source implementation of each of the three algorithms, with a 118-h trace data set and ground truth map for unrestricted use by the automatic map generation community.

Days of Our Lives: Assessing Day Similarity from Location Traces

We develop and test algorithms for assessing the similarity of a person’s days based on location traces recorded from GPS. An accurate similarity measure could be used to find anomalous behavior, to cluster similar days, and to predict future travel. We gathered an average of 46 days of GPS traces from 30 volunteer subjects. Each subject was shown random pairs of days and asked to assess their similarity. We tested eight different similarity algorithms in an effort to accurately reproduce our subjects’ assessments, and our statistical tests found two algorithms that performed better than the rest. We also successfully applied one of our similarity algorithms to clustering days using location traces.

TransitGenie: a context-aware, real-time transit navigator

A transit navigation system is described that integrates real-time transit and user tracking with existing transit schedules to improve the transit riding experience.

Thrifty tracking: online GPS tracking with low data uplink usage

A typical online GPS tracking system uses a cellular uplink to report the location of a device to a central server, and in a study based on 1.6 billion location updates we find at least 90% are sent with a fixed 1--300 second period. Through experiments with the cost of cellular data transmission we also find that every packet sent incurs significant overhead. With these observations in mind, we describe a thrifty tracking system that allows the specification of a target error or budget-bound, while it optimizes the other. In our experiments, thrifty tracking outperforms the status quo by up to 20X while providing improved guarantees and flexibility.

Demo: Tracking transit with EasyTracker

EasyTracker is an automated system that assists small public or private transit agencies in deploying bus tracking and arrival time prediction. This demo will showcase how data from GPS sensors embedded in smartphones can be automatically processed in order to accurately estimate routes, bus stop locations, schedules, and make annotated maps with real-time bus tracking and arrival time predictions. We will also demonstrate a website portal which transit agencies can use to further interact with their bus transit systems.

Trading Off Accuracy, Timeliness, and Uplink Usage in Online GPS Tracking

In an online GPS tracking system, a fundamental trade-off exists between timeliness, the average time interval between a recorded change in location and a change in the reported location; accuracy, the average error between the actual location and the reported location; and uplink usage, the average amount of data used per second of tracking. While tracking efficiency has been addressed in the literature, our thrifty tracking system presents the first unified view of timeliness, accuracy, and uplink usage, allowing the user to specify desired targets for any two of these objectives, while optimizing the third. We also provide a closed-form characterization of this three-way trade-off, and demonstrate that our system converges to the predicted performance.