Jaimyoung Kwon

Day-to-Day Travel-Time Trends and Travel-Time Prediction from Loop-Detector Data

An approach is presented for estimating future travel times on a freeway using flow and occupancy data from single-loop detectors and historical travel-time information. Linear regression, with the stepwise-variable-selection method and more advanced tree-based methods, is used. The analysis considers forecasts ranging from a few minutes into the future up to an hour ahead. Leave-a-day-out cross-validation was used to evaluate the prediction errors without underestimation. The current traffic state proved to be a good predictor for the near future, up to 20 min, whereas historical data are more informative for longer-range predictions. Tree-based methods and linear regression both performed satisfactorily, showing slightly different qualitative behaviors for each condition examined in this analysis. Unlike preceding works that rely on simulation, real traffic data were used. Although the current implementation uses measured travel times from probe vehicles, the ultimate goal is an autonomous system that relies strictly on detector data. In the course of presenting the prediction system, the manner in which travel times change from day to day was examined, and several metrics to quantify these changes were developed. The metrics can be used as input for travel-time prediction, but they also should be beneficial for other applications, such as calibrating traffic models and planning models.

Detecting Errors and Imputing Missing Data for Single-Loop Surveillance Systems

Single-loop detectors provide the most abundant source of traffic data in California, but loop data samples are often missing or invalid. A method is described that detects bad data samples and imputes missing or bad samples to form a complete grid of clean data, in real time. The diagnostics algorithm and the imputation algorithm that implement this method are operational on 14,871 loops in six districts of the California Department of Transportation. The diagnostics algorithm detects bad (malfunctioning) single-loop detectors from their volume and occupancy measurements. Its novelty is its use of time series of many samples, instead of basing decisions on single samples, as in previous approaches. The imputation algorithm models the relationship between neighboring loops as linear and uses linear regression to estimate the value of missing or bad samples. This gives a better estimate than previous methods because it uses historical data to learn how pairs of neighboring loops behave. Detection of bad loops and imputation of loop data are important because they allow algorithms that use loop data to perform analysis without requiring them to compensate for missing or incorrect data samples.

Statistical Methods for Jointly Estimating the Decay Constant of 40K and the Age of a Dating Standard

To simultaneously evaluate the decay constant of 40K (λ) and the age of a standard (tstd) using isotopic data from geologic materials, we applied a series of statistical methods. The problem of estimating the most probable intercept of many nonlinear curves in λ and tstd space is formulated by an errors-in-variables nonlinear regression model. Then a maximum likelihood method is applied to the model for a point estimate, which is equivalent to the nonlinear least square method when measurement error distributions are Gaussian. Uncertainties and confidence regions of the estimates can be approximated using three methods: the asymptotic normal approximation, the parametric bootstrap method and Bonferroni confidence regions. Five pairs of published data for samples with ages from 2 ka to 4.5 Ga were used to estimate λ and the age of Fish Canyon sanidine (tFCs). The statistical procedure yields most probable estimates of λ (5.4755 ± 0.0170 × 10−10 (1σ)/year) and tFCs (28.269 ± 0.0661 (1σ) Ma) which are in between previously published values. These results indicate the power of our approach to provide improved constraints on these parameters, although the preliminary nature of some of the input data require further review before the values can be adopted.

Hybrid algorithm for indoor positioning using wireless LAN

Locating an indoor mobile station based on a wireless communication infrastructure, has practical applications. The most widely employed methods today use an RF propagation loss (PL) model or location fingerprinting (LF). The PL method is known to perform poorly compared to LF. But LF requires an extensive training dataset and cannot adapt well to configuration changes or receiver breakdown. In this paper, we develop a hybrid method that combines the strength of these two methods. It first formulates the RF PL in a nonlinear, censored regression model and adjusts the regression function to the observed signal strength in the fingerprint dataset. In the absence of a training dataset, the hybrid method coincides with the PL method, and, as the spatial granularity of the training dataset increases, the result of the algorithm approaches the result of the LF method. It balances the flexibility and accuracy of the two traditional methods, makes intelligent use of missing values, produces error bounds, and can be made dynamic. We evaluate the performance of the algorithm by applying it to a real site and observe satisfactory positioning accuracy.

Estimation of Truck Traffic Volume from Single Loop Detectors with Lane-to-Lane Speed Correlation

An algorithm for real-time estimation of truck traffic in multilane freeways was proposed. The algorithm used data from single loop detectors—the most widely installed surveillance technology for urban freeways in the United States. The algorithm worked for those freeway locations that have a truck-free lane and exhibit high lane-to-lane speed correlation. These conditions are met by most urban freeway locations. The algorithm produced real-time estimates of the truck traffic volumes at the location. It also can be used to produce alternative estimates of the mean effective vehicle length, which can improve speed estimates from single loop detector data. The algorithm was tested with real freeway data and produced estimates of truck traffic volumes with only 5.7% error. It also captured the daily patterns of truck traffic and mean effective vehicle length. Applied to loop data on Interstate 710 near Long Beach, California, during the dockworkers’ lockout October 1 to 9, 2002, the algorithm found a 32% reduction in five-axle truck volume.

Decomposition of Travel Time Reliability into Various Sources: Incidents, Weather, Work Zones, Special Events, and Base Capacity

An empirical, corridor-level method is proposed to divide the travel time unreliability or variability over a freeway section into the following components: incidents, weather, work zones, special events, and inadequate base capacity or bottlenecks. The method consists of three steps: (a) corridor-level aggregation of travel time and source data, (b) quantile regression to fit the 95th percentile of travel time on the source variables, and (c) calculation of the contribution of individual sources to the buffer time. It could be applied to other percentile-based travel time reliability measures such as planning time and 90th and 95th percentiles. Once the source data are defined, the method can be automatically applied to any site with minimum calibration. When applied to a 30.5-mi section of northbound I-880 in the San Francisco Bay Area, California, the method revealed that traffic accidents contributed 15.1% during the morning and 25.5% during the afternoon, among others, and most of the remaining reliability came from recurrent bottlenecks. Quantifying the components of travel time variability at individual freeway sites is essential in developing effective strategies to mitigate congestion.

TOPL: TOOLS FOR OPERATIONAL PLANNING OF TRANSPORTATION NETWORKS

TOPL is a suite of software tools for specifying freeway operational improvement strategies, such as ramp metering, demand and incident management, and for quickly estimating the benefits of such improvements. TOPL is based on the macroscopic cell transmission model. The paper summarizes the theory of the cell transmission model and describes the procedure to carry out a TOPL application. The procedure is illustrated for the 26-mile long I-210W freeway in California, whose model is calibrated using loop detector measurements of volume and speed. The measurements show that congestion originates in a bottleneck and moves upstream, as predicted by the theory. The simulations show that appropriate ramp metering can dramatically reduce total congestion delay and mainline travel time.Copyright

A new methodology for evaluating incident detection algorithms

We present a novel, off-line approach for evaluating incident detection algorithms. Previous evaluations have focused on determining the detection rate versus false alarm rate curve -- a process which we argue is inherently fraught with difficulties. Instead, we propose a cost-benefit analysis where cost mimics the real costs of implementing the algorithm and benefit is in terms of reduction in congestion. We argue that these quantities are of more practical interest than the traditional rates. Moreover, these costs, estimated on training data, can be used both as a mechanism to fine-tune a single algorithm as well as a meaningful quantity for direct comparisons between different types of incident detection algorithms. We demonstrate our approach with a detailed example. Key words: Incident detection

Identification of genes induced in proteoid roots of white lupin under nitrogen and phosphorus deprivation, with functional characterization of a formamidase

White lupin (Lupinus albus L.) is considered a model system for understanding plant acclimation to nutrient deficiency. It acclimates to phosphorus (P) and iron (Fe) deficiency by the development of short, densely clustered lateral roots called proteoid (or cluster) roots; proteoid-root development is further influenced by nitrogen (N) supply. In an effort to better understand proteoid root function under various nutrient deficiencies, we used nylon filter arrays to analyze 2,102 expressed sequence tags (ESTs) from proteoid roots of P-deficient white lupin. These have been previously analyzed for up-regulation in −P proteoid roots, and were here analyzed for up-regulation in proteoid roots of N-deprived plants. We identified a total of 19 genes that displayed up-regulation in proteoid roots under both P and N deprivation. One of these genes showed homology to putative formamidases. The corresponding open reading frame was cloned, overexpressed in E. coli, and the encoded protein was purified; functional characterization of the recombinant protein confirmed formamidase activity. Though many homologues of bacterial and fungal formamidases have been identified in plants, to our knowledge, this is the first report of a functional characterization of a plant formamidase.

Statistical methods for detecting spatial configuration errors in traffic surveillance sensors

With large-scale deployment of traffic surveillance sensors becoming commonplace, it becomes critical to maintain correct information about the spatial configuration of the sensors. The problem is burdensome when hundreds or thousands of sensors are deployed. One common configuration error is the switching of directions of highway loop detectors that share the same cabinet. Proposed are semiautomatic and automatic methods for detecting such errors, on the basis of the strong correlations between measurements made by spatially close sensors. The semiautomatic method uses a multidimensional scaling (MDS) map of sensors, which visually displays the similarity between sensor measurements and enables one to easily identify sensor mislabeling. The automatic method uses a scoring scheme that computes the probability of sensor mislabeling from the pairwise distance or similarity matrix. The algorithm, tested on data from a four-lane freeway consisting of 64 sensor locations—10 of which had switched locations—successfully detected all errors with 5.6% false detection rate, even with poor data quality. The MDS map can be used for other applications, such as detection of sensor malfunctions.