Nikola Mitrovic

Online map-matching based on Hidden Markov model for real-time traffic sensing applications

In many Intelligent Transportation System (ITS) applications that crowd-source data from probe vehicles, a crucial step is to accurately map the GPS trajectories to the road network in real time. This process, known as map-matching, often needs to account for noise and sparseness of the data because (1) highly precise GPS traces are rarely available, and (2) dense trajectories are costly for live transmission and storage. We propose an online map-matching algorithm based on the Hidden Markov Model (HMM) that is robust to noise and sparseness. We focused on two improvements over existing HMM-based algorithms: (1) the use of an optimal localizing strategy, the variable sliding window (VSW) method, that guarantees the online solution quality under uncertain future inputs, and (2) the novel combination of spatial, temporal and topological information using machine learning. We evaluated the accuracy of our algorithm using field test data collected on bus routes covering urban and rural areas. Furthermore, we also investigated the relationships between accuracy and output delays in processing live input streams. In our tests on field test data, VSW outperformed the traditional localizing method in terms of both accuracy and output delay. Our results suggest that it is viable for low latency applications such as traffic sensing.

Spatiotemporal Patterns in Large-Scale Traffic Speed Prediction

The ability to accurately predict traffic speed in a large and heterogeneous road network has many useful applications, such as route guidance and congestion avoidance. In principle, data-driven methods, such as support vector regression (SVR), can predict traffic with high accuracy because traffic tends to exhibit regular patterns over time. However, in practice, the prediction performance can significantly vary across the network and during different time periods. Insight into those spatiotemporal trends can improve the performance of intelligent transportation systems. Traditional prediction error measures, such as the mean absolute percentage error, provide information about the individual links in the network but do not capture global trends. We propose unsupervised learning methods, such as k-means clustering, principal component analysis, and self-organizing maps, to mine spatiotemporal performance trends at the network level and for individual links. We perform prediction for a large interconnected road network and for multiple prediction horizons with an SVR-based algorithm. We show the effectiveness of the proposed performance analysis methods by applying them to the prediction data of the SVR.

Low-dimensional models for missing data imputation in road networks

Intelligent transport systems (ITS) require data with high spatial and temporal resolution for applications such as modeling, traffic management, prediction and route guidance. However, field data is usually quite sparse. This problem of missing data severely limits the effectiveness of ITS. Missing values are usually imputed by either using historical data of the road or current information from neighboring links. In most scenarios, information from some or all of neighboring links might not be available. Furthermore, historical data may also be incomplete. To overcome these issues, we propose methods which can construct low-dimensional representation of large and diverse networks, in presence of missing historical and neighboring data. We use these low-dimensional models to reconstruct data profiles for road segments, and impute missing values. To this end we use Fixed Point Continuation with Approximate SVD (FPCA) and Canonical Polyadic (CP) decomposition for incomplete tensors to solve the problem of missing data. We apply these methods to expressways and a large urban road network to assess their performance for different scenarios.

Matrix and Tensor Based Methods for Missing Data Estimation in Large Traffic Networks

Intelligent transportation systems (ITSs) gather information about traffic conditions by collecting data from a wide range of on-ground sensors. The collected data usually suffer from irregular spatial and temporal resolution. Consequently, missing data is a common problem faced by ITSs. In this paper, we consider the problem of missing data in large and diverse road networks. We propose various matrix and tensor based methods to estimate these missing values by extracting common traffic patterns in large road networks. To obtain these traffic patterns in the presence of missing data, we apply fixed-point continuation with approximate singular value decomposition, canonical polyadic decomposition, least squares, and variational Bayesian principal component analysis. For analysis, we consider different road networks, each of which is composed of around 1500 road segments. We evaluate the performance of these methods in terms of estimation accuracy, variance of the data set, and the bias imparted by these methods.

CUR decomposition for compression and compressed sensing of large-scale traffic data

Intelligent Transportation Systems (ITS) often operate on large road networks, and typically collect traffic data with high temporal resolution. Consequently, ITS need to handle massive volumes of data, and methods to represent that data in more compact representations are sorely needed. Subspace methods such as Principal Component Analysis (PCA) can create accurate low-dimensional models. However, such models are not readily interpretable, as the principal components usually involve a large number of links in the traffic network. In contrast, the CUR matrix decomposition leads to low-dimensional models where the components correspond to individual links in the network; the resulting models can be easily interpreted, and can also be used for compressed sensing of the traffic network. In this paper, the CUR matrix decomposition is applied for two purposes: (1) compression of traffic data; (2) compressed sensing of traffic data. In the former, only data from a “random” subset of links and time instances is stored. In the latter, data for the entire traffic network is inferred from measurements at a “random” subset of links. Numerical results for a large traffic network in Singapore demonstrate the feasibility of the proposed approach.

Low-Dimensional Models for Compressed Sensing and Prediction of Large-Scale Traffic Data

Advanced sensing and surveillance technologies often collect traffic information with high temporal and spatial resolutions. The volume of the collected data severely limits the scalability of online traffic operations. To overcome this issue, we propose a low-dimensional network representation where only a subset of road segments is explicitly monitored. Traffic information for the subset of roads is then used to estimate and predict conditions of the entire network. Numerical results show that such approach provides 10 times faster prediction at a loss of performance of 3% and 1% for 5- and 30-min prediction horizons, respectively.

Unsupervised learning based performance analysis of n-support vector regression for speed prediction of a large road network

Many intelligent transportation systems (ITS) applications require accurate prediction of traffic parameters. Previous studies have shown that data driven machine learning methods like support vector regression (SVR) can effectively and accurately perform this task. However, these studies focus on highways, or a few road segments. We propose a robust and scalable method using v-SVR to tackle the problem of speed prediction of a large heterogeneous road network. The traditional performance measures such as mean absolute percentage error (MAPE) and root mean square error (RMSE) provide little insight into spatial and temporal characteristics of prediction methods for a large network. This inadequacy can be a serious hurdle in effective implementation of prediction models for route guidance, congestion avoidance, dynamic traffic assignment and other ITS applications. We propose unsupervised learning techniques by employing k-means clustering, principal component analysis (PCA), and self organizing maps (SOM) to overcome this insufficiency. We establish the effectiveness of the developed methods by evaluation of spatial and temporal characteristics of prediction performance of the proposed variable window v-SVR method.

Wavelets on graphs with application to transportation networks

The technological advancements in Intelligent Transport Systems have made it possible to acquire large amounts of traffic data in real-time. As a result, various data-mining techniques are being used to extract useful traffic patterns. The research presented in this article focuses on the detection of disruptive traffic events such as congestion. In most transportation studies, traffic parameters are typically modeled as time series. However, these techniques fail to incorporate the spatial dependencies between different traffic variables. In this work, the traffic quantities such as speeds are considered as the signals defined at the vertices of a network line graph. Furthermore, the graph wavelet operators are applied to the spatial signals to generate the wavelet coefficients at different wavelet scales. By analyzing these wavelet coefficients, useful information such as origin, propagation, and the span of traffic congestion are inferred. For analysis, we consider two major expressways in Singapore. The analysis shows that the abrupt changes in the speed can be captured by using the wavelet coefficients at the higher scales. On the other hand, the high magnitude coefficients at the lower wavelet scales reflect the smooth flow of the traffic across the network.

Near-Lossless Compression for Large Traffic Networks

With advancements in sensor technologies, intelligent transportation systems can collect traffic data with high spatial and temporal resolution. However, the size of the networks combined with the huge volume of the data puts serious constraints on system resources. Low-dimensional models can help ease these constraints by providing compressed representations for the networks. In this paper, we analyze the reconstruction efficiency of several low-dimensional models for large and diverse networks. The compression performed by low-dimensional models is lossy in nature. To address this issue, we propose a near-lossless compression method for traffic data by applying the principle of lossy plus residual coding. To this end, we first develop a low-dimensional model of the network. We then apply Huffman coding (HC) in the residual layer. The resultant algorithm guarantees that the maximum reconstruction error will remain below a desired tolerance limit. For analysis, we consider a large and heterogeneous test network comprising of more than 18 000 road segments. The results show that the proposed method can efficiently compress data obtained from a large and diverse road network, while maintaining the upper bound on the reconstruction error.

Bayesian Support Vector Regression for traffic speed prediction with error bars

Traffic prediction algorithms can help improve the performance of Intelligent Transportation Systems (ITS). To this end, ITS require algorithms with high prediction accuracy. For more robust performance, the traffic systems also require a measure of uncertainty associated with prediction data. Data driven algorithms such as Support Vector Regression (SVR) perform traffic prediction with overall high accuracy. However, they do not provide any information about the associated uncertainty. The prediction error can only be calculated once field data becomes available. Consequently, the applications which use prediction data, remain vulnerable to variations in prediction error. To overcome this issue, we propose Bayesian Support Vector Regression (BSVR). BSVR provides error bars along with the predicted traffic states. We perform sensitivity and specificity analysis to evaluate the efficiency of BSVR in anticipating variations in prediction error. We perform multi-horizon prediction and analyze the performance of BSVR for expressways as well as general road segments.