Search

Spatiotemporal problems are ubiquitous and of vital importance in many research fields. Despite the potential already demonstrated by deep learning methods in modeling spatiotemporal data, typical approaches tend to focus solely on conditional expectations of the output variables being modeled. In this article, we propose a multioutput multiquantile deep learning approach for jointly modeling several conditional quantiles together with the conditional expectation as a way to provide a more complete “picture” of the predictive density in spatiotemporal problems. Using two large-scale data sets from the transportation domain, we empirically demonstrate that, by approaching the quantile regression problem from a multitask learning perspective, it is possible to solve the embarrassing quantile crossings problem while simultaneously significantly outperforming state-of-the-art quantile regression methods. Moreover, we show that jointly modeling the mean and several conditional quantiles not only provides a rich description about the predictive density that can capture heteroscedastic properties at a neglectable computational overhead but also leads to improved predictions of the conditional expectation due to the extra information and the regularization effect induced by the added quantiles. [ABSTRACT FROM AUTHOR]

Flexible calibration of dynamic traffic assignment (DTA) systems in real time has important applications in effective traffic management. However, the existing approaches are either limited to small networks or to a specific class of parameters. In this light, this study presents a framework to systematically reduce the dimension of the generic online calibration problem, making it more scalable. Specifically, a state–space formulation of the problem in the reduced dimension space is proposed. Following this the problem is solved using the constrained extended Kalman filter, which is made tractable because of the low dimensionality of the formulated problem. The effectiveness of the proposed approach is demonstrated using a real-world network leading to better state estimation by 13% and better state predictions by 11%—with a 50 fold dimensionality reduction. Insights into choosing the right degree of dimensionality reduction are also discussed. This work has the potential for a more widespread application of real-time DTA systems in practice.

Effective real-time traffic management strategies often require dynamic traffic assignment systems that are calibrated online. But the computationally intensive nature of online calibration limits their application to smaller networks. This paper presents a dimensionality reduction of the online calibration problem that is based on principal components to overcome this limitation. To demonstrate this approach, the origin–destination flow estimation problem is formulated in relation to its principal components. The efficacy of the procedure was tested with real data on the Singapore Expressway network in an open-loop framework. A reduction in the problem dimension by a factor of 50 was observed with only a 2% loss in estimation accuracy. Further, the computational times were reduced by an order of 100. The procedure led to better predictions, as the principal components captured the structural spatial relationships. This work has the potential to make the online calibration problem more scalable.

The calibration of dynamic traffic assignment (DTA) models involves the estimation of model parameters to best replicate real-world measurements. Good calibration is essential to estimate and predict accurately traffic states, which are crucial for traffic management applications to alleviate congestion. A widely used approach to calibrate simulation-based DTA models is the extended Kalman filter (EKF). The EKF assumes that the DTA model parameters are unconstrained, although they are in fact constrained; for instance, origin–destination (O-D) flows are nonnegative. This assumption is typically not problematic for small- and medium-scale networks in which the EKF has been successfully applied. However, in large-scale networks (which typically contain numbers of O-D pairs with small magnitudes of flow), the estimates may severely violate constraints. In consequence, simply truncating the infeasible estimates may result in the divergence of EKF, leading to extremely poor state estimations and predictions. To address this issue, a constrained EKF (CEKF) approach is presented; it imposes constraints on the posterior distribution of the state estimators to obtain the maximum a posteriori (MAP) estimates that are feasible. The MAP estimates are obtained with a heuristic followed by the coordinate descent method. The procedure determines the optimum and are computationally faster by 31.5% over coordinate descent and by 94.9% over the interior point method. Experiments on the Singapore expressway network indicated that the CEKF significantly improved model accuracy and outperformed the traditional EKF (up to 78.17%) and generalized least squares (up to 17.13%) approaches in state estimation and prediction.

Because of environmental and economic stress, current strong investment in adaptive transport systems can efficiently use capacity, minimizing costs and environmental impacts. The common vision is of a system that dynamically changes itself (the supply) to anticipate the needs of travelers (the demand). In some occasions, unexpected and unwanted demand patterns are noticed in the traffic network; these patterns lead to system failures and cost implications. Significantly, low speeds or excessively low flows at an unforeseeable time are only some of the phenomena that are often noticed and need to be explained for a transport system to develop a better future response. The objective of this research was the formulation of a methodology that could identify anomalies on traffic networks and correlate them with special events by using Internet data. The main subject of interest in this study was the investigation of why traffic congestion was occurring as well as why demand fluctuated on days when there were no apparent reasons for such phenomena. The system was evaluated by using Google’s public data set for taxi trips in New York City. A “normality” baseline was defined at the outset and then used in the subsequent study of the demand patterns of individual days to detect outliers. With the use of this approach it was possible to detect fluctuations in demand and to analyze and correlate them with disruptive event scenarios such as extreme weather conditions, public holidays, religious festivities, and parades. Kernel density analysis was used so that the affected areas, as well as the significance of the observed differences compared with the average day, could be depicted.

Public transportation authorities across the world are implementing various peak and off-peak pricing strategies to manage travel demand and improve the overall system performance. In this study, an activity-based demand framework was used to evaluate two off-peak pricing strategies currently in use in Singapore. These strategies consisted of a free prepeak travel on mass rapid transit (MRT) and an off-peak discount for an integrated transit (public buses and MRT). Smart card data collected before and after the implementation of the first policy were used to calibrate the behavioral models involved, to capture travelers’ preferences and choices properly. To evaluate both pricing strategies, a comprehensive set of key performance indicators was considered and included the changes in peak ridership, average trip fare, operator’s revenue, the number of public transportation trips, and mode share. The results indicate that off-peak discount pricing strategies are a viable policy option for spreading demand peaks and that they are more effective during the afternoon peak period. This study also demonstrates the capabilities and the advantages of an agent-based modeling platform, SimMobility, as a tool for policy analysis.

Cascading failures, such as the Northeast outage of 2003, can keep millions of people without power for extended periods of time. A combination of simultaneous occurrence of rare events in the physical grid, the human inability to apprehend and act on the real circumstances, and failures in the IT infrastructure result- ing in poor situation awareness may lead to power outages. The formation and propagation of cascading events is nonlinear. There is a precursor phase, in which some system recovery is possible and human oper- ators are able to act and recover failed lines, followed by an escalation phase, in which the number of failed lines increases rapidly, and by a cascading phase that corresponds to the blackout. Although stochastic power grid simulation is an appropriate tool to study cascading events, human intervention, which is mostly ignored in simulation, can nonetheless influence the recovery and the extension of the blackouts. In this pa- per we model the probability of human failure by applying a human reliability modeling approach, the, Standardized Plant Analysis Risk-Human Reliability Analysis (SPAR-H), to the analysis of failure scenari- os narrated by grid operators with different levels of expertise. The SPAR-H methodology is a simplified approach to human error quantification that accounts for individual performance-shaping factors that affect perception, processing and response to events in complex environments. To implement this approach, we started by interviewing system operators from New England and the South East, and took note of the narra- tives of the emergencies that happened during their shift. Next, we applied the SPAR-H methodology to each task and calculated dependency between the cognitive and action-based tasks. We then calculated the final failure probability and compared the empirical evaluation of error probability with what actually oc- curred. Using Monte Carlo methods and the SPAR-H equations, we finally estimated the HEP that can be readily integrated into the stochastic simulation of a physical power grid. With the intention to calculate the HEP specific to each cascading phase, we propose to increase the number of narrated events and classify each, according to a phase (precursor and escalation). We will then improve the model by calculating the frequency associated with each level of severity for each factor, and by creating a probability distribution for each factor, conditional to a specific phase.

The detection limit (about 0.017 μg mL−1) for voltammetric determination of iodide (peak at +0.87 V vs. Ag/AgCl at pH 2) at a glutaraldehyde‐cross‐linked poly‐L‐lysine modified glassy carbon electrode involving oxidation to iodine was found to be several orders of magnitude lower than that for the voltammetric determination on a bare glassy carbon electrode. This method was applied successfully to the determination of iodide in two medicinal formulations. Idoxuridine was determined indirectly at the same electrode by accumulating it first at −0.8 V vs. Ag/AgCl. At this potential the CI bond in the adsorbed idoxuridine is reduced giving iodide, which is then determined at the modified electrode. The method was successfully applied to the determination of idoxuridine in a urine sample.