Travel Time Delay Research Articles

Water flow pattern recognition and fracture characterization in near-surface chalks using time-lapse ground-penetrating radar: An experiment

Predicting subsurface water flow is inherently complex due to the heterogeneity of geological formations. Water infiltration occurs through various pathways, such as pore throats, fractures, or faults, depending on the sediment type. In an effort to enhance our understanding of water flow within heterogeneous, fractured chalks formations, we integrate ground penetrating radar (GPR), acoustic measurements, and photogrammetric surveys to examine the flow patterns. We first conduct a field experiment of injecting water into chalks and monitor the water flow through a time-lapse GPR reflection survey. Then we identify the water-affected zone by analyzing the GPR data, and correlate the water-affected zone with the identified fractures on the exposed strata. Finally, we verify the water effect through measured acoustic velocities of acquired samples. Our analysis of the GPR data encompasses direct subtraction, traveltime delay, electromagnetic (EM) velocity changes, and water saturation variations. These analyses consistently indicate a predominant southeastward flow direction. This finding is corroborated by the fractures identified through photogrammetry, which closely match the water-affected zone, suggesting that water flow is indeed fracture-driven. The measured acoustic velocities are utilized to estimate water saturation using an iso-frame rock physics model, which shows consistency with the saturation estimated through EM velocity, further validating the monitoring results. These findings advance the understanding of fluid flow and its interaction with fracture system in heterogeneous chalks, and provide a reference for future studies by demonstrating the efficacy of integrating photogrammetry-identified fractures, EM velocity, and acoustic velocity to verify monitoring outcomes.

Travel Time Variability in Urban Mobility: Exploring Transportation System Reliability Performance Using Ridesharing Data

Travel time variability (TTV) is a crucial indicator of transportation network performance, assessing travel time reliability and delays. This study investigates TTV metrics within the context of shared mobility using probe data from transportation network companies (TNCs) in Chicago, Los Angeles, and Dallas–Fort Worth. Eight reliability metrics are analyzed and compared for each origin–destination (OD) pair in the network, including standard deviation (SD), the Planning Time Index (PTI), the Travel Time Index (TTI), the Buffer Index (BI), On-time Measures PR (alpha), and the Misery Index (MI), to evaluate their effectiveness in clustering OD pairs using K-means clustering. The findings confirm that SD, PTI, and MI are particularly effective in measuring travel time reliability and clustering within urban systems. This study identifies the most unbalanced supply–demand OD pairs/regions in each city, noting that low/medium-SD clusters around metropolitan airports indicate stable travel times even in high-demand zones, while high-SD clusters in downtown areas reveal significant traffic demands and unreliability. These patterns become more pronounced in study areas with multiple city centers. This study highlights the need for targeted strategies to enhance travel time reliability, particularly in regions like Dallas–Fort Worth, where public transportation alternatives are limited.

Effects of 2.5-D ultra-low and ultra-high velocity zones on flip-reverse-stacking (FRS) of the ScS wavefield

SUMMARY Within the last decade, thin ultra-low velocity zone (ULVZ) layering, sitting directly on top of the core–mantle boundary (CMB), has begun to be investigated using the flip-reverse-stack (FRS) method. In this method, pre- and post-cursor arrivals that are symmetrical in time about the ScS arrival, but with opposite polarities, are stacked. This same methodology has also been applied to high velocity layering, with indications that ultra-high velocity zones (UHVZs) may also exist. Thus far, studies using the FRS technique have relied on 1-D synthetic predictions to infer material properties of ULVZs. 1-D ULVZ models predominantly show a SdS precursor that reflects off the top of the ULVZ and an ScscS reverberation within the ULVZ that arrives as a post-cursor. 1-D UHVZ models are more complex and have a different number of arrivals depending on a variety of factors including UHVZ thickness, velocity contrast, and lateral extent. 1-D modelling approaches assume that lower mantle heterogeneity is constant and continuous everywhere across the lower mantle. However, lower mantle features display lateral heterogeneity and are either finite in extent or display local thickness variations. We examine the interaction of the ScS wavefield with ULVZs and UHVZs in 2.5-D geometries of finite extent. We show that multiple additional arrivals exist that are not present in 1-D predictions. In particular, multipath ScS arrivals as well as additional post-cursor arrivals are generated. Subsequent processing by the FRS method generates complicated FRS traces with multiple peaks. Furthermore, post-cursor arrivals can be generated even when the ScS ray path does not directly strike the heterogeneity from above. Analysing these predictions for 2.5-D models using 1-D modelling techniques demonstrates that a cautious approach must be adopted in utilization and interpretation of FRS traces to determine if the ScS wavefield is interacting with a ULVZ or UHVZ through a direct strike on the top of the feature. In particular, traveltime delays or advances of the ScS arrival should be documented and symmetrical opposite polarity arrivals should be demonstrated to exist around ScS. The latter can be quantified by calculation of a time domain multiplication trace. Because multiple post-cursor arrivals are generated by finite length heterogeneities, interpretation should be confined to single layer models rather than to interpret the additional peaks as internal layering. Furthermore, strong trade-offs exist between S-wave velocity perturbation and thickness making estimations of ULVZ or UHVZ elastic parameters highly uncertain. We test our analysis methods using data from an event occurring in the Fiji-Tonga region recorded in North America. The ScS bounce points for this event sample the CMB region to the southeast of Hawaii, in a region where ULVZs have been identified in several recent studies. We see additional evidence for a ULVZ in this region centred at 14°N and 153°W with a lateral scale of at least 250 km × 360 km. Assuming a constant S-wave velocity decrease of −10 or −20 per cent with respect to the PREM model implies a ULVZ thickness of up to 16 or 9 km, respectively.

Unveiling community adaptations to extreme heat events using mobile phone location data

The escalating frequency, duration, and intensity of extreme heat events have posed a significant threat to human society in recent decades. Understanding the dynamic patterns of human mobility under extreme heat will contribute to accurately assessing the risk of extreme heat exposure. This study leverages an emerging geospatial data source, anonymous cell phone location data, to investigate how people in different communities adapt travel behaviors responding to extreme heat events. Taking the Greater Houston Metropolitan Area as an example, we develop two indices, the Mobility Disruption Index (MDI) and the Activity Time Shift Index (ATSI), to quantify diurnal mobility changes and activity time shift patterns at the city and intra-urban scales. The results reveal that human mobility decreases significantly in the daytime of extreme heat events in Houston while the proportion of activity after 8 p.m. is increased, accompanied with a delay in travel time in the evening. Moreover, these mobility-decreasing and activity-delaying effects exhibited substantial spatial heterogeneity across census block groups. Causality analysis using the Geographical Convergent Cross Mapping (GCCM) model combined with correlation analyses indicates that people in areas with a high proportion of minorities and poverty are less able to adopt heat adaptation strategies to avoid the risk of heat exposure. These findings highlight the fact that besides the physical aspect of environmental justice on heat exposure, the inequity lies in the population's capacity and knowledge to adapt to extreme heat. This research is the first of the kind that quantifies multi-level mobility for extreme heat responses, and sheds light on a new facade to plan and implement heat mitigations and adaptation strategies beyond the traditional approaches.

Assessing Multifaceted Effects of Speed Humps and Bumps: Travel Time, Safety, and Environmental Considerations

This study focuses on investigating the significant impacts of speed breakers on various parameters, including travel time delays, vehicle speeds, fuel consumption, pavement maintenance costs, and vehicular exhaust emissions. Field data was collected and analyzed to assess the effects of different types of traffic calming measures on these parameters. The findings provide valuable insights into the implications of speed breakers on road safety, environmental pollution, and overall road infrastructure management. The results reveal that the implementation of speed humps, speed bumps, and triple bumps effectively slows down vehicles, as evidenced by considerable reductions in the 85th percentile speeds. The reduction percentages were 41.65% for speed humps, 73.52% for speed bumps, and 86.27% for triple bumps. This indicates the effectiveness of these traffic calming measures in improving road safety by reducing vehicle speeds. However, the presence of speed breakers also leads to increased travel time delays. On average, traversing stretches with speed humps, speed bumps, and triple bumps resulted in delays of 9.31, 16.42, and 29.51 seconds, respectively. While the individual delay times may appear relatively short, the cumulative effect of multiple speed obstacles along a road needs to be considered. Another significant impact observed is the increased fuel consumption associated with speed breakers. The study found that for every 100 km of travel, motorcycles and passenger cars consumed approximately 12.07 km and 27.37 km of additional fuel, respectively, when the density of speed breakers was 1.33/km. This translates to a fuel consumption increase of 13.73% for motorcycles and 37.74% for passenger cars. Furthermore, the presence of speed humps was found to contribute to pavement deterioration, as indicated by decreased Pavement Condition Index (PCI) values. The study also revealed that sections with speed humps incurred significantly higher maintenance costs compared to sections without speed humps. The increase in maintenance cost ranged from 100 to 264% across different road sections, with higher traffic volumes leading to greater cost escalation. Additionally, the study confirms that lower vehicle speeds, particularly between 0-15 km/hr, are associated with higher emissions of pollutants, including carbon monoxide (CO) and other pollutants. This highlights the environmental implications of speed breakers and their contribution to urban air pollution. Doi: 10.28991/CEJ-2024-010-07-07 Full Text: PDF

Reliability and Unreliability Aspects of Travel Time Analysis on the Stochastic Network Using the Target-Oriented Perspective

This study proposes a target-oriented method to study travelers’ route choice behavior under travel time variability, and discusses the resulted equilibrium flow patterns. Both travel time reliability and travel time unreliability are considered in this new method, and accordingly, there are two targets. The first one is target for travel time to ensure travel time reliability, and based on this target, another one is target for excess delay to mitigate travel time unreliability. In this model, travel time and excess delay (i.e., the random vector) are stochastically correlated with each other, which is modeled with the copula function based on Sklar’s Theorem, and the exact form of the copula is obtained by the proved comonotonicity relationship of this random vector. The target interaction, i.e., the complementarity relationship, is also modeled based on the utility functions, the meaning of which is that travelers have the will to make more targets achieved so as to obtain more utility. Furthermore, with this model, this paper formulates the user equilibrium as a variational inequality problem to study the long-term effect of the route choice behavior, and solves it with the method of successive average. Finally, numerical testings on the traffic network are conducted to show the convergence of the solution algorithm, and to illustrate the impact of targets on the equilibrium results. Results show that the flow change can be five times more than that with less risk-averse travelers.

Modeling the economic cost of congestion in Addis Ababa City, Ethiopia

Road traffic which results in significant time lags, increased fuel consumption, and financial losses, remains a noteworthy challenge in developed and developing countries. As a result, the Ethiopian Government and the City Administration of Addis Ababa have built extensive road networks and imposed restrictions on driving, vehicle acquisition and parking. However, despite all these efforts, drivers and passengers waste significant time on long traffic queues, resulting in unpredictable and delayed travel. The current study investigated the cost of travel time delay, vehicle operating costs, time reliability, and the factors influencing these variables. The study used questionnaires, measurements, and traffic counting techniques to collect data from nine road segments. The sample comprised 3240 participants. The cost functions of both drivers and passengers were examined using a multiple linear regression model, with estimation performed using ordinary least squares. According to the findings, the economic costs of congestion depend on the number of lanes, the length of the road segment, the volume of traffic, and the respondents’ income level. The study also revealed that travel, vehicle operation, and unreliability costs account for 74%, 6%, and 20%, respectively, of the total congestion costs.

A matching pursuit approach to the geophysical inverse problem of seismic travel time tomography under the ray theory approximation

Summary Seismic travel time tomography is a geophysical imaging method to infer the 3-D interior structure of the solid Earth. Most commonly formulated as a linearized inverse problem, it maps differences between observed and expected wave travel times to interior regions where waves propagate faster or slower than the expected average. The Earth’s interior is typically parametrized by a single kind of localized basis function. Here we present an alternative approach that uses matching pursuits on large dictionaries of basis functions. Within the past decade the (Learning) Inverse Problem Matching Pursuits ((L)IPMPs) have been developed. They combine global and local trial functions. An approximation is built in a so-called best basis, chosen iteratively from an intentionally overcomplete set or dictionary. In each iteration, the choice for the next best basis element reduces the Tikhonov–Phillips functional. This is in contrast to classical methods that use either global or local basis functions. The LIPMPs have proven their applicability in inverse problems like the downward continuation of the gravitational potential as well as the MEG-/EEG-problem from medical imaging. Here, we remodel the Learning Regularized Functional Matching Pursuit (LRFMP), which is one of the LIPMPs, for travel time tomography in a ray theoretical setting. In particular, we introduce the operator, some possible trial functions and the regularization. We show a numerical proof of concept for artificial travel time delays obtained from a contrived model for velocity differences. The corresponding code is available online.

Mathematical Modeling on Integrated Vehicle Assignment and Rebalancing in Ride-hailing System with Uncertainty Using Fuzzy Linear Programming

The general public frequently uses taxis as local transportation to get from one location to another. Ride-hailing is an innovation in taxi services that lets customers use their smartphones to find drivers, find prices, and submit requests. The two parts of ride-hailing are vehicle assignment and rebalancing. The task of the assignment is to allocate resources as efficiently as possible to fulfil demand. Demand and supply in the area are brought into balance through the rebalancing process. The rebalancing step is frequently carried out independently of the assignment process in ride-hailing pickup systems. However, it can be integrated into a single optimization process to improve system performance. The linear programming approach can integrate assignment and rebalancing. The batch assignment is an assignment algorithm in which each supply and demand is collected within a specific time window. The assignment is done after the vehicles and requests are collected. Fuzzy linear programming is used to deal with environmental uncertainty. Uncertain demand can reduce the reliability of ride-hailing pickup systems in addressing the problem of allocating empty vehicles to areas of high demand. The pickup time may change over time due to traffic conditions. Assignment integration and rebalancing modelling with fuzzy parameters are carried out to obtain a ride-hailing pickup system model that can handle demand uncertainty, pickup travel time, and travel delay times and increase the ride-hailing effectiveness of the pickup system. Numerical simulations were carried out on the assignment and rebalancing integration model without and with uncertainty parameters—numerical simulations based on publicly available taxi travel request data. The model selection for the ride-hailing system can be determined based on the numerical simulation results. This research can potentially advance and expand knowledge, especially in transportation.

Cooperative traffic signal control through a counterfactual multi-agent deep actor critic approach

Signal control has been effective to alleviate urban traffic congestion. Massive related works about signal timing optimization have been proposed and led to many signal control methods and systems. In recent years, reinforcement learning (RL) algorithms have attracted the increasing attention of researchers in the area of signal control optimization, since they can learn the optimal timing policy themselves by analyzing changing patterns between the traffic condition and signal timing plans. Multi-intersection traffic signal control problems face more challenges than classical single intersection problems such as the dimensionality issue as the joint action space of multiple agents grows exponentially with the number of intersections. The existing state-of-art multi-agent reinforcement learning algorithms developed for other areas may not suit well with traffic signal control given the complex spatial–temporal nature of traffic flows. In this paper, we propose a novel multi-agent counterfactual actor–critic with scheduler (MACS) framework for multiple interactions. In this method, decentralized actors control the traffic signals and the centralized critic combines recurrent policies with feed-forward critics. Additionally, a scheduler module that exchanges information among agents helps the individual agent better understand the entire environment. The proposed method was evaluated using simulation experiments based on a real-world urban street network in Shenzhen, China. Results showed that the proposed method outperforms the classic model-based method and several existing RL-based methods according to different measures such as queue length, average travel time delay, and throughput.

Methodology for Monitoring Border Crossing Delays with Connected Vehicle Data: United States and Mexico Land Crossings Case Study

International trade is a critical part of the United States economy. Land border crossings between the United States and Mexico accounts for a large proportion of the USD 779 billion in trade between these two countries. Monitoring and managing the operations of these land border crossings is critical for ensuring efficient trade and providing appropriate security. This paper examines the opportunity to use connected vehicle data to monitor the travel time delay of passenger vehicles crossing the border for system level assessment across 26 border crossing locations over an analysis period of 25 days in August 2020. A sample size of 51,341 trips from the US to Mexico and 41,708 trips from Mexico to the US were used in this study. Furthermore, 97% trips to the US and 76% trips to Mexico experienced delays. The average delay was 34 min for trips to the US compared to only 2 min for trips to Mexico. In terms of the predictability of border crossing times, there was also substantial variation by direction. The interquartile range of vehicle delay from the US to Mexico was 2 min, while the interquartile range of delay for vehicles travelling from Mexico to the US was 46 min. Border crossings were also ranked using four performance metrics—trip counts, median delay, delayed trip counts and total delays in vehicle hours. Methods for summarizing delay trends by time of the day and day of the week to identify time windows of interest are also presented. Land border crossing operations have a significant influence on security and economic efficiency. We believe the techniques presented in this paper provide a scalable methodology for providing near real-time factual data on border crossing delays that provide important information for land border transport-managing stakeholders to make informed management decisions that balance security and economic efficiency.

An edge architecture for enabling autonomous aerial navigation with embedded collision avoidance through remote nonlinear model predictive control

In this article, we present an edge-based architecture for enhancing the autonomous capabilities of resource-constrained aerial robots by enabling a remote nonlinear model predictive control scheme, which can be computationally heavy to run on the aerial robots' onboard processors. The nonlinear model predictive control is used to control the trajectory of an unmanned aerial vehicle while detecting, and preventing potential collisions. The proposed edge architecture enables trajectory recalculation for resource-constrained unmanned aerial vehicles in relatively real-time, which will allow them to have fully autonomous behaviors. The architecture is implemented with a remote Kubernetes cluster on the edge side, and it is evaluated on an unmanned aerial vehicle as our controllable robot, while the robotic operating system is used for managing the source codes, and overall communication. With the utilization of edge computing and the architecture presented in this work, we can overcome computational limitations, that resource-constrained robots have, and provide or improve features that are essential for autonomous missions. At the same time, we can minimize the relative travel time delays for time-critical missions over the edge, in comparison to the cloud. We investigate the validity of this hypothesis by evaluating the system's behavior through a series of experiments by utilizing either the unmanned aerial vehicle or the edge resources for the collision avoidance mission.

Max-Pressure Control utilizing Total Movement Delay

Traffic signal control algorithms utilizing max-pressure principles can rely on feedback obtained from either queue lengths or travel times. This paper considers max-pressure algorithms based on travel times, aiming to address traffic-related challenges, including (i) unbounded link capacity, (ii) mitigating the “the Last Packet effect”, and (iii) enhancing queue modeling to account for spatial queue distribution. Hence, we propose a new max-pressure algorithm based on total movement delays, which requires local information only to determine appropriate phase activation, and aims at minimizing delays. First, we introduce a new traffic model based in queuing theory. Unlike previous approaches that rely on the store-and-forward model, our model accurately depicts vehicle travel time delays. Second, based on the developed model, we introduce a new traffic oriented max-pressure algorithm. The objective of maximizing throughput is compared to minimizing travel time delays, through the comparison of different a-cyclic time-based max-pressure control schemes with existing a-cyclic queue-based algorithms control schemes.

Impacts of public transit delays and disruptions on equity seeking groups in Toronto – A time-expanded graph approach

This study employed graph theory techniques to evaluate the global efficiency of Toronto's transit network. Using global efficiency, which measures the ease with which riders could reach their destinations, we explored if disadvantaged riders suffered the same level of travel time delays as the general population during both minor (recurrent) service disturbances and major (non-recurrent) disruptions. We modified the global efficiency measure by applying weights representing origin-destination trip flows for six population groups of interest separately. We used a time-expanded graph, which we built using real-time data scraped from the public APIs of the Toronto Transit Commission. This approach allowed us to create travel time estimates that considered the effects of real-world headway performance on travel times over 16 days during the winter of 2019. Our analysis found that disadvantaged groups experienced similar reductions in global efficiency and similar increases in travel time during both routine delays and major transit disruptions compared to the general population. Groups with predominantly suburban travel patterns, such as Black and Low-Income riders, had slightly lower drops in global efficiency than the general population, showing their better resilience to delays and disruption.

The implications of S-wave attenuation in geothermal reservoirs

An increasing number of research studies have found that S-waves have significant attenuation and travel time delays as they pass through geothermal reservoirs and magma chambers. For this study, we used a micro-seismic dense array of recording stations to model the underground structure of a geothermal volume and analyzed recorded seismic waves. We analyze waveform data from twenty-five stations and identify whether P- and S-waves have been attenuated. We utilize tomographic solutions of P- and S-wave velocities and derived rock physics parameters, including Bulk, Shear, Lambda, Young's Moduli, and Poisson's ratio, to identify underground geological structures. We speculate the cause of S-wave attenuation and travel time delays. From the rock physics interpretations, the subterranean geothermal reservoir may be further explored to identify fractures, porosity, saturation, and permeability properties. We found that unusually high S-wave attenuation is usually related to a geothermal reservoir zone or magma, and few, if any, earthquakes occur within these zones. We test a rapid, simple, and inexpensive means to provide an early preliminary analysis of geothermal potential.

Exploring trade-offs in public bus electrification under stochastic conditions

In this article, we address the question of electric bus planning and operation under stochastic travel time and energy consumption. Uncertainties in the environment may cause disruptions to the planning and operation of electric buses, and a transportation planner must anticipate such conditions and be able to respond appropriately. One of the preconditions for planning robust strategies is understanding the existence and impact of multiple trade-off scenarios, which is the basis for this study. We model the travel time delays and trip energy consumption using estimated probability density functions and use a stochastic, mixed-integer formulation with chance constraints to evaluate several trade-off scenarios for electric bus fleets under uncertainty. The results show the existence of trade-off scenarios that lead to varying degrees of impacts related to network and environment. Careful fleet planning, dispatch, and charge control enable us to make the balance between these trade-offs and achieve better operational performances under uncertainty.

Traffic Signal Control With Adaptive Online-Learning Scheme Using Multiple-Model Neural Networks.

This article proposes a new traffic signal control algorithm to deal with unknown-traffic-system uncertainties and reduce delays in vehicle travel time. Unknown-traffic-system dynamics are approximated using a recurrent neural network (NN). To accurately identify the traffic system model, an online-learning scheme is developed to switch among a set of candidate NNs (i.e., multiple-model NNs) based on their estimation errors. Then, a bank of optimal signal-timing controllers is designed based on the online identification of the traffic system. Simulation studies have been carried out for the obtained control strategies using multiple-model NNs, and the desired results have been obtained. Moreover, compared with the widely used actuated traffic signal control schemes, it is shown that the proposed method can reduce vehicle travel delays and improve traffic system robustness.

On the Blackbody Spectrum of Kilonovae

The early spectra of the kilonova AT2017gfo have a remarkably smooth blackbody continuum, which reveals information on the thermal properties and radioactive heating within the ejecta. However, the widespread use of a single-temperature blackbody to fit kilonova data is theoretically invalid, because (1) the significant travel-time delays for a rapidly cooling surface result in a broad distribution of temperatures and (2) the relativistic Doppler correction varies across different surface elements. Thus, the observed spectrum should be a modified blackbody with a range of temperatures over the surface. In this paper we quantify the impact of these effects and illustrate the typical wavelength-dependent spectral corrections. We apply the multitemperature blackbody framework to the first-epoch X-shooter AT2017gfo spectrum, to deconvolve the underlying physical temperature at the photosphere from the relativistic Doppler shift. We show that the cooling and Doppler effects individually result in a variation of temperatures over the photosphere of up to 30%, but in combination these effects nearly cancel and produce the single-temperature blackbody observed. Finally, we show that fitting the UV, optical, or near-infrared separately yields blackbody temperatures consistent at the percent level, which puts stringent limits on any proposed modification of the spectral continuum.

X-Ray/UVOIR Frequency-resolved Time Lag Analysis of Mrk 335 Reveals Accretion Disk Reprocessing

UV and optical continuum reverberation mapping is a powerful tool for probing the accretion disk and inner broad-line region. However, recent reverberation mapping campaigns in the X-ray, UV, and optical have found lags consistently longer than those expected from the standard disk reprocessing picture. The largest discrepancy to date was recently reported in Mrk 335, where UV/optical lags are up to 12 times longer than expected. Here, we perform a frequency-resolved time lag analysis of Mrk 335, using Gaussian processes to account for irregular sampling. For the first time, we compare the Fourier frequency-resolved lags directly to those computed using the popular interpolated cross-correlation function method applied to both the original and detrended light curves. We show that the anticipated disk reverberation lags are recovered by the Fourier lags when zeroing in on the short-timescale variability. This suggests that a separate variability component is present on long timescales. If this separate component is modeled as reverberation from another region beyond the accretion disk, we constrain a size scale of roughly 15 lt-days from the central black hole. This is consistent with the size of the broad-line region inferred from Hβ reverberation lags. We also find tentative evidence for a soft X-ray lag, which we propose may be due to light travel time delays between the hard X-ray corona and distant photoionized gas that dominates the soft X-ray spectrum below 2 keV.

Discovering Causes of Traffic Congestion via Deep Transfer Clustering

Traffic congestion incurs long delay in travel time, which seriously affects our daily travel experiences. Exploring why traffic congestion occurs is significantly important to effectively address the problem of traffic congestion and improve user experience. Traditional approaches to mine the congestion causes depend on human efforts, which is time consuming and cost-intensive. Hence, we aim at discovering the known and unknown causes of traffic congestion in a systematic way. However, to achieve it, there are three challenges: (1) traffic congestion is affected by several factors with complex spatio-temporal relations; (2) there are a few samples of congestion data with known causes due to the limitation of human label; (3) more unknown congestion causes are unexplored since several factors contribute to traffic congestion. To address above challenges, we design a congestion cause discovery system consisting of two modules: (1) congestion feature extraction module, which extracts the important features distinguishing between different causes of congestion; and (2) congestion cause discovery module, which designs a deep semi-supervised learning based framework to discover the causes of traffic congestion with limited labeled data. Specifically, in pre-training stage, it first leverages a few labeled data as prior knowledge to pre-train the model. Then, in clustering stage, we propose two different clustering methods to discover the congestion causes. For the first clustering method, we extend the classic deep embedded clustering model to produce clusters via soft assignment. For the second one, we iteratively use k -means to group the latent features extracted from the pre-trained model, and use the cluster results as pseudo-labels to fine-tune the network. Extensive experiments show that the performance of our methods is superior to the state-of-the-art baselines, which demonstrates the effectiveness of the proposed cause discovery system. Additionally, our system is deployed and used in the practical production environment at Amap.