26 results found
Yu L, Guo F, Sivakumar A, et al., 2023, Few-Shot traffic prediction based on transferring prior knowledge from local network, Transportmetrica B: Transport Dynamics, Vol: 11, Pages: 1664-1686, ISSN: 2168-0566
Short-term traffic prediction has been widely studied in the community of Intelligent Transport Systems for decades. Despite the advances in machine learning-based prediction techniques, a challenging problem that affects the applications of such methods in practice is the prevalence of insufficient data across an entire road network. To address this few-shot traffic prediction problem at a local network scale, we develop a hybrid framework in conjunction with the prior knowledge transferring algorithm and two widely used models, i.e. Long-short Term Memory and Spatial–Temporal Graph Convolutional Neural Network. The proposed modelling framework is trained and tested using five-minute interval traffic flow data collected from London under different few-shot learning scenarios. Results show that transferring local network prior knowledge can improve the accuracy of both one-step prediction and multi-step prediction under inadequate data conditions, regardless of the deep-learning tool used.
Wang H, Pawlak J, Faghih Imani A, et al., 2023, When does it pay off to use electricity demand data with rich information about households and their activities? A comparative machine learning approach to demand modelling, Energy and Buildings, Vol: 295, Pages: 1-15, ISSN: 0378-7788
Energy demand modelling has been widely applied in various contexts, including power plant generation, building energy simulation and demand-side management. However, it is still an ongoing research topic in terms of the choice of modelling method, feature engineering for data-driven methods, the application contexts and the type of data used. In the residential sector, survey-based and meter-based approaches are categorised according to the type of input data used, i.e. the activity records from the time use survey and energy consumption from meters respectively. These two paradigms are not necessarily easy to combine, which warrants the questions of when one may be preferred over the other and whether they need to be combined despite the significant data requirements. Other details also have a huge impact on the data structure and performance of the energy demand model, including the choice of influential factors, the historical time window of factors selected, the split between training and test data, and the choice of machine learning (ML) algorithm. There is a lack of comparative research to guide researchers and practitioners in developing energy demand modelling capability, specifically as it pertains to these issues. This study analyses three groups of test scenarios in a multi-household residential context based in the UK. Six ML algorithms (LightGBM, Random forest, ANN, SVM, KNN and LSTM), with eight sets of various influential features, at four different historical time window widths and two train-test splits were compared. An appropriate methodology was designed to capture the temporal impact of activities on energy demand and represent the overlap and interaction of activities. The results show that the combination of meter-based and survey-based energy demand models performs better in terms of modelling accuracy and robustness against sudden load variation. Particularly, integrating energy tariffs, household and individual attributes, appliance usage a
Gao H, Liu K, Wang J, et al., 2023, Modular Bus Unit Scheduling for an Autonomous Transit System under Range and Charging Constraints, Applied Sciences (Switzerland), Vol: 13
Recent advances in vehicle technology offer new opportunities for an electric, automated, modular bus (MB) unit with an adjustable capacity to be applied to transit systems, promising to tackle the resource allocation challenges of traditional buses in coping with uneven travel demand. Drawing on the concept of modular vehicles, this paper introduces a novel scheduling system in which MB units can be combined/separated from fulfilling imbalanced trip demands through capacity adjustments. We develop an optimization model for determining the optimal formation and trip sequence of MB units. In particular, given that the vehicles are electrically powered, battery range limits and charging plans are considered in the system scheduling process. A column-generation-based heuristic algorithm is designed to efficiently solve this model, with constraints related to travel demand and charging station capacity incorporated into the master problem and the trip sequence for modular units with limited energy solved by the subproblem. Taking real data from transit operations for numerical examples, the proposed model performs well in terms of both algorithmic performance and practical applications. The generated optimal MB dispatching scheme can significantly reduce the operating cost from $1534.31 to $1144.26, a decrease of approximately 25% compared to conventional electric buses. The sensitivity analysis on the MB dispatch cost and battery capacity provides some insights for both the scenario configuration and the battery selection for MB system implementation.
Zhai X, Guo F, Krishnan R, 2023, An Online Optimal Bus Signal Priority Strategy to Equalise Headway in Real-Time, INFORMATION, Vol: 14
Li T, Guo F, Krishnan R, et al., 2022, An analysis of the value of optimal routing and signal timing control strategy with connected autonomous vehicles, Journal of Intelligent Transportation Systems, Pages: 1-15, ISSN: 1547-2450
With the emergence of connected and automated technologies, Connected Autonomous Vehicles (CAVs) are able to communicate and interact with other vehicles and signal controllers. Vehicle to Vehicle (V2V) and Vehicle to Infrastructure (V2I) communications open up an opportunity to improve routing and signal timing efficiency with additional information from CAVs, such as prior travel time and signal green time. Most of the existing research on routing and signal timing for Human Driven Vehicles (HDVs) has to face the fact that human drivers only have partial knowledge about travel costs and traffic status on the road network, which typically reduces the system efficiency. In this paper, the impacts of additional information from CAVs on routing and signal timing efficiency in terms of total travel time have been investigated. An Optimal Routing and Signal Timing (ORST) control strategy for CAVs has been proposed and compared with four existing routing and signal timing strategies where drivers have different levels of information. The results of the simulation demonstrate that with additional information from CAVs, ORST can reduce about 49% of the total travel time compared with Stochastic User Equilibrium (SUE) and about 10% of the total travel time compared with User Equilibrium (UE).
Li J, Xie N, Zhang K, et al., 2022, Network-scale traffic prediction via knowledge transfer and regional MFD analysis, TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES, Vol: 141, ISSN: 0968-090X
Li J, Guo F, Sivakumar A, et al., 2021, Transferability Improvement in Short-term Traffic Prediction using Stacked LSTM Network, Transportation Research Part C: Emerging Technologies, Vol: 124, ISSN: 0968-090X
Short-term traffic flow forecasting is a key element in Intelligent Transport Systems (ITS) to provide proactive traffic state information to road network operators. A variety of methods to predict traffic variables in the short-term can be found in the literature, ranging from time-series algorithms, machine learning tools and deep learning methods to a selective hybrid of these approaches. Despite the advances in prediction techniques, a challenging problem that affects the application of such methods in the real world is the prevalence of insufficient data across an entire network. It is rare that extensive historical training data required for model training are available for all the links in a city. In order to address this data insufficiency problem, this paper applies transfer learning techniques to machine learning methods in short-term traffic prediction. All the traffic data used in this paper were collected from Highways England road networks in the UK. The results show that through improving the transferability of machine learning-based models, the computational burden due to the model training process can be significantly reduced and the prediction accuracy under data deficient scenarios can be improved for one-step ahead prediction. However, the prediction accuracy gradually decreases in multi-step ahead prediction. It is also found that the accuracy of the proposed hybrid method is highly dependent upon consistency between datasets but less dependent on geographical attributes of links.
Li J, Zhang K, Shen L, et al., 2021, A Domain Adaptation Framework for Short-term Traffic Prediction, IEEE Intelligent Transportation Systems Conference (ITSC), Publisher: IEEE, Pages: 3564-3569, ISSN: 2153-0009
Li J, Guo F, Wang Y, et al., 2020, Short-term traffic prediction with deep neural networks and adaptive transfer learning, 23rd International Conference on Intelligent Transportation Systems (ITSC), Publisher: IEEE, Pages: 1-6
A key problem in short-term traffic prediction is the prevailing data missing scenarios across the entire traffic network. To address this challenge, a transfer learning framework is currently used in the literature, which could improve the prediction accuracy on the target link that suffers severe data missing problems by using information from source links with sufficient historical data. However, one of the limitations in these transfer-learning based models is their high dependency on the consistency between datasets and the complex data selection process, which brings heavy computation burden and human efforts. In this paper, we propose an adaptive transfer learning method in short-term traffic flow prediction model to alleviate the complex data selection process. Specifically, a self-adaptive neural network with a novel domain adaptation loss is developed. The domain adaptation loss is able to calculate the distance between the source data and the corresponding target data in each training batch, which can help the network to adaptively filter inconsistent source data and learn target link related information in each training batch. The Maximum Mean Discrepancy (MMD) measurement, which has been fully validated and applied in transfer learning research, is used in combination with the Gaussian kernel to measure the distance between datasets in each training batch. A series of experiments are designed and conducted using 15-minute interval traffic flow data from the Highways England, UK. The results have demonstrated that the proposed adaptive transfer learning method is less affected by the inconsistency between datasets and provides more accurate short-term traffic flow prediction.
Li T, Guo F, Krishnan R, et al., 2020, Right-of-way reallocation for mixed flow of autonomous vehicles and human driven vehicles, Transportation Research Part C: Emerging Technologies, Vol: 115, ISSN: 0968-090X
Autonomous Vehicles (AVs) are bringing challenges and opportunities to urban traffic systems. One of the crucial challenges for traffic managers and local authorities is to understand the nonlinear change in road capacity with increasing AV penetration rate, and to efficiently reallocate the Right-of-Way (RoW) for the mixed flow of AVs and Human Driven Vehicles (HDVs). Most of the existing research suggests that road capacity will significantly increase at high AV penetration rates or an all-AV scenario, when AVs are able to drive with smaller headways to the leading vehicle. However, this increase in road capacity might not be significant at a lower AV penetration rate due to the heterogeneity between AVs and HDVs. In order to investigate the impacts of mixed flow conditions (AVs and HDVs), this paper firstly proposes a theoretical model to demonstrate that road capacity can be increased with proper RoW reallocation. Secondly, four different RoW reallocation strategies are compared using a SUMO simulation to cross-validate the results in a numerical analysis. A range of scenarios with different AV penetration rates and traffic demands are used. The results show that road capacity on a two-lane road can be significantly improved with appropriate RoW reallocation strategies at low or medium AV penetration rates, compared with the do-nothing RoW strategy.
Dong Y, Guo F, Sivakumar A, et al., 2020, Short-term traffic prediction under disruptions using deep learning, TRAFFIC INFORMATION AND CONTROL, Editors: Li, He, Publisher: INST ENGINEERING TECH-IET, Pages: 79-114, ISBN: 978-1-83953-025-8
Wu Z, Guo F, Polak J, et al., 2019, Evaluating grid-interactive electric bus operation and demand response with load management tariff, Applied Energy, Vol: 255, Pages: 1-12, ISSN: 0306-2619
Electric Vehicles are expected to play a vital role in the transition of smart energy systems. Lots of recent research has explored numerous underlying mechanisms to achieve the synergetic interactions in the electricity balancing process. In this paper, the grid-interactive operation of electric buses is first time integrated within a dynamic market frame using the Distribution Locational Marginal Price algorithm for load congestion management. Since the defined problem correlates the opportunity charging flexibility with the bus mobility over a network, the tempo-spatial distribution of energy needs can be reflected in the dynamic of service planning. The interactions between bus operators and suppliers are quantitatively modelled by a bi-level optimisation process to represent the electric bus service planning and electricity market clearing separately. The effectiveness of the proposed load management has been demonstrated using data collected from an integrated real-world bus network. Experiments show that engagement of electric bus charging load in demand response is helpful to alleviate the network congestion and to reduce the power loss by 7.2% in the distribution network. However, alleviated charging loads have exhibited counter-intuitive ability for load shifting. The restricted electric bus operational requirements leads to a 8.17% loss of charging demand, while the reliance on large batteries has increased by 10.57%. However, the sensitivity analysis also shows that as the battery cost declines, the such discourage implications on grid-interactive electric bus operation will decrease once the battery cost below 190/kWh. The optimal grid-ebus integration have to consider the trade-off between range add-up, reduced battery cost and additional benefits.
Zhu L, Krishnan R, Guo F, et al., 2019, Early identification of recurrent congestion in heterogeneous urban traffic, IEEE Intelligent Transportation Systems Conference - ITSC 2019, Publisher: IEEE, Pages: 1-6
Urban traffic congestion has become a criticalissue that not only affects the quality of daily lives but alsoharms the environment and economy. Traffic patterns arerecurrent in nature, so is congestion. However, little attentionhas been paid to the development of methods that wouldenable early warning of the formation of congestion and itspropagation. This paper proposes a method for automatedearly congestion detection operating over time horizons rangingfrom half an hour to three hours. The method uses a deeplearning technique, Convolutional Neural Networks (CNN), andadapts it to the specific context of urban roads. Empiricalresults are reported from a busy traffic corridor in the city ofBath. Comprehensive evaluation metrics, including DetectionRate, False Positive Rate and Mean Time to Detection, areused to evaluate the performance of the proposed methodcompared to more conventional machine learning methodsincluding Feed-forward Neural Network and Random Forest.The results indicate that recurrent congestion can indeed bepredicted before it occurs and demonstrates that CNN basedmethod offers superior detection accuracy compared to theconventional machine learning methods in this context.
Zhu L, Krishnan R, Sivakumar A, et al., 2019, Traffic monitoring and anomaly detection based on simulation of Luxembourg Road network, IEEE Intelligent Transportation Systems Conference - ITSC 2019, Publisher: IEEE, Pages: 1-6
Traffic incidents which commonly result fromtraffic accidents, anomalous construction events and inclementweather can cause a wide range of negative impacts on urbanroad networks. Developing a high efficiency and transferabletraffic incident detection system plays an important role insolving the imbalance caused by traffic incidents betweentraffic demand and capacity. However, the existing literatureon transferability of traffic incident detection is rather limited.The objective of this paper is to provide an accurateand transferable incident detection approach based on therelationship between traffic variables and observed trafficincidents, in particular at a network level. We propose a deeplearning based method which has been calibrated using partof the collected traffic variables and the pre-assigned trafficincidents and then tested against the rest of the dataset. Theproposed method is compared to other benchmarks commonlyused in traffic incident detection, in terms of detection rate, falsepositive rate, f-measurement and detection time. The resultsindicate that the proposed method is significantly promising fortraffic incident detection with high accuracy and transferabilitycompared to the more widely used techniques in the literature.
Dong Y, Polak J, Sivakumar A, et al., 2019, Disaggregate short-term location prediction based on recurrent neural network and an agent-based platform, Transportation Research Record: Journal of the Transportation Research Board, Vol: 2673, Pages: 657-668, ISSN: 0361-1981
With the growing popularity of mobile and sensory devices, there has been a strong research interest in short-term disaggregate-level location prediction. Such predictive models have huge application potential in several sectors to change and improve people’s daily life and experience. Existing methods in this research stream have mainly focused on the prediction of sequence of location, with valuable temporal information overlooked. In addition, data limitations have constrained the development and understanding from different algorithms. In this paper, the authors propose a recurrent neural network-based method (RNN and LSTM, long short-term memory) for the next and future location prediction. This model predicts the sequence in time, thus it can predict both when and where an individual will be in the future and the duration of the stay at each location. The predictive model is developed based on an agent-based simulation platform that can produce realistic spatial-temporal trajectory data at the individual level. Analysis of the simulated data has shown that RNN and LSTM are capable of predicting future locations with better results than other comparative methods, especially for agents with high location variability. Online prediction with true location information fed into the model later in the day would greatly improve the predicted results. However, significant variations can be observed at the zonal level, with all methods performing much better on frequently visited locations than less visited locations or irregular visits.
Li T, Guo F, Krishnamoorthy R, et al., 2019, Right-of-Way Reallocation for Mixed Flow of Autonomous Vehicles and Human Driven Vehicles, 51st Annual Conference of the Universities-Transport-Study-Group (UTSG)
Zhu L, Guo F, Polak J, et al., 2018, Urban link travel time estimation using traffic states based data fusion, IET Intelligent Transport Systems, Vol: 12, Pages: 651-663, ISSN: 1751-956X
Estimated travel time is a key input for many intelligent transport systems (ITS) applications and traffic management functions. There are numerous studies that show that fusing data from different sources such as global positioning system (GPS), Bluetooth, mobile phone network (MPN), and inductive loop detector (ILD) can result in more accurate travel time estimation. However, to date, there has been little research investigating the contribution of individual data sources to the quality of the final estimate or how this varies according to source-specific data quality under different traffic states. Here, three different data sources, namely bus-based GPS (bGPS) data, ILD data, and MPN data, of varying quality are combined using three different data fusion techniques of varying complexity. In order to quantify the accuracy of travel time estimation, travel time calculated using automatic number plate recognition (ANPR) data are used as the `ground truth'. The final results indicate that fusing multiple data together does not necessarily enhance the accuracy of travel time estimation. The results also show that even in dense urban areas, bGPS data, when combined with ILD data, can provide reasonable travel time estimates of general traffic stream under different traffic states.
Guo F, Polak JW, Krishnamoorthy R, 2018, Predictor fusion for short-term traffic forecasting, Transportation Research Part C: Emerging Technologies, Vol: 92, Pages: 90-100, ISSN: 0968-090X
Zhu L, Guo F, Krishnamoorthy R, et al., 2018, The Use of Convolutional Neural Networks for Traffic Incident Detection at a Network Level, Transportation Research Board 97th Annual Meeting
Zhu L, Guo F, Krishnan R, et al., 2018, A Deep Learning Approach for Traffic Incident Detection in Urban Networks, 21st IEEE International Conference on Intelligent Transportation Systems (ITSC), Publisher: IEEE, Pages: 1011-1016, ISSN: 2153-0009
Zhu L, Guo F, Krishnamoorthy R, et al., 2017, Automated Early Detection of Congestion on Urban Roads: A Deep Learning Approach, 50th Annual Conference of the Universities-Transport-Study-Group (UTSG)
Guo F, Krishnan R, Polak JW, 2017, The influence of alternative data smoothing prediction techniques on the performance of a two-stage short-term urban travel time prediction framework, Journal of Intelligent Transportation Systems: Technology, Planning, and Operations, Vol: 21, Pages: 214-226, ISSN: 1547-2450
This article investigates the impact of alternative data smoothing and traffic prediction methods on the accuracy of the performance of a two-stage short-term urban travel time prediction framework. Using this framework, we test the influence of the combination of two different data smoothing and four different prediction methods using travel time data from two substantially different urban traffic environments and under both normal and abnormal conditions. This constitutes the most comprehensive empirical evaluation of the joint influence of smoothing and predictor choice to date. The results indicate that the use of data smoothing improves prediction accuracy regardless of the prediction method used and that this is true in different traffic environments and during both normal and abnormal (incident) conditions. Moreover, the use of data smoothing in general has a much greater influence on prediction performance than the choice of specific prediction method, and this is independent of the specific smoothing method used. In normal traffic conditions, the different prediction methods produce broadly similar results but under abnormal conditions, lazy learning methods emerge as superior.
Zhu L, Guo F, Polak J, et al., 2017, Multi-sensor Fusion Based on the Data from Bus GPS, Mobile Phone and Loop Detectors in Travel Time Estimation, Transportation Research Board 96th Annual Meeting
Guo F, Krishnan R, Polak J, 2013, A computationally efficient two-stage method for short-term traffic prediction on urban roads, TRANSPORTATION PLANNING AND TECHNOLOGY, Vol: 36, Pages: 62-75, ISSN: 0308-1060
Guo F, Krishnan R, Polak JW, 2012, Short-term traffic prediction under normal and incident conditions using singular spectrum analysis and the k-nearest neighbour method
Short-term traffic prediction is an important area in Intelligent Transport Systems (ITS) research. A number of ITS applications such as Advanced Traveller Information Systems (ATIS), Dynamic Route Guidance (DRG) and Urban Traffic Control (UTC) can benefit from improved prediction of traffic variables for the short-term future. Traffic prediction during abnormal condition, such as incidents, is especially important to these applications. However, this is an area not well-researched. This paper presents a novel improvement to a k-Nearest Neighbour (kNN) based traffic predictor with Singular Spectrum Analysis (SSA) technique based data preprocessing. This SSA-kNN framework is implemented for short-term traffic prediction under both normal and incident traffic conditions. A key feature of this approach is the data pre-processing step, which is designed to accommodate the extremely noisy sensor inputs that arise during incident conditions. This paper compares the prediction accuracy of the SSA-kNN approach with three other commonly used machine learning methods, kNN, Grey System Model (GM) and Support Vector Regression (SVR). Moreover, the sensitivity of traffic prediction accuracy to various kNN design parameters is explored. The results show that the proposed SSA-kNN based approach has the best prediction accuracy among the methods used in this study, especially during non-recurring incidents. The concept behind the proposed method can be extended to other machine learning tools to improve the accuracy of short-term traffic forecasting models.
Guo F, Polak JW, Krishnan R, 2010, Comparison of modelling approaches for short term traffic prediction under normal and abnormal conditions, Pages: 1209-1214-1209-1214, ISSN: 2153-0009
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.