Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8916960
Jun Lin, Stanley Y. P. Chien, Yaobin Chen, Chi-Chih Chen, Rini Sherony
Road departure mitigation system (RDMS) is a new vehicle active safety technology. Unlike Lane Departure Warning System (LDWS) and Lane Keeping Assistant System (LKAS), which relies on the lane marking to detect road edge, RDMS may not rely on the lane markings and can use the road edge and roadside objects to detect vehicle road departure. Since metal guardrail is a very common type of roadside boundary in the United States, RDMS may use metal guardrail as a reference to detect the road edge. Using real metal guardrails to test the performance of RDMS is difficult. One way to perform RDMS testing is to use a metal guardrail surrogate that has the similar properties to the real metal guardrail when sensed by the most common automotive sensors, such as radar, LIDAR, camera, etc., but will not damage the vehicle if being impacted. This paper describes the study of the 77GHz and 24GHz Radar Cross Section (RCS) of the real metal guardrail. The result will be used as the radar specifications for designing metal guardrail surrogate for the evaluation of RDMS.
{"title":"24 GHz and 77 GHz Radar Characteristics of Metal Guardrail for the Development of Metal Guardrail Surrogate for Road Departure Mitigation System Testing","authors":"Jun Lin, Stanley Y. P. Chien, Yaobin Chen, Chi-Chih Chen, Rini Sherony","doi":"10.1109/ITSC.2019.8916960","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8916960","url":null,"abstract":"Road departure mitigation system (RDMS) is a new vehicle active safety technology. Unlike Lane Departure Warning System (LDWS) and Lane Keeping Assistant System (LKAS), which relies on the lane marking to detect road edge, RDMS may not rely on the lane markings and can use the road edge and roadside objects to detect vehicle road departure. Since metal guardrail is a very common type of roadside boundary in the United States, RDMS may use metal guardrail as a reference to detect the road edge. Using real metal guardrails to test the performance of RDMS is difficult. One way to perform RDMS testing is to use a metal guardrail surrogate that has the similar properties to the real metal guardrail when sensed by the most common automotive sensors, such as radar, LIDAR, camera, etc., but will not damage the vehicle if being impacted. This paper describes the study of the 77GHz and 24GHz Radar Cross Section (RCS) of the real metal guardrail. The result will be used as the radar specifications for designing metal guardrail surrogate for the evaluation of RDMS.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"22 1","pages":"3340-3346"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77810824","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917357
S. Shafiei, M. Saberi, H. Vu
Time-dependent origin-destination (TDOD) demand estimation is often formulated as a bi-level quadratic optimization in which the estimated demand in the upper-level problem is evaluated iteratively through a dynamic traffic assignment (DTA) model in the lower level. When congestion forms and propagates in the network, traditional solutions assuming a linear relation between demand flow and link flow become inaccurate and yield biased solutions. In this study, we study a sensitivity-based method taking into account the impact of other OD flows on the links’ traffic volumes and densities. Thereafter, we compare the performance of the proposed method with several well-established solution methods for TDOD demand estimation problem. The methods are applied to a benchmark study urban network and a major freeway corridor in Melbourne, Australia. We show that the incorporation of traffic density into flow-based models improves the accuracy of the estimated OD flows and assist solution algorithm in avoiding converging to a sub-optimal result. Moreover, the final results obtained from the proposed sensitivity-based method contains less amount of error while the method exceeds the problem’s computational intensity compared to the traditional linear method.
{"title":"Nonlinearity in Time-Dependent Origin-Destination Demand Estimation in Congested Networks","authors":"S. Shafiei, M. Saberi, H. Vu","doi":"10.1109/ITSC.2019.8917357","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917357","url":null,"abstract":"Time-dependent origin-destination (TDOD) demand estimation is often formulated as a bi-level quadratic optimization in which the estimated demand in the upper-level problem is evaluated iteratively through a dynamic traffic assignment (DTA) model in the lower level. When congestion forms and propagates in the network, traditional solutions assuming a linear relation between demand flow and link flow become inaccurate and yield biased solutions. In this study, we study a sensitivity-based method taking into account the impact of other OD flows on the links’ traffic volumes and densities. Thereafter, we compare the performance of the proposed method with several well-established solution methods for TDOD demand estimation problem. The methods are applied to a benchmark study urban network and a major freeway corridor in Melbourne, Australia. We show that the incorporation of traffic density into flow-based models improves the accuracy of the estimated OD flows and assist solution algorithm in avoiding converging to a sub-optimal result. Moreover, the final results obtained from the proposed sensitivity-based method contains less amount of error while the method exceeds the problem’s computational intensity compared to the traditional linear method.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"1 1","pages":"3892-3897"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89439121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8916965
Dae Jung Kim, Jin Sung Kim, Seung-Hi Lee, C. Chung
In this paper, we present a comparative study of two machine learning methods to estimate the road surface condition without directly estimating tire-road friction coefficient. It is well known that using either a vehicle model-based approach or an end-to-end artificial intelligent method is not satisfactory to estimate the tire-road friction coefficient due to sensor noise, parameter uncertainty, and disturbances. To cope with this problem, three feature vectors obtained based on the vehicle dynamics are utilized for support vector machine (SVM) and deep neural network (DNN) with a time-window approach. The effectiveness of the proposed method is verified using experimental data obtained with a test vehicle on proving grounds. From the experimental study, we observed that the road surface condition estimation using DNN is superior to that using SVM.
{"title":"A Comparative Study of Estimating Road Surface Condition Using Support Vector Machine and Deep Neural Networ","authors":"Dae Jung Kim, Jin Sung Kim, Seung-Hi Lee, C. Chung","doi":"10.1109/ITSC.2019.8916965","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8916965","url":null,"abstract":"In this paper, we present a comparative study of two machine learning methods to estimate the road surface condition without directly estimating tire-road friction coefficient. It is well known that using either a vehicle model-based approach or an end-to-end artificial intelligent method is not satisfactory to estimate the tire-road friction coefficient due to sensor noise, parameter uncertainty, and disturbances. To cope with this problem, three feature vectors obtained based on the vehicle dynamics are utilized for support vector machine (SVM) and deep neural network (DNN) with a time-window approach. The effectiveness of the proposed method is verified using experimental data obtained with a test vehicle on proving grounds. From the experimental study, we observed that the road surface condition estimation using DNN is superior to that using SVM.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"2 1","pages":"1066-1071"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89870925","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917330
Hendrik Königshof, Niels Ole Salscheider, C. Stiller
We propose a 3D object detection and pose estimation method for automated driving using stereo images. In contrast to existing stereo-based approaches, we focus not only on cars, but on all types of road users and can ensure real-time capability through GPU implementation of the entire processing chain. These are essential conditions to exploit an algorithm for highly automated driving. Semantic information is provided by a deep convolutional neural network and used together with disparity and geometric constraints to recover accurate 3D bounding boxes. Experiments on the challenging KITTI 3D object detection benchmark show results that are within the range of the best image-based algorithms, while the runtime is only about a fifth. This makes our algorithm the first real-time image-based approach on KITTI.
{"title":"Realtime 3D Object Detection for Automated Driving Using Stereo Vision and Semantic Information","authors":"Hendrik Königshof, Niels Ole Salscheider, C. Stiller","doi":"10.1109/ITSC.2019.8917330","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917330","url":null,"abstract":"We propose a 3D object detection and pose estimation method for automated driving using stereo images. In contrast to existing stereo-based approaches, we focus not only on cars, but on all types of road users and can ensure real-time capability through GPU implementation of the entire processing chain. These are essential conditions to exploit an algorithm for highly automated driving. Semantic information is provided by a deep convolutional neural network and used together with disparity and geometric constraints to recover accurate 3D bounding boxes. Experiments on the challenging KITTI 3D object detection benchmark show results that are within the range of the best image-based algorithms, while the runtime is only about a fifth. This makes our algorithm the first real-time image-based approach on KITTI.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"40 11 1","pages":"1405-1410"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88062276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917087
C. Rapson, Boon-Chong Seet, M. Naeem, J. Lee, R. Klette
A vehicle’s braking lights can help to infer its future trajectory. Visible light communication using vehicle lights can also transmit other safety information to assist drivers with collision avoidance (whether the drivers be human or autonomous). Both these use cases require accurate localisation of vehicle lights by computer vision. Due to the large variation in lighting conditions (day, night, fog, snow, etc), the shape and brightness of the light itself, as well as difficulties with occlusions and perspectives, conventional methods are challenging and deep learning is a promising strategy. This paper presents a comparison of deep learning methods which are selected based on their potential to evaluate real-time video. The detection accuracy is shown to have a strong dependence on the size of the vehicle light within the image. A cascading approach is taken, where a downsampled image is used to detect vehicles, and then a second routine searches for vehicle lights at higher resolution within these Regions of Interest. This approach is demonstrated to improve detection, especially for small objects. Using YOLOv3 for the first stage and Tiny_YOLO for the second stage achieves satisfactory results across a wide range of conditions, and can execute at 37 frames per second. The ground truth for training and evaluating the methods is available for other researchers to use and compare their results.
{"title":"A Performance Comparison of Deep Learning Methods for Real-time Localisation of Vehicle Lights in Video Frames","authors":"C. Rapson, Boon-Chong Seet, M. Naeem, J. Lee, R. Klette","doi":"10.1109/ITSC.2019.8917087","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917087","url":null,"abstract":"A vehicle’s braking lights can help to infer its future trajectory. Visible light communication using vehicle lights can also transmit other safety information to assist drivers with collision avoidance (whether the drivers be human or autonomous). Both these use cases require accurate localisation of vehicle lights by computer vision. Due to the large variation in lighting conditions (day, night, fog, snow, etc), the shape and brightness of the light itself, as well as difficulties with occlusions and perspectives, conventional methods are challenging and deep learning is a promising strategy. This paper presents a comparison of deep learning methods which are selected based on their potential to evaluate real-time video. The detection accuracy is shown to have a strong dependence on the size of the vehicle light within the image. A cascading approach is taken, where a downsampled image is used to detect vehicles, and then a second routine searches for vehicle lights at higher resolution within these Regions of Interest. This approach is demonstrated to improve detection, especially for small objects. Using YOLOv3 for the first stage and Tiny_YOLO for the second stage achieves satisfactory results across a wide range of conditions, and can execute at 37 frames per second. The ground truth for training and evaluating the methods is available for other researchers to use and compare their results.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"153 1","pages":"567-572"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86637238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917322
Rashmi Munjal, William Liu, Xue Jun Li, Jairo Gutiérrez
With the explosive increase in the number of mobile devices such as smartphones or laptops, the design of mobile applications becomes increasingly complex, power hungry and resource consuming. Therefore, conventional networks are facing serious problems such as traffic overload and energy consumption due to high traffic demands. As a result, network designers are looking for more options to accommodate numerous data requirements. Aiming to find a promising way to tackle this problem, we are investigating heterogeneous networking architectures, which utilize the existing public transport network as an alternative communication network along with infrastructure-based networks. We propose a heterogeneous network architecture called Software Defined Connectivity (SDC) that utilizes the flow of transport network such as buses, trains, and ferries to start the forwarding process from nearby parking/offloading spots to disseminate data along with conventional networks. Results show that the SDC architecture helps in data offloading over public transport vehicles as per the profiles of each user with significant savings of energy.
{"title":"Big Data Offloading using Smart Public Vehicles with Software Defined Connectivity","authors":"Rashmi Munjal, William Liu, Xue Jun Li, Jairo Gutiérrez","doi":"10.1109/ITSC.2019.8917322","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917322","url":null,"abstract":"With the explosive increase in the number of mobile devices such as smartphones or laptops, the design of mobile applications becomes increasingly complex, power hungry and resource consuming. Therefore, conventional networks are facing serious problems such as traffic overload and energy consumption due to high traffic demands. As a result, network designers are looking for more options to accommodate numerous data requirements. Aiming to find a promising way to tackle this problem, we are investigating heterogeneous networking architectures, which utilize the existing public transport network as an alternative communication network along with infrastructure-based networks. We propose a heterogeneous network architecture called Software Defined Connectivity (SDC) that utilizes the flow of transport network such as buses, trains, and ferries to start the forwarding process from nearby parking/offloading spots to disseminate data along with conventional networks. Results show that the SDC architecture helps in data offloading over public transport vehicles as per the profiles of each user with significant savings of energy.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"24 1","pages":"3361-3366"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86375001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8916845
I. Bae, Jin Hyo Kim, J. Moon, Shiho Kim
Comfort driving has emerged as an important topic in the autonomous car research field. This study focuses on lane change maneuvering (LCM) of autonomous vehicles to provide a comfortable driving experience for passengers. For this purpose, we propose an LCM algorithm for determining a desired trajectory by evaluating the allowable lateral acceleration value obtained from Bezier curves at a local path planning stage for comfortable and smooth motion of the vehicle. The performance of the proposed LCM algorithm was verified through computer simulations and real driving tests.
{"title":"Lane Change Maneuver based on Bezier Curve providing Comfort Experience for Autonomous Vehicle Users","authors":"I. Bae, Jin Hyo Kim, J. Moon, Shiho Kim","doi":"10.1109/ITSC.2019.8916845","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8916845","url":null,"abstract":"Comfort driving has emerged as an important topic in the autonomous car research field. This study focuses on lane change maneuvering (LCM) of autonomous vehicles to provide a comfortable driving experience for passengers. For this purpose, we propose an LCM algorithm for determining a desired trajectory by evaluating the allowable lateral acceleration value obtained from Bezier curves at a local path planning stage for comfortable and smooth motion of the vehicle. The performance of the proposed LCM algorithm was verified through computer simulations and real driving tests.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"136 1","pages":"2272-2277"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88534627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917386
Yusuke Muramatsu, Yuki Tsuji, Alexander Carballo, S. Thompson, Hiroyuki Chishiro, Shinpei Kato
3D object detection is becoming increasingly significant for emerging autonomous vehicles. Safety decision making and motion planning depend highly on the result of 3D object detection. Recent 3D detection models are optimized for cars, cyclists and pedestrians with multiple models. This is not desirable because multiple models require significant resources, which are also used for other algorithms, such as localization or object tracking. We present SECOND-DX for providing multi-class support for 3D object detection with only a single model and it enables the detection of all three classes of 3D objects scanned using LiDAR sensors in real time. We conducted experiments involving the KITTI 3D object dataset to show that SECOND-DX is more accuracy overall evaluation metrics without compromising execution speed when compared with algorithms extended to support multi-class detection with a single model. Additionally, SECOND-DX can detect pedestrian classes comparable with that of current models that are optimized to support only cyclists and pedestrians.
{"title":"SECOND-DX: Single-model Multi-class Extension for Sparse 3D Object Detection","authors":"Yusuke Muramatsu, Yuki Tsuji, Alexander Carballo, S. Thompson, Hiroyuki Chishiro, Shinpei Kato","doi":"10.1109/ITSC.2019.8917386","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917386","url":null,"abstract":"3D object detection is becoming increasingly significant for emerging autonomous vehicles. Safety decision making and motion planning depend highly on the result of 3D object detection. Recent 3D detection models are optimized for cars, cyclists and pedestrians with multiple models. This is not desirable because multiple models require significant resources, which are also used for other algorithms, such as localization or object tracking. We present SECOND-DX for providing multi-class support for 3D object detection with only a single model and it enables the detection of all three classes of 3D objects scanned using LiDAR sensors in real time. We conducted experiments involving the KITTI 3D object dataset to show that SECOND-DX is more accuracy overall evaluation metrics without compromising execution speed when compared with algorithms extended to support multi-class detection with a single model. Additionally, SECOND-DX can detect pedestrian classes comparable with that of current models that are optimized to support only cyclists and pedestrians.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"3 1","pages":"2675-2680"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88649732","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8916957
Ismael Estalayo, E. Osaba, I. Laña, J. Ser
Within the current technological landscape sketched out by Intelligent Transport Systems (ITS), traffic flow prediction and route planning are two of the cornerstones on which the scientific community has been focused for years. Applications leveraging advances in these fields range from individual mobility planning to the establishment of optimal delivery routes, with doubtless benefits yielded to an immense strata of society. Intuitively, combining both prediction and route planning in a single, robust system could boost even further their paramount importance within the ITS field. However, most approaches reported so far in literature develop route planning techniques relying on actual traffic data (current or past observations) rather than on future traffic estimations, which could reliably represent the traffic flow status while the route is being performed. Unfortunately, research efforts around the monolithic hybridization of traffic prediction and route planning are still scarce. This manuscript embraces this noted issue as its main motivation by proposing an advanced routing platform endowed with a Long Short-Term Memory (LSTM) model for traffic forecasting purposes. The predictive output of this model serves as the input to a route planner, which constructs optimal green routes minimizing not only the total travel time, but also the CO2 emissions of the vehicle. The system has been tested using Open Trip Planner and real data collected over the city of Århus (Denmark), from which three different types of routes have been built and analyzed along a selection of predictive time horizons. The obtained results are promising and underscore the need for considering traffic predictions along the route for an improved usability of current route planning frameworks.
{"title":"Deep Recurrent Neural Networks and Optimization Meta-Heuristics for Green Urban Route Planning with Dynamic Traffic Estimates","authors":"Ismael Estalayo, E. Osaba, I. Laña, J. Ser","doi":"10.1109/ITSC.2019.8916957","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8916957","url":null,"abstract":"Within the current technological landscape sketched out by Intelligent Transport Systems (ITS), traffic flow prediction and route planning are two of the cornerstones on which the scientific community has been focused for years. Applications leveraging advances in these fields range from individual mobility planning to the establishment of optimal delivery routes, with doubtless benefits yielded to an immense strata of society. Intuitively, combining both prediction and route planning in a single, robust system could boost even further their paramount importance within the ITS field. However, most approaches reported so far in literature develop route planning techniques relying on actual traffic data (current or past observations) rather than on future traffic estimations, which could reliably represent the traffic flow status while the route is being performed. Unfortunately, research efforts around the monolithic hybridization of traffic prediction and route planning are still scarce. This manuscript embraces this noted issue as its main motivation by proposing an advanced routing platform endowed with a Long Short-Term Memory (LSTM) model for traffic forecasting purposes. The predictive output of this model serves as the input to a route planner, which constructs optimal green routes minimizing not only the total travel time, but also the CO2 emissions of the vehicle. The system has been tested using Open Trip Planner and real data collected over the city of Århus (Denmark), from which three different types of routes have been built and analyzed along a selection of predictive time horizons. The obtained results are promising and underscore the need for considering traffic predictions along the route for an improved usability of current route planning frameworks.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"10 1","pages":"1336-1342"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83617593","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917180
Lingbin Ning, Yidong Li, Min Zhou, Haifeng Song, Hai-rong Dong
Train timetable rescheduling (TTR) aims to address the recovery of train operation order in reordering and retiming strategies during disturbances. Considering this problem, this paper introduces a deep reinforcement learning (DRL) approach to minimize the average total delay for all trains along the railway line. Specifically, the detailed train operation in block sections and stations is illustrated to establish a learning environment involving its state sets, action sets, and the reward function. The learning agent is responsible for adjusting running times, dwell times and departure sequences for trains and conflicts are resolved simultaneously. Numerical experiments are performed on an adapted timetable carried out on the Beijing-Shanghai high-speed railway line. The experimental results indicate that the proposed approach reduces the average total delay by 46.38% in real time, compared to the First-Come-First-Served (FCFS) method.
{"title":"A Deep Reinforcement Learning Approach to High-speed Train Timetable Rescheduling under Disturbances","authors":"Lingbin Ning, Yidong Li, Min Zhou, Haifeng Song, Hai-rong Dong","doi":"10.1109/ITSC.2019.8917180","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917180","url":null,"abstract":"Train timetable rescheduling (TTR) aims to address the recovery of train operation order in reordering and retiming strategies during disturbances. Considering this problem, this paper introduces a deep reinforcement learning (DRL) approach to minimize the average total delay for all trains along the railway line. Specifically, the detailed train operation in block sections and stations is illustrated to establish a learning environment involving its state sets, action sets, and the reward function. The learning agent is responsible for adjusting running times, dwell times and departure sequences for trains and conflicts are resolved simultaneously. Numerical experiments are performed on an adapted timetable carried out on the Beijing-Shanghai high-speed railway line. The experimental results indicate that the proposed approach reduces the average total delay by 46.38% in real time, compared to the First-Come-First-Served (FCFS) method.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"17 1","pages":"3469-3474"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78920184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：