Pub Date : 2019-10-01DOI: 10.1109/ITSC.2019.8917475
Xin Nie, Meifang Yang, R. W. Liu
Deep learning-based object detection has recently received significant attention among scholars and practitioners. However, the acquired images often suffer from visual quality degradation under severe weather conditions, which could lead to negative effects on object detection in practical applications. Most previous studies proposed to implement object detection based on the assumption that image restoration techniques (e.g., image dehazing and low-light image enhancement, etc.) could improve visual quality while boosting detection accuracy. In contrast, we assumed that the image restoration techniques (i.e., image preprocessing) may also degrade the fine image details resulting in failing to promote object detection performance. In this work, according to the physical imaging process under severe weather conditions, we directly proposed to synthetically generate the degraded images with training labels to enlarge the original training datasets, which commonly contain only clear natural images under normal weather conditions. The advanced YOLOv3 model was then trained and tested on the enlarged dataset which contain both synthetic and realistic ship images generated under different weather conditions. Experiments have been conducted to compare the proposed method with other competing methods which implement training model only with clear images and testing model with (or without) image preprocessing. Results illustrated that our model could achieve superior detection performance under different conditions.
{"title":"Deep Neural Network-Based Robust Ship Detection Under Different Weather Conditions","authors":"Xin Nie, Meifang Yang, R. W. Liu","doi":"10.1109/ITSC.2019.8917475","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917475","url":null,"abstract":"Deep learning-based object detection has recently received significant attention among scholars and practitioners. However, the acquired images often suffer from visual quality degradation under severe weather conditions, which could lead to negative effects on object detection in practical applications. Most previous studies proposed to implement object detection based on the assumption that image restoration techniques (e.g., image dehazing and low-light image enhancement, etc.) could improve visual quality while boosting detection accuracy. In contrast, we assumed that the image restoration techniques (i.e., image preprocessing) may also degrade the fine image details resulting in failing to promote object detection performance. In this work, according to the physical imaging process under severe weather conditions, we directly proposed to synthetically generate the degraded images with training labels to enlarge the original training datasets, which commonly contain only clear natural images under normal weather conditions. The advanced YOLOv3 model was then trained and tested on the enlarged dataset which contain both synthetic and realistic ship images generated under different weather conditions. Experiments have been conducted to compare the proposed method with other competing methods which implement training model only with clear images and testing model with (or without) image preprocessing. Results illustrated that our model could achieve superior detection performance under different conditions.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"69 1","pages":"47-52"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81729844","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.2019.8916953
Songyi Zhang, Yu Chen, Shi-tao Chen, N. Zheng
With the progress of autonomous driving technology in recent years, motion planning has been an issue in the navigation of self-driving cars. To achieve an optimal path that meets the requirements of both smoothness and safety, vehicle kinematics and dynamics constraints should be considered. This paper proposes a novel motion planning method based on Hybrid A* for real-time and curvature-contentious path planning with local post smoothing in complex dynamic environments: (1)our method introduces parametric clothoid curves precomputed offline as basic motion primitives for rapid online planning; (2)the path obtained using our method is G2-continuous (i.e., curvature continuous) and does not have a considerable effect on the search time consumption, while also considering possible collisions and motion constraints of nonholonomic car-like vehicles; (3) the node re-expansion issue of conventional Hybrid A* is discussed and resolved by the proposed quintic spine-based local smoothing approach for complete path continuity. Hence, post smoothing and collision checking for the overall resulting path. Simulation and on-road tests have been performed to evaluate the efficiency of the proposed method. The method can be widely implemented in numerous complex scenarios.
随着近年来自动驾驶技术的进步，运动规划一直是自动驾驶汽车导航中的一个问题。为了获得满足平稳性和安全性要求的最优路径，需要考虑车辆的运动学和动力学约束。本文提出了一种基于Hybrid a *的运动规划方法，用于复杂动态环境下具有局部后平滑的实时曲率争议路径规划:(1)将离线预计算的参数化曲面曲线作为基本运动基元，实现快速在线规划;(2)该方法得到的路径是g2连续的(即曲率连续)，对搜索耗时影响不大，同时考虑了非完整类车可能发生的碰撞和运动约束;(3)讨论了传统Hybrid A*的节点再展开问题，并提出了基于五次棘的局部光滑完全路径连续性方法。因此，对整个结果路径进行后平滑和碰撞检查。仿真和道路试验验证了该方法的有效性。该方法可广泛应用于众多复杂场景。
{"title":"Hybrid A*-based Curvature Continuous Path Planning in Complex Dynamic Environments","authors":"Songyi Zhang, Yu Chen, Shi-tao Chen, N. Zheng","doi":"10.1109/ITSC.2019.8916953","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8916953","url":null,"abstract":"With the progress of autonomous driving technology in recent years, motion planning has been an issue in the navigation of self-driving cars. To achieve an optimal path that meets the requirements of both smoothness and safety, vehicle kinematics and dynamics constraints should be considered. This paper proposes a novel motion planning method based on Hybrid A* for real-time and curvature-contentious path planning with local post smoothing in complex dynamic environments: (1)our method introduces parametric clothoid curves precomputed offline as basic motion primitives for rapid online planning; (2)the path obtained using our method is G2-continuous (i.e., curvature continuous) and does not have a considerable effect on the search time consumption, while also considering possible collisions and motion constraints of nonholonomic car-like vehicles; (3) the node re-expansion issue of conventional Hybrid A* is discussed and resolved by the proposed quintic spine-based local smoothing approach for complete path continuity. Hence, post smoothing and collision checking for the overall resulting path. Simulation and on-road tests have been performed to evaluate the efficiency of the proposed method. The method can be widely implemented in numerous complex scenarios.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"18 1","pages":"1468-1474"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81821917","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.2019.8917082
Wenhao Wu, Bing Bu, Wei Zhang
Communication-based train control (CBTC) are automated train control systems using information communication technologies to ensure the safe operation of rail vehicles. With the development of information technology, massive commercial software, hardware products and standard communication equipment are applied to urban rail transit systems, which introduces a crowd of security threats to CBTC systems. This paper proposes a generalized stochastic Petri net model to simulate dynamic interaction between the attacker and defender to evaluate the security of CBTC systems. According to the characteristics of the system and attack-defense methods, we divide our model to the penetration phase and the disruption phase. In each phase, we provide effective means of attack and corresponding defensive measures, and the system state is determined correspondingly. The model parameters are obtained by conducting attack and defense exercises on the semi-physical simulation platform. The system transition probability is derived with the model parameter and the Nash equilibrium of the game between the attacker and defender. The system availability is obtained by calculating the steady probability of each state which can be derived from the GSPN model solution. Our analytic results reveal the seriousness of the system security situation and the significance of defensive measures for system security.
{"title":"Attacks and Counter Defense Mechanisms for CBTC Systems: System Modeling and Availability Analysis","authors":"Wenhao Wu, Bing Bu, Wei Zhang","doi":"10.1109/ITSC.2019.8917082","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917082","url":null,"abstract":"Communication-based train control (CBTC) are automated train control systems using information communication technologies to ensure the safe operation of rail vehicles. With the development of information technology, massive commercial software, hardware products and standard communication equipment are applied to urban rail transit systems, which introduces a crowd of security threats to CBTC systems. This paper proposes a generalized stochastic Petri net model to simulate dynamic interaction between the attacker and defender to evaluate the security of CBTC systems. According to the characteristics of the system and attack-defense methods, we divide our model to the penetration phase and the disruption phase. In each phase, we provide effective means of attack and corresponding defensive measures, and the system state is determined correspondingly. The model parameters are obtained by conducting attack and defense exercises on the semi-physical simulation platform. The system transition probability is derived with the model parameter and the Nash equilibrium of the game between the attacker and defender. The system availability is obtained by calculating the steady probability of each state which can be derived from the GSPN model solution. Our analytic results reveal the seriousness of the system security situation and the significance of defensive measures for system security.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"17 1","pages":"2521-2526"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81838443","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.2019.8917397
Takeshi Hirai, T. Murase
This paper presents the performance characteristics (notably, the adverse effects on communication quality) of Sensing-based Semi-Persistent Scheduling (Sensing-based SPS) of PC5-based Cellular-V2X mode 4 in crash warning application under congested environments. Sensing-based SPS is a new Cellular-V2X standard. This scheduling protocol has two interference estimation mechanisms to select a transmission slot considered as low interference from a history of slot usages. As the mode 4's channel becomes severely congested, frequent collision errors become to cause missing slot information and, the estimated interference becomes to differ from the actual one. Hence, the two mechanisms become to select a slot with higher interference. Our simulation results revealed that Sensing-based SPS was of 9% lower communication quality than that of a method with no estimation mechanisms in a congested situation. Our behavior analysis and computer simulations pointed out new disadvantages of Sensing-based SPS and presented a necessity to improve the algorithm of Sensing-based SPS.
{"title":"Performance Characteristics of Sensing-based SPS of PC5-based C-V2X Mode 4 in Crash Warning Application under Congestion","authors":"Takeshi Hirai, T. Murase","doi":"10.1109/ITSC.2019.8917397","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917397","url":null,"abstract":"This paper presents the performance characteristics (notably, the adverse effects on communication quality) of Sensing-based Semi-Persistent Scheduling (Sensing-based SPS) of PC5-based Cellular-V2X mode 4 in crash warning application under congested environments. Sensing-based SPS is a new Cellular-V2X standard. This scheduling protocol has two interference estimation mechanisms to select a transmission slot considered as low interference from a history of slot usages. As the mode 4's channel becomes severely congested, frequent collision errors become to cause missing slot information and, the estimated interference becomes to differ from the actual one. Hence, the two mechanisms become to select a slot with higher interference. Our simulation results revealed that Sensing-based SPS was of 9% lower communication quality than that of a method with no estimation mechanisms in a congested situation. Our behavior analysis and computer simulations pointed out new disadvantages of Sensing-based SPS and presented a necessity to improve the algorithm of Sensing-based SPS.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"18 1","pages":"189-194"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82643186","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.2019.8917454
Maleen Jayasuriya, G. Dissanayake, Ravindra Ranasinghe, N. Gandhi
This paper presents a low cost, resource efficient localisation approach for autonomous driving in GPS denied environments. One of the most challenging aspects of traditional landmark based localisation in the context of autonomous driving, is the necessity to accurately and frequently detect landmarks. We leverage the state of the art deep learning framework, YOLO (You Only Look Once), to carry out this important perceptual task using data obtained from monocular cameras. Extracted bearing only information from the YOLO framework, and vehicle odometry, is fused using an Extended Kalman Filter (EKF) to generate an estimate of the location of the autonomous vehicle, together with it’s associated uncertainty. This approach enables us to achieve real-time sub metre localisation accuracy, using only a sparse map of an outdoor urban environment. The broader motivation of this research is to improve the safety and reliability of Personal Mobility Devices (PMDs) through autonomous technology. Thus, all the ideas presented here are demonstrated using an instrumented mobility scooter platform.
本文提出了一种低成本、资源高效的GPS环境下自动驾驶定位方法。在自动驾驶背景下，传统的基于地标的定位最具挑战性的一个方面是，必须准确、频繁地检测地标。我们利用最先进的深度学习框架YOLO (You Only Look Once)，使用从单目摄像机获得的数据来执行这一重要的感知任务。使用扩展卡尔曼滤波(EKF)将从YOLO框架中提取的仅方位信息与车辆里程数融合，以生成自动驾驶车辆的位置估计值及其相关的不确定性。这种方法使我们能够实现实时亚米定位精度，仅使用室外城市环境的稀疏地图。本研究的更广泛动机是通过自主技术提高个人移动设备(PMDs)的安全性和可靠性。因此，这里提出的所有想法都是使用仪表移动滑板车平台进行演示的。
{"title":"Leveraging Deep Learning Based Object Detection for Localising Autonomous Personal Mobility Devices in Sparse Maps","authors":"Maleen Jayasuriya, G. Dissanayake, Ravindra Ranasinghe, N. Gandhi","doi":"10.1109/ITSC.2019.8917454","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917454","url":null,"abstract":"This paper presents a low cost, resource efficient localisation approach for autonomous driving in GPS denied environments. One of the most challenging aspects of traditional landmark based localisation in the context of autonomous driving, is the necessity to accurately and frequently detect landmarks. We leverage the state of the art deep learning framework, YOLO (You Only Look Once), to carry out this important perceptual task using data obtained from monocular cameras. Extracted bearing only information from the YOLO framework, and vehicle odometry, is fused using an Extended Kalman Filter (EKF) to generate an estimate of the location of the autonomous vehicle, together with it’s associated uncertainty. This approach enables us to achieve real-time sub metre localisation accuracy, using only a sparse map of an outdoor urban environment. The broader motivation of this research is to improve the safety and reliability of Personal Mobility Devices (PMDs) through autonomous technology. Thus, all the ideas presented here are demonstrated using an instrumented mobility scooter platform.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"31 1","pages":"4081-4086"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80506032","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.2019.8917381
A. Parra, A. Zubizarreta, Joshué Pérez
Intelligent Transportation Systems (ITS) is currently one of the most active research areas, being electric vehicles (EVs) and their vehicle dynamics enhancement key topics. For this purpose, the development of optimal Advanced Driver-Assistance Systems (ADAS) and Advanced Vehicle Dynamics Control Systems is required. However, as electrified propulsion systems offer multiple topologies (and higher complexity), this task becomes much more difficult. In this context, the use of intelligent control techniques has been proposed as a suitable alternative to offer both performance and flexibility.In order to demonstrate the advantages of intelligent approaches and their ability to adapt to different scenarios, this work presents a comparative study of the performance of Intelligent Control based torque vectoring (TV) algorithms in electric vehicles with three different powertrain topologies: Front Wheel Driven (FWD), Rear Wheel Driven (RWD) and Four/All Wheel Driven (AWD). The same TV approach has been used for all topologies, and a skid-pad test has been selected as a critical manoeuvre for evaluating the lateral dynamics of each topology, which has been simulated using a high fidelity vehicle simulator.Results show that the same intelligent control approach can be used for different topologies without retuning its parameters, enhancing the vehicle dynamics for all cases. This demonstrates the flexibility of intelligent approaches due to their reduced model dependency. Additionally, results show that each architecture promotes a different type of dynamic behaviour in the vehicle: understeering behaviour for the FWD, oversteering behaviour for the RWD and a neutral behaviour for the AWD.
{"title":"A comparative study of the effect of Intelligent Control based Torque Vectoring Systems on EVs with different powertrain architectures","authors":"A. Parra, A. Zubizarreta, Joshué Pérez","doi":"10.1109/ITSC.2019.8917381","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917381","url":null,"abstract":"Intelligent Transportation Systems (ITS) is currently one of the most active research areas, being electric vehicles (EVs) and their vehicle dynamics enhancement key topics. For this purpose, the development of optimal Advanced Driver-Assistance Systems (ADAS) and Advanced Vehicle Dynamics Control Systems is required. However, as electrified propulsion systems offer multiple topologies (and higher complexity), this task becomes much more difficult. In this context, the use of intelligent control techniques has been proposed as a suitable alternative to offer both performance and flexibility.In order to demonstrate the advantages of intelligent approaches and their ability to adapt to different scenarios, this work presents a comparative study of the performance of Intelligent Control based torque vectoring (TV) algorithms in electric vehicles with three different powertrain topologies: Front Wheel Driven (FWD), Rear Wheel Driven (RWD) and Four/All Wheel Driven (AWD). The same TV approach has been used for all topologies, and a skid-pad test has been selected as a critical manoeuvre for evaluating the lateral dynamics of each topology, which has been simulated using a high fidelity vehicle simulator.Results show that the same intelligent control approach can be used for different topologies without retuning its parameters, enhancing the vehicle dynamics for all cases. This demonstrates the flexibility of intelligent approaches due to their reduced model dependency. Additionally, results show that each architecture promotes a different type of dynamic behaviour in the vehicle: understeering behaviour for the FWD, oversteering behaviour for the RWD and a neutral behaviour for the AWD.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"92 1","pages":"480-485"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80310523","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.2019.8917130
Zebang Shen, Yichong Xu, Muchen Sun, Alexander Carballo, Qingguo Zhou
Due to multi-path effects, GNSS-based localization methods are not always reliable in urban transportation scenes. To solve this problem, matching-based methods, which compare the real-time sensor data with a prior map, are widely used for urban autonomous driving. In these methods, high-precision noise-free 3D map plays an essential role in vehicle localization. This paper proposes a 3D map optimization framework to generate such map with high efficiency and low memory consumption. First, a deep learning based method is designed to automatically filter out non-map objects in mobile laser scans during mapping. Then, a method named dynamic local NDT is introduced for mapping and localization to improve efficiency and reduce memory usage. Furthermore, a road segmentation method is exploited for further optimization. The proposed framework only relies on LIDAR and GNSS-INS, which makes it simple and easily conducted. The mapping and positioning experimental results show that the proposed framework outperforms the conventional NDT method.
{"title":"3D Map Optimization with Fully Convolutional Neural Network and Dynamic Local NDT","authors":"Zebang Shen, Yichong Xu, Muchen Sun, Alexander Carballo, Qingguo Zhou","doi":"10.1109/ITSC.2019.8917130","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917130","url":null,"abstract":"Due to multi-path effects, GNSS-based localization methods are not always reliable in urban transportation scenes. To solve this problem, matching-based methods, which compare the real-time sensor data with a prior map, are widely used for urban autonomous driving. In these methods, high-precision noise-free 3D map plays an essential role in vehicle localization. This paper proposes a 3D map optimization framework to generate such map with high efficiency and low memory consumption. First, a deep learning based method is designed to automatically filter out non-map objects in mobile laser scans during mapping. Then, a method named dynamic local NDT is introduced for mapping and localization to improve efficiency and reduce memory usage. Furthermore, a road segmentation method is exploited for further optimization. The proposed framework only relies on LIDAR and GNSS-INS, which makes it simple and easily conducted. The mapping and positioning experimental results show that the proposed framework outperforms the conventional NDT method.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"116 1","pages":"4404-4411"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80349050","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.2019.8917479
C. Shih, Pao-Wei Huang, E-Ton Yen, Pei-Kuei Tsung
Motion prediction is an essential feature for autonomous vehicle to understand the intention of nearby vehicles so as to plan its route. Several works have been proposed to predict the motion of nearby vehicles using Kalman filter, RNN, and other machine learning methods. However, many of them rely on either the communication among vehicles or the global information of all the vehicles on the road and have limited applicability. This paper presents the design and implementation of a new machine learning model to predicate the motion of nearby vehicles by observing their motions in last few seconds. The proposed network consists of encoderdecoder, LSTM, and attention model to tackle the challenges so as to predict the vehicle speed using limited observations. The network was trained based on data set collected on public roads. Comparing with Kalman filter, the developed method reduces the prediction error up to 50% and the prediction error is up to 6.5KPH(1.8 meter per second) under all evaluated scenarios, which is less than the tolerance of speedometers on vehicles.
运动预测是自动驾驶汽车了解附近车辆意图从而规划其路线的重要功能。已经提出了一些使用卡尔曼滤波、RNN和其他机器学习方法来预测附近车辆的运动的工作。然而，它们中的许多要么依赖于车辆之间的通信，要么依赖于道路上所有车辆的全局信息，适用性有限。本文提出了一种新的机器学习模型的设计和实现，该模型通过观察最近几秒钟的运动来预测附近车辆的运动。该网络由编码器-解码器、LSTM和注意力模型组成，以解决利用有限的观测值预测车辆速度的挑战。该网络是基于在公共道路上收集的数据集进行训练的。与卡尔曼滤波相比，所开发的方法将预测误差降低了50%，在所有评估情景下的预测误差均达到6.5KPH(1.8 m / s)，小于车辆上车速表的容差。
{"title":"Vehicle speed prediction with RNN and attention model under multiple scenarios","authors":"C. Shih, Pao-Wei Huang, E-Ton Yen, Pei-Kuei Tsung","doi":"10.1109/ITSC.2019.8917479","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917479","url":null,"abstract":"Motion prediction is an essential feature for autonomous vehicle to understand the intention of nearby vehicles so as to plan its route. Several works have been proposed to predict the motion of nearby vehicles using Kalman filter, RNN, and other machine learning methods. However, many of them rely on either the communication among vehicles or the global information of all the vehicles on the road and have limited applicability. This paper presents the design and implementation of a new machine learning model to predicate the motion of nearby vehicles by observing their motions in last few seconds. The proposed network consists of encoderdecoder, LSTM, and attention model to tackle the challenges so as to predict the vehicle speed using limited observations. The network was trained based on data set collected on public roads. Comparing with Kalman filter, the developed method reduces the prediction error up to 50% and the prediction error is up to 6.5KPH(1.8 meter per second) under all evaluated scenarios, which is less than the tolerance of speedometers on vehicles.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"34 1","pages":"369-375"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80389506","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.2019.8917014
David Oswald, G. Scora, Nigel Williams, Peng Hao, M. Barth
In addition to providing safety and mobility benefits, Connected and Automated Vehicles (CAVs) have the potential to reduce fuel consumption and emissions. As new CAV applications are developed, it is valuable to estimate these potential environmental benefits, typically using vehicle activity data and emissions models. To date, most researchers in the U.S. have used the MOVES vehicle emissions model, developed and maintained by the U.S. Environmental Protection Agency (EPA). However, because MOVES uses a binning approach, it is likely underestimating the true energy and emissions savings that occur when CAV applications smooth traffic flow. To illustrate this problem, we measure and model the fuel consumption and CO2 emissions for a real-world CAV application: Eco-Approach and Departure (EAD) at signalized intersections. Real-world measurements are compared to a MOVES-based estimate, as well as to an estimate provided by the physical-based Comprehensive Modal Emissions Model (CMEM). Results show that MOVES consistently underestimates the energy and emissions benefits of the CAV application, primarily since the bin sizes in MOVES are too large to catch the nuances of traffic smoothing. On the other hand, CMEM provided a more accurate energy and emissions estimate, primarily since it uses analytical functions to model emissions and does not suffer from the same binning problem.
{"title":"Evaluating the Environmental Impacts of Connected and Automated Vehicles: Potential Shortcomings of a Binned-Based Emissions Model","authors":"David Oswald, G. Scora, Nigel Williams, Peng Hao, M. Barth","doi":"10.1109/ITSC.2019.8917014","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917014","url":null,"abstract":"In addition to providing safety and mobility benefits, Connected and Automated Vehicles (CAVs) have the potential to reduce fuel consumption and emissions. As new CAV applications are developed, it is valuable to estimate these potential environmental benefits, typically using vehicle activity data and emissions models. To date, most researchers in the U.S. have used the MOVES vehicle emissions model, developed and maintained by the U.S. Environmental Protection Agency (EPA). However, because MOVES uses a binning approach, it is likely underestimating the true energy and emissions savings that occur when CAV applications smooth traffic flow. To illustrate this problem, we measure and model the fuel consumption and CO2 emissions for a real-world CAV application: Eco-Approach and Departure (EAD) at signalized intersections. Real-world measurements are compared to a MOVES-based estimate, as well as to an estimate provided by the physical-based Comprehensive Modal Emissions Model (CMEM). Results show that MOVES consistently underestimates the energy and emissions benefits of the CAV application, primarily since the bin sizes in MOVES are too large to catch the nuances of traffic smoothing. On the other hand, CMEM provided a more accurate energy and emissions estimate, primarily since it uses analytical functions to model emissions and does not suffer from the same binning problem.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"1 1","pages":"3639-3644"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83388538","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.2019.8917291
Khaled Saleh, A. Abobakr, D. Nahavandi, Julie Iskander, M. Attia, Mostafa Hossny, S. Nahavandi
One of the main barriers against the full deployment of autonomous vehicles in urban traffic environments is the understanding of the intentions and behaviours of the human around them. Moreover, understanding and predicting intentions of vulnerable road users such as cyclists is still one of the most challenging tasks. In this work, we are proposing a novel framework for the task of intent prediction of cyclists via hand signalling from point cloud scans. We utilised our developed data generation pipeline for generating synthetic point cloud scans of cyclists doing a set of hand signals in urban traffic environments. Then, we feed a sequence of the generated point cloud scans to our framework which jointly segments all cyclists instances and predicts their most probable intended actions in an end-to-end fashion. Our proposed framework has achieved superior results with 83% in F1-Measure score over the testing split of our generated dataset. Additionally, the proposed framework outperformed other compared baseline approaches with more than 39% improvement in F1-Measure score.
{"title":"Cyclist Intent Prediction using 3D LIDAR Sensors for Fully Automated Vehicles","authors":"Khaled Saleh, A. Abobakr, D. Nahavandi, Julie Iskander, M. Attia, Mostafa Hossny, S. Nahavandi","doi":"10.1109/ITSC.2019.8917291","DOIUrl":"https://doi.org/10.1109/ITSC.2019.8917291","url":null,"abstract":"One of the main barriers against the full deployment of autonomous vehicles in urban traffic environments is the understanding of the intentions and behaviours of the human around them. Moreover, understanding and predicting intentions of vulnerable road users such as cyclists is still one of the most challenging tasks. In this work, we are proposing a novel framework for the task of intent prediction of cyclists via hand signalling from point cloud scans. We utilised our developed data generation pipeline for generating synthetic point cloud scans of cyclists doing a set of hand signals in urban traffic environments. Then, we feed a sequence of the generated point cloud scans to our framework which jointly segments all cyclists instances and predicts their most probable intended actions in an end-to-end fashion. Our proposed framework has achieved superior results with 83% in F1-Measure score over the testing split of our generated dataset. Additionally, the proposed framework outperformed other compared baseline approaches with more than 39% improvement in F1-Measure score.","PeriodicalId":6717,"journal":{"name":"2019 IEEE Intelligent Transportation Systems Conference (ITSC)","volume":"12 1","pages":"2020-2026"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83398521","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：