Pub Date : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00008
S. Ranka, A. Rangarajan, L. Elefteriadou, Sivaramnakrishnan Srinivasan, Emmanuel Poasadas, Dan Hoffman, Raj Ponnulari, Jeremy Dilmore, Tom Byron
This smart city traffic management approach seeks to use edge-based video-stream processing (using multicore and GPU processors) at intersections and in public vehicles (city buses, fire trucks, ambulances, school buses) to convert video data into space-time trajectories of individual vehicles and pedestrians that are transmitted to a cloud-based system. Key information is then synthesized in the cloud from them to create a real-time city-wide traffic palette. Real-time or offline processing both at the edge and the cloud will then be leveraged to optimize intersection operations, manage network traffic, identify near-collisions between various units of traffic, provide street parking information, and a host of other applications. Additional information such as weather and environment will also be leveraged. The use of edge-based real-time machine learning (ML) techniques and videostream processing has several significant advantages. (1) Because there is no need to store copious amounts of video (few minutes typically suffice for edge-based processing), it automatically addresses concerns of public agencies who do not want person-identifiable information to be stored for reasons of citizen privacy and legality. (2) The processing of the video stream at the edge will allow for the use of low bandwidth communication using wireline and wireless networks to a central system such as a cloud, resulting in a compressed and holistic picture of the entire city. (3) The real-time nature of processing enables a wide variety of novel transportation applications at the intersection, street, and system levels that were not possible hitherto, significantly impacting safety and mobility.
{"title":"A Vision of Smart Traffic Infrastructure for Traditional, Connected, and Autonomous Vehicles","authors":"S. Ranka, A. Rangarajan, L. Elefteriadou, Sivaramnakrishnan Srinivasan, Emmanuel Poasadas, Dan Hoffman, Raj Ponnulari, Jeremy Dilmore, Tom Byron","doi":"10.1109/MetroCAD48866.2020.00008","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00008","url":null,"abstract":"This smart city traffic management approach seeks to use edge-based video-stream processing (using multicore and GPU processors) at intersections and in public vehicles (city buses, fire trucks, ambulances, school buses) to convert video data into space-time trajectories of individual vehicles and pedestrians that are transmitted to a cloud-based system. Key information is then synthesized in the cloud from them to create a real-time city-wide traffic palette. Real-time or offline processing both at the edge and the cloud will then be leveraged to optimize intersection operations, manage network traffic, identify near-collisions between various units of traffic, provide street parking information, and a host of other applications. Additional information such as weather and environment will also be leveraged. The use of edge-based real-time machine learning (ML) techniques and videostream processing has several significant advantages. (1) Because there is no need to store copious amounts of video (few minutes typically suffice for edge-based processing), it automatically addresses concerns of public agencies who do not want person-identifiable information to be stored for reasons of citizen privacy and legality. (2) The processing of the video stream at the edge will allow for the use of low bandwidth communication using wireline and wireless networks to a central system such as a cloud, resulting in a compressed and holistic picture of the entire city. (3) The real-time nature of processing enables a wide variety of novel transportation applications at the intersection, street, and system levels that were not possible hitherto, significantly impacting safety and mobility.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122065727","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 : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00020
T. Horeis, T. Kain, Julian-Steffen Müller, F. Plinke, J. Heinrich, Maximilian Wesche, Hendrik Decke
The development of system architectures, fulfilling a diverse set of technical and economic requirements, is known to be a challenging task when designing a new vehicle. The demands particularly concerning the system’s reliability, availability, and safety, are, however, remarkably increasing when glancing towards full vehicle autonomy, since this level of automatization excludes any takeover actions by passengers. To satisfy the requirements, a fail-operational system design that includes several fallback paths is required. Since classical approaches, which require adding fallback paths with various redundant and segregated components, contradict the harsh cost constraints prevailing in the automotive sector, further use of those approaches is not desirable. Hence, various new concepts are developed to dissolve this contradiction, i.e., reducing the number of hardware and software components, while on the other hand, keeping the level of reliability high. The problem, though, is that the systems resulting from applying those concepts are highly complex and can not be sufficiently analyzed with today’s tools regarding the availability, safety, and reliability of the system. Therefore, in this paper, we introduce AT-CARS (Analyzing Tool for Complex, Autonomous, and Reliable Systems), a tool capable of analyzing various complex systems architectures designed for autonomous vehicles. Our tool aims to support system engineers responsible for determining suitable system architectures that fulfill the expected safety requirements while satisfying the monetary conditions by providing measurements concerning availability, safety, and reliability. Those parameters are determined by a state-based Monte Carlo simulation, which supports dynamic failure management procedures.
{"title":"A Reliability Engineering Based Approach to Model Complex and Dynamic Autonomous Systems","authors":"T. Horeis, T. Kain, Julian-Steffen Müller, F. Plinke, J. Heinrich, Maximilian Wesche, Hendrik Decke","doi":"10.1109/MetroCAD48866.2020.00020","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00020","url":null,"abstract":"The development of system architectures, fulfilling a diverse set of technical and economic requirements, is known to be a challenging task when designing a new vehicle. The demands particularly concerning the system’s reliability, availability, and safety, are, however, remarkably increasing when glancing towards full vehicle autonomy, since this level of automatization excludes any takeover actions by passengers. To satisfy the requirements, a fail-operational system design that includes several fallback paths is required. Since classical approaches, which require adding fallback paths with various redundant and segregated components, contradict the harsh cost constraints prevailing in the automotive sector, further use of those approaches is not desirable. Hence, various new concepts are developed to dissolve this contradiction, i.e., reducing the number of hardware and software components, while on the other hand, keeping the level of reliability high. The problem, though, is that the systems resulting from applying those concepts are highly complex and can not be sufficiently analyzed with today’s tools regarding the availability, safety, and reliability of the system. Therefore, in this paper, we introduce AT-CARS (Analyzing Tool for Complex, Autonomous, and Reliable Systems), a tool capable of analyzing various complex systems architectures designed for autonomous vehicles. Our tool aims to support system engineers responsible for determining suitable system architectures that fulfill the expected safety requirements while satisfying the monetary conditions by providing measurements concerning availability, safety, and reliability. Those parameters are determined by a state-based Monte Carlo simulation, which supports dynamic failure management procedures.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131171789","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 : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00011
Zhinan Qiao, Andrew Sansom, M. McGuire, Andrew Kalaani, Xu Ma, Qing Yang, Song Fu
Recently, more and more attention has been paid to the connected object detection for better performance. One of the most interesting fields is learning from multiple resources in a connected fashion. In this paper, we present a connected object detection method using multiple cameras for the smart transportation system. The proposed architecture consists of three parts: an alignment framework, a deep multi-view fusion network and an object detection network. Experiments are conducted to illustrate the performance of our proposed architecture.
{"title":"Accurate Object Detection in Smart Transportation Using Multiple Cameras","authors":"Zhinan Qiao, Andrew Sansom, M. McGuire, Andrew Kalaani, Xu Ma, Qing Yang, Song Fu","doi":"10.1109/MetroCAD48866.2020.00011","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00011","url":null,"abstract":"Recently, more and more attention has been paid to the connected object detection for better performance. One of the most interesting fields is learning from multiple resources in a connected fashion. In this paper, we present a connected object detection method using multiple cameras for the smart transportation system. The proposed architecture consists of three parts: an alignment framework, a deep multi-view fusion network and an object detection network. Experiments are conducted to illustrate the performance of our proposed architecture.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115603654","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 : 2020-02-01DOI: 10.1109/metrocad48866.2020.00006
{"title":"MetroCAD 2020 Committees","authors":"","doi":"10.1109/metrocad48866.2020.00006","DOIUrl":"https://doi.org/10.1109/metrocad48866.2020.00006","url":null,"abstract":"","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"2014 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128021038","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 : 2020-02-01DOI: 10.1109/metrocad48866.2020.00003
{"title":"MetroCAD 2020 Breaker Page","authors":"","doi":"10.1109/metrocad48866.2020.00003","DOIUrl":"https://doi.org/10.1109/metrocad48866.2020.00003","url":null,"abstract":"","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"126 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115885675","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 : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00024
Yunyi Jia, B. Ayalew
Automated vehicles have immense potentials for improving the safety, efficiency and environmental problems in our existing transportation systems. Despite the tremendous ongoing efforts from both industry and academia, fully autonomous vehicles have not yet been widely deployed in public traffic. In the foreseeable future, automated vehicles will very likely be expected to operate in traffic that involve heterogeneous agents including automated vehicles, human-driven vehicles and pedestrians. Such heterogeneity will bring new challenges to the safety of the traffic system. This paper reviews some existing works related to heterogeneous traffic systems and presents a vision of cyber-human-physical heterogeneous traffic systems that can substantially enhance overall safety.
{"title":"Cyber-Human-Physical Heterogeneous Traffic Systems for Enhanced Safety","authors":"Yunyi Jia, B. Ayalew","doi":"10.1109/MetroCAD48866.2020.00024","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00024","url":null,"abstract":"Automated vehicles have immense potentials for improving the safety, efficiency and environmental problems in our existing transportation systems. Despite the tremendous ongoing efforts from both industry and academia, fully autonomous vehicles have not yet been widely deployed in public traffic. In the foreseeable future, automated vehicles will very likely be expected to operate in traffic that involve heterogeneous agents including automated vehicles, human-driven vehicles and pedestrians. Such heterogeneity will bring new challenges to the safety of the traffic system. This paper reviews some existing works related to heterogeneous traffic systems and presents a vision of cyber-human-physical heterogeneous traffic systems that can substantially enhance overall safety.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123725716","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 : 2020-02-01DOI: 10.1109/metrocad48866.2020.00005
{"title":"MetroCAD 2020 Opinion","authors":"","doi":"10.1109/metrocad48866.2020.00005","DOIUrl":"https://doi.org/10.1109/metrocad48866.2020.00005","url":null,"abstract":"","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115001733","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 : 2020-02-01DOI: 10.1109/metrocad48866.2020.00023
{"title":"MetroCAD 2020 Index","authors":"","doi":"10.1109/metrocad48866.2020.00023","DOIUrl":"https://doi.org/10.1109/metrocad48866.2020.00023","url":null,"abstract":"","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132107650","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 : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00022
Zhiliu Yang, Chen Liu
Existing global optimization mapping is prone to time-consuming parameter fine tuning of scan matching. In this work, we propose a novel map estimation method via cascading scan-to-map local matching with Deep Neural Network (DNN). Scan-to-map local matching firstly acts as a teacher to provide a coarse pose estimation, then a DNN is trained in an unsupervised learning fashion by exploiting the self-contradictory occupancy status of the point clouds. On the other hand, in order to cope with the mismatch problem caused by variable point number of a scan and fixed input size of DNN, a data hiding strategy is proposed. Experiments are conducted on three LiDAR datasets we collected from real-world scenarios. The visualization results of final maps demonstrate that our method, teacher-forcing unsupervised learning, is able to produce 2D occupancy map very close to the real world, which outperforms pure DeepMapping as well as ICP-warm-started DeepMapping. We further demonstrated that our results are comparable with those from traditional Bundle Adjustment (BA) method, without the need for parameter fine tuning.
{"title":"2D Map Estimation via Teacher-Forcing Unsupervised Learning","authors":"Zhiliu Yang, Chen Liu","doi":"10.1109/MetroCAD48866.2020.00022","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00022","url":null,"abstract":"Existing global optimization mapping is prone to time-consuming parameter fine tuning of scan matching. In this work, we propose a novel map estimation method via cascading scan-to-map local matching with Deep Neural Network (DNN). Scan-to-map local matching firstly acts as a teacher to provide a coarse pose estimation, then a DNN is trained in an unsupervised learning fashion by exploiting the self-contradictory occupancy status of the point clouds. On the other hand, in order to cope with the mismatch problem caused by variable point number of a scan and fixed input size of DNN, a data hiding strategy is proposed. Experiments are conducted on three LiDAR datasets we collected from real-world scenarios. The visualization results of final maps demonstrate that our method, teacher-forcing unsupervised learning, is able to produce 2D occupancy map very close to the real world, which outperforms pure DeepMapping as well as ICP-warm-started DeepMapping. We further demonstrated that our results are comparable with those from traditional Bundle Adjustment (BA) method, without the need for parameter fine tuning.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130108756","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 : 2020-02-01DOI: 10.1109/MetroCAD48866.2020.00019
Yongling Sun, Xiaosong Yang, Hai Xiao, H. Feng
with the fast-growing advancements in highly automated driving technologies, cost-effective evaluation and validation of functional modules and full-stack driving system have become great challenges for automakers prior to the release of new intelligent vehicle models. Simulation based on high-fidelity rendering and physics engine has been widely used as a powerful tool to develop self-driving systems. From an OEM perspective, a flexible, extensible, and scalable virtual platform is proposed to integrate individual functional modules, support typical scenario libraries, and evaluate single functional algorithm or an entire domain controller system for rapid algorithm iterations and high-efficiency system testing. The platform demonstrates that a stand-alone machine can support up to 32-channel simulations in parallel depending on system resources, a semi-automatic method is introduced to generate a couple of scenarios based on map data and standard road network format, and typical perception algorithms are visualized and evaluated. In addition, such evaluation system can be deployed to the cloud to support large-scale simulations and testing automation.
{"title":"An intelligent driving simulation platform: architecture, implementation and application","authors":"Yongling Sun, Xiaosong Yang, Hai Xiao, H. Feng","doi":"10.1109/MetroCAD48866.2020.00019","DOIUrl":"https://doi.org/10.1109/MetroCAD48866.2020.00019","url":null,"abstract":"with the fast-growing advancements in highly automated driving technologies, cost-effective evaluation and validation of functional modules and full-stack driving system have become great challenges for automakers prior to the release of new intelligent vehicle models. Simulation based on high-fidelity rendering and physics engine has been widely used as a powerful tool to develop self-driving systems. From an OEM perspective, a flexible, extensible, and scalable virtual platform is proposed to integrate individual functional modules, support typical scenario libraries, and evaluate single functional algorithm or an entire domain controller system for rapid algorithm iterations and high-efficiency system testing. The platform demonstrates that a stand-alone machine can support up to 32-channel simulations in parallel depending on system resources, a semi-automatic method is introduced to generate a couple of scenarios based on map data and standard road network format, and typical perception algorithms are visualized and evaluated. In addition, such evaluation system can be deployed to the cloud to support large-scale simulations and testing automation.","PeriodicalId":117440,"journal":{"name":"2020 International Conference on Connected and Autonomous Driving (MetroCAD)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123140151","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：