Publications

[1] Title: Pedestrian Detection for Autonomous Cars: Occlusion Handling by Classifying Body Parts

Authors: Muhammad Mobaidul Islam, Abdullah Al Redwan Newaz, Balakrishna Gokaraju, and Ali Karimoddini

Publisher: Proc. of 2020 the IEEE SMC Conference.

https://ieeexplore.ieee.org/document/9282839

Abstract—In this work, we address the problem of detecting body parts of pedestrians using deep neural networks. In particular, we consider the occluded pedestrian detection problem in autonomous driving settings. While state-of-the-art deep neural models perform reasonably well for detecting full body pedestrians, their performances are not satisfactory for occluded pedestrians. Introducing a new training strategy along with a fusion mechanism, we enhance the performance of the SSD-Mobilenet and the Faster R-CNN by utilizing body parts information to handle occluded pedestrians. We evaluate our method by training these two deep neural networks using a public dataset as well as our dataset. The performance of the two developed models is compared both in terms of detection accuracy and runtime efficiency.

[2] Title: Eco-Vehicular Edge Networks for Connected Transportation: A Distributed Multi-Agent Reinforcement Learning Approach

Authors: Md Ferdous Pervej and Shih-Chun Lin

Publisher: 2020 IEEE 92nd Vehicular Technology Conference (VTC2020)

https://ieeexplore.ieee.org/document/9348507

Abstract—This paper introduces an energy-efficient, software defined vehicular edge network for the growing intelligent connected transportation system. A joint user-centric virtual cell formation and resource allocation problem is investigated to bring eco-solutions at the edge. This joint problem aims to combat against the power-hungry edge nodes while maintaining assured reliability and data rate. More specifically, by prioritizing the downlink communication of dynamic eco-routing, highly mobile autonomous vehicles are served with multiple low-powered access points (APs) simultaneously for ubiquitous connectivity and guaranteed reliability of the network. The formulated optimization is exceptionally troublesome to solve within a polynomial time, due to its complicated combinatorial structure. Hence, a distributed multi-agent reinforcement learning (D-MARL) algorithm is proposed for eco-vehicular edges, where multiple agents cooperatively learn to receive the best reward.

[3] Title: Wireless Networked Multi-Robot Systems in Smart Factories

Authors: Kwang-Cheng Chen, Shih-Chun Lin, Jen-Hao Hsiao, Chun-Hung Liu, Andreas F. Molisch, and Gerhard Fettweis

Publisher: Proceedings of the IEEE, 2020.

https://ieeexplore.ieee.org/document/9272626

Abstract—Smart manufacturing based on artificial intelligence and information communication technology will become the main contributor to the digital economy of the upcoming decades. In order to execute flexible production, smart manufacturing must holistically integrate wireless networking, computing, and automatic control technologies. This paper discusses the challenges of this complex system engineering, from a wireless networking perspective. Starting from enabling flexible re-configuration of a smart factory, we discuss existing wireless technology and the trends of wireless networking evolution to facilitate multi-robot smart factories. Furthermore, the special sequential decision-making of a multi-robot manufacturing system is examined. Social learning can be used to extend the resilience of precision operation in a multi-robot system by taking network topology into consideration, which also introduces a new vision for the cybersecurity of smart factories. A summary of highlights of technological opportunities for holistic facilitation of wireless networked multi-robot smart factories rounds off this paper.

[4] Title: CAV Impacts on Traffic Intersection Capacity

Authors: Li Song, and Wei Fan

Publisher: NC-CAV, Report No. TCE2020-03-001, January 2021.

https://www.nccav.com/resources/reports

Abstract—This report provides a comprehensive review of the current state-of-the-art and state-of-the-practice on connected and autonomous vehicle’s technology and its impacts on traffic intersection capacity. This should give a clear picture of connected and automated vehicle (CAV) technology, deployment and market penetration rate prediction of CAVs, traffic flow controls of CAVs, control strategies for CAVs at intersections, intersection capacity analysis methods, and intersection modeling scenarios and parameters used in the existing evaluation studies on the impacts of CAVs at intersections. It summarizes a comprehensive review of the current state-of-the-art and\ state-of-the-practice on studies related to the connected vehicles, automated vehicles, and connected and automated vehicles. Deployment and market penetration rate prediction, traffic flow control strategy for CAVs, intersection control strategy for CAVs, intersection capacity analysis methods (empirical-based and simulation-based methods), and intersection modeling scenarios and parameters of CAVs are introduced and reviewed. This part is intended to provide a fundamental and solid reference to develop control strategies for CAVs at the intersection and conduct effective simulation strategies and impact analysis for future tasks.

[5] Title: Impacts of Automated, Connected, Electric, and Shared Vehicles on Transportation Revenue Collection

Authors: Morgan Crowder, Steven Jiang, James Poslusny, Nicolas Norboge, Steve Bert, Daniel Findley

Publisher: NC-CAV, Report No. TCE2020-03-002, January 2021.

https://www.nccav.com/resources/reports/litsurveyimpacttransportationrevenue

Abstract—CAVs will provide numerous benefits when adopted, including reduced travel time, increased safety, and increased fuel efficiency. As CAVs continue to develop it will become important to act proactively to reduce the uncertainty around the adoption of these new systems. With increased fuel efficiency, sources of revenue such as the Motor Fuel Tax will decrease, which is especially a concern for North Carolina. Additionally, it may be necessary to revisit current statutes surrounding vehicles to accommodate this new technology. However, it should be possible to prepare for this changing landscape and adapt, as necessary. The adoption of CAVs will produce a wide variety of changes and have a large number of applications. Freight trucking will become more efficient, personal trips will be lengthened, and people who do not already have access to vehicle travel will be able to use vehicles. Public transportation will be able to reduce their overhead costs, as well as delivery services. This report will review the impacts of adopting CAVs on transportation revenue collection.

[6] Impacts of Connected Autonomous Vehicles on Traditional and Emerging Transportation Infrastructure, Literature Review

Authors: R. Thomas Chase (NCSU), Guangchuan Yang (NCSU), Ishtiak Ahmed (NCSU), Shoaib Samandar (NCSU), Shih-Chun Lin (NCSU), Chia-Hung Lin (NCSU), Chien-Yuan Wang (NCSU), John Kelly (NC A&T), and Mansi Bhavsar (NC A&T)

Publisher: NC-CAV, Report No. TCE2020-03-003, January 2021.

https://www.nccav.com/resources/reports/litsurveyimpacts-of-connected-and-autonomous-vehicles-on-traditional-and-em

Abstract: CAVs will require both the development of the new road infrastructures and the uplift in the existing road infrastructure and it is important analyze the readiness of the existing transportation infrastructure and maintenance programs to support CAV deployment and will investigate the emerging infrastructure required for the adoption of future CAV technologies. This review includes the state of the practice and state-of-the-art systems and technology to support CAVs as well as current and planned national research which NCDOT may use for planning and policy decisions in the future.

[7] Title: A C-V2X Platform Using Transportation Data and Spectrum Aware Slidelink Access.

Authors: C.-H. Lin, S.-C. Lin, C.-Y. Wang, and T. Chase

Publisher: Proceeding of 2021 IEEE International Conference on Systems, Man, And Cybernetics (SMC). https://ieeexplore.ieee.org/document/9659109

Abstract— Intelligent transportation systems and autonomous vehicles are expected to bring new experiences with enhanced efficiency and safety to road users in the near future. However, an efficient and robust vehicular communication system should act as a strong backbone to offer the needed infrastructure connectivity. Deep learning (DL)-based algorithms are widely adopted recently in various vehicular communication applications due to their achieved low latency and fast reconfiguration properties. Yet, collecting actual and sufficient transportation data to train DL-based vehicular communication models is costly and complex. This paper introduces a cellular vehicle-to-everything (C-V2X) verification platform based on an actual traffic simulator and spectrum-aware access. This integrated platform can generate realistic transportation and communication data, benefiting the development and adaptivity of DL-based solutions. Accordingly, vehicular spectrum recognition and management are further investigated to demonstrate the potentials of dynamic slidelink access. Numerical results show that our platform can effectively train and realize DL-based C-V2X algorithms. The developed slidelink communication can adopt different operating bands with remarkable spectrum detection performance, validating its practicality in real-world vehicular environments.

[8] Title: A Deep Learning Approach for Lane Detection.

Authors: Tesfamchael Getahun, Ali Karimoddini, and Priyantha Muda

Publisher: Proceeding of 24th IEEE International Conference on Intelligent Transportation Systems (ITSC 2021) https://ieeexplore.ieee.org/document/9564965

Abstract— Vision based lane detection is considered as one of the key components of advanced driver assistance systems (ADAS) and autonomous vehicle technologies. Majority of the approaches for lane detection systems rely on manually-tuned lane feature extraction methods which tend to under-perform when environmental conditions change. Alternatively, state-ofthe- art methods use variety of deep learning techniques for

lane feature extraction and prediction, demonstrating better performance than conventional lane detectors. However, deep learning approaches are computationally demanding and often fail to meet real-time requirements of autonomous vehicles. This paper proposes a lane detection system using a light weight convolutional neural network model as a feature extractor exploiting the potential of deep learning while meeting real-time needs. The model is trained using a small image patch of dimension 16 x 64 pixels and a non-overlapping sliding window approach is employed to achieve fast inference. Then, the predictions are clustered to get the lane boundaries. The proposed method is also integrated to the perception system of our autonomous vehicle and runs at 28 fps in CPU on image resolution of 768 x 1024 meeting real-time requirements needed for self-driving cars.

[9] Title: XGBoost: a Tree-based Approach for Traffic Volume Prediction

Authors: Benjamin Lartey, Abdollah Homaifar, Abenezer Girma, Ali Karimoddini and Daniel Opoku

Publisher: Proceeding of 2021 IEEE International Conference on Systems, Man, And Cybernetics (SMC). https://ieeexplore.ieee.org/document/9658959

Abstract— The growth in the transportation sector has led to an enormous increase in the number of vehicles which resulted in serious transportation issues including road congestion. Hence, estimating the number of vehicles on a road enables traffic managers to take appropriate decisions to curb congestion. In this paper, we propose to use an extreme gradient boosting (XGBoost) algorithm to efficiently and accurately predict the hourly traffic volume. We investigated the effectiveness of the proposed method for different scenarios including how well it performs during extreme weather conditions and holidays. We further investigated the effect of ridge and LASSO regularization on the performance of XGBoost. We then propose a new approach for setting the LASSO regularization parameter in terms of the number of observations and predictors. The performance and computational efficiency of the proposed approach is evaluated on data collected from Interstate-94, Minnesota and the result is compared with existing methods. Results show that the proposed method provides a good balance between performance and computational efficiency.

[10] Title: A Pedestrian Detection and Tracking Framework for Autonomous Cars: Efficient Fusion of Camera and LiDAR Data

Authors: Muhammad Mobaidul Islam, Abdullah Al Redwan Newaz, and Ali Karimoddini

Publisher: Proceeding of 2021 IEEE International Conference on Systems, Man, And Cybernetics (SMC). https://ieeexplore.ieee.org/document/9658639

Abstract— Pedestrian detection plays a critical role in autonomous driving and is an actively researched task in many different domains. To this end, pedestrian detection is considered either as a subclass of object detection or a part of semantic segmentation. There have been several methods with impressive object detection and semantics segmentation performances. Our hypothesis is that we can take advantage of capabilities of these individual methods and improve their performances, when they are properly fused. We propose a fusion method that combines anchor-based prediction with high-level semantic information. Unlike traditional pedestrian detectors that rely on homogeneous prediction models, our method provides the joint prediction with asymmetric multiple networks by fusing pair-wise pixel relevant information. Besides, parallel implementation of our framework entails low runtime overhead. We extensively evaluate the efficiency and robustness of the proposed fusion method on our autonomous driving dataset. Results show our fusion method improves detection rate more than an order of magnitude while achieving competitive runtime efficiency.

[11] Title: A Computationally Effective Pedestrian Detection using Constrained Fusion with Body Parts for Autonomous Driving

Authors: Muhammad Mobaidul Islam, Abdullah Al Redwan Newaz, Renran Tian, Abdollah Homaifar, Ali Karimoddini

Publisher: Proceeding of 2021 IEEE International Conference on Robotic Computing.

https://ieeexplore.ieee.org/abstract/document/9699953

Abstract— In this work, we address the problem of detecting body parts of pedestrians using the Single Shot Detection (SSD) method. In particular, we consider the occluded pedestrian detection problem in autonomous driving scenarios where the balance performance between accuracy and speed is crucial. While SSD performs reasonably well for the full-body pedestrian detection problem, its performance is not satisfactory for handling occlusion. To achieve a real-time performance along with robust detection, we introduce a simple yet effective body parts based single shot pedestrian detection architecture (BP-SSD). Augmenting the decision layer of BP-SSD with our proposed constraint optimization technique, we demonstrate that BP-SSD can accurately detect pedestrians. We evaluate our method using an autonomous driving dataset. Experimental results show that the proposed BP-SSD along with the proposed constraint optimization technique outperforms the generic SSD based pedestrian detection method.

[12] Title: SD-VEC: Distributed Computing Architectures for Ultra-Low Latency Connected Transportation

Authors: Shih-Chun Lin, Kwang-Cheng Chen, Ali Karimoddini, and Nicola Rohrseitz

Publisher: IEEE Communications Magazine.

https://ieeexplore.ieee.org/abstract/document/9681625

Abstract— New paradigm shifts and 6G technological revolution in vehicular services have emerged toward unmanned driving, automated transportation, and self-driving vehicles. As the technology for autonomous vehicles becomes mature, real challenges come from reliable, safe, real-time connected transportation operations to achieve ubiquitous and prompt information exchanges with massive connected and autonomous vehicles.

This article aims at introducing novel wireless distributed architectures that embed the edge computing capability inside software-defined vehicular networking infrastructure. Such edge networks consist of open-loop grant-free communications and computing-based control frameworks, which enable dynamic eco-routing with ultra-low latency and mobile data-driven orchestration. Thus, this work advances the frontiers of machine learning potentials and next-generation mobile system realization in vehicular networking applications.

[13] Title: TULVCAN: Terahertz Ultra-broadband Learning Vehicular Channel-Aware Networking

Authors: C.-H. Lin, S.-C. Lin, and E. Blasch

Publisher: Proc. of IEEE INFOCOM Workshop, Virtual Conference, May 2021.

https://ieeexplore.ieee.org/document/9484613

Abstract— Due to spectrum scarcity and increasing wireless capacity demands, terahertz (THz) communications at 0.1-10THz and the corresponding spectrum characterization have emerged to meet diverse service requirements in future 5G and 6G wireless systems. However, conventional compressed sensing techniques to

reconstruct the original wideband spectrum with under-sampled measurements become inefficient as local spectral correlation is deliberately omitted. Recent works extend communication methods with deep learning-based algorithms but lack strong ties to THz channel properties. This paper introduces novel THz channel-aware spectrum learning solutions that fully disclose the uniqueness of THz channels when performing such ultrabroadband sensing in vehicular environments. Specifically, a joint design of spectrum compression and reconstruction is proposed through a structured sensing matrix and two-phase reconstruction based on high spreading loss and molecular absorption at THz frequencies. An end-to-end learning framework, namely compression and reconstruction network (CRNet), is further developed with the mean-square-error loss function to improve sensing accuracy while significantly reducing computational complexity. Numerical results show that the CRNet solutions outperform the latest generative adversarial network (GAN) realization with a much higher cosine and structure similarity measures, smaller learning errors, and 56n% less required training overheads. This THz Ultra-broadband Learning Vehicular Channel-Aware

Networking (TULVCAN) work successfully achieves effective THz spectrum learning and hence allows frequency-agile access.

[14] Title: Dynamic Power Allocation and Virtual Cell Formation for Throughput-Optimal Vehicular Edge Networks in Highway Transportation

Authors: Md Ferdous Pervej, Shih-Chun Lin

Publisher: ICC Workshop 2020.

https://ieeexplore.ieee.org/document/9145348

Abstract— This paper investigates highly mobile vehicular networks from users' perspectives in highway transportation. Particularly, a centralized software-defined architecture is introduced in which centralized resources can be assigned, programmed, and controlled using the anchor nodes (ANs) of the edge servers. Unlike the legacy networks, where a typical user is served from only one access point (AP), in the proposed system model, a vehicle user is served from multiple APs simultaneously. While this increases the reliability and the spectral efficiency of the assisted users, it also necessitates an accurate power allocation in all transmission time slots. As such, a joint user association and power allocation problem is formulated to achieve enhanced reliability and weighted user sum rate. However, the formulated problem is a complex combinatorial problem, remarkably hard to solve. Therefore, fine-grained machine learning algorithms are used to efficiently optimize joint user associations and power allocations of the APs in a highly mobile vehicular network. Furthermore, a distributed single-agent reinforcement learning algorithm, namely SARL-MARL, is proposed which obtains nearly identical genie-aided optimal solutions within a nominal number of training episodes than the baseline solution. Simulation results validate that our solution outperforms existing schemes and can attain genie-aided optimal performances.

[15] Title: GCN-CNVPS: Novel Method for Cooperative Neighboring Vehicle Positioning System Based on Graph Convolution Network

Authors: Chia-Hung Lin, Yo-Hui Fang, Hsin-Yuan Chang, Yu-Chien Lin, Wei-Ho Chung, Shih-Chun Lin, and Ta-Sung Lee

Publisher: IEEE Access, vol. 9, pp. 153429-153441, 2021, doi: 10.1109/ACCESS.2021.3127914.

https://ieeexplore.ieee.org/document/9614197

Networking (TULVCAN) work successfully achieves effective THz spectrum learning and hence allows frequency-agile access.

[16] Title: Connected Autonomous Vehicles: State of Practice

Authors: Muhammad Mobaidul Islam, Abdullah Al Redwan Newaz, Li Song, Benjamin Lartey, Shih-Chun Lin, Wei Fan, Ali Hajbabaie, Mubbashar A. Khan, Thomas Chase, Alireza Partovi, Abdollah Homaifar, and Ali Karimoddini

Publisher: Submitted for publication

Abstract— Connected Autonomous Vehicles (CAVs) have the potential to deal with the steady increase in road traffic while solving transportation related issues such as traffic congestion, pollution, and road safety. Therefore, CAVs are becoming increasingly popular and viewed as the next generation transportation solution. Although modular advancements have been achieved in the development of CAVs, these efforts are not fully integrated to operationalize CAVs in realistic driving scenarios. This paper surveys a wide range of efforts reported in the literature, and summarizes the current state of practice in the field of CAVs in terms of autonomy technologies, communication backbone, and computation needs. Furthermore, this paper provides a general guidance on how transportation infrastructures need to be prepared in order to effectively operationalize CAVs. The paper also identifies challenges that need to be addressed in near future for effective and reliable adoption of CAVs.