Browsing by Author "Huang, Cheng"

Now showing 1 - 7 of 7
  • Thumbnail Image
    ItemOpen Access
    Distributed trajectory management for urban air mobility operations with ground-based edge intelligence
    (IEEE, 2023-11-10) Huang, Cheng; Petrunin, Ivan; Tsourdos, Antonios
    Trajectory management is a critical undertaking in urban air mobility (UAM) to ensure safe, secure, and efficient operations. Cooperative targets have the capability to report their information while managing non-cooperative targets presents a challenge in the UAM operational environment (UOE). Consequently, ground-based non-cooperative surveillance assumes a vital role in monitoring anomalies. Given the difficulties associated with implementing centralized management in a large metropolitan area, this study proposes a distributed management architecture that leverages ground-based edge intelligence to enhance resilience in performing relevant tasks. It demonstrates that employing a developed edge computing system yields superior efficiency for heterogeneous sensors and their corresponding algorithms, such as detection, fusion, and tactical conflict management, compared to typical cloud servers. Furthermore, the proposed architecture incorporates an adaptive load balancing scheme, which monitors the real-time tasks and balances tasks among multiple edge devices to enhance the efficient resource management of the edge intelligence system. Ultimately, the distributed system offers energy-saving benefits and guarantees performance, making it suitable for providing services to diverse stakeholders involved in UAM.
  • Thumbnail Image
    ItemOpen Access
    Enhancing object detection and localization through multi-sensor fusion for smart city infrastructure
    (IEEE, 2024-06-26) Syamal, Soujanya; Huang, Cheng; Petrunin, Ivan
    The rapid advancement in autonomous systems and smart city infrastructure demands sophisticated object detection and localization capabilities to ensure safety, efficiency, and reliability. Traditional single sensor approaches often fall short, especially under complex environmental conditions. This paper introduces the CLR-Localiser, a novel multi-sensor fusion framework that synergistically integrates data from cameras, LiDAR and radar sensors mounted on roadside infrastructure to enhance object detection and 3D localization. Leveraging the complementary strengths of each sensor type, the CLR-Localiser employs an early fusion approach and deep learning techniques, including convolutional neural networks for object detection and regression networks for precise localization. We rigorously validated the performance of the CLR-Localiser against the benchmark Kitti dataset, and a custom dataset specifically designed for this research, demonstrating significant improvements in detection accuracy, localization precision, and object-tracking capabilities under diverse conditions. Our findings highlight the CLR-Localiser's potential to overcome the limitations of conventional monocular and single-sensor methods, offering a robust solution for autonomous driving, robotics, surveillance, and industrial automation applications. The development and validation of the CLR-Localiser not only prove the technical feasibility of early sensor data fusion but also pave the way for future advancements in multi-sensor fusion technology for enhanced environmental perception in autonomous systems.
  • Thumbnail Image
    ItemOpen Access
    Integrated frameworks of unsupervised, supervised and reinforcement learning for solving air traffic flow management problem
    (IEEE, 2021-11-15) Huang, Cheng; Xu, Yan
    This paper studies the demand-capacity balancing (DCB) problem in air traffic flow management (ATFM) with collaborative multi-agent reinforcement learning (MARL). To attempt the proper ground delay for resolving airspace hotspots, a multi-agent asynchronous advantage actor-critic (MAA3C) framework is firstly constructed with the long short-term memory network (LSTM) for the observations, in which the number of agents varies across training steps. The unsupervised learning and supervised learning are then introduced for better collaboration and learning among the agents. Experimental results demonstrate the scalability and generalization of the proposed frameworks, by means of applying the trained models to resolve different simulated and real-world DCB scenarios, with various flights number, sectors number and capacity settings.
  • Thumbnail Image
    ItemOpen Access
    Object detection for ground-based non-cooperative surveillance in urban air mobility utilizing lidar-camera fusion
    (AIAA, 2022-01-19) Huang, Cheng; Petrunin, Ivan; Tsourdos, Antonios
    Public safety and security are critical components in the Concept of Operations (ConOps) for Urban Air Mobility (UAM). The potential flight conflicts posed to air and ground objects need to be assessed, especially near critical regions and infrastructures, e.g. vertiports. In this sense, all targets, whether cooperative or non-cooperative air and ground targets, should be detected and tracked for conflict and risk assessment. To achieve this goal, ground-based non-cooperative sensors like cameras and lidar are utilized for situational awareness in this paper. In addition, a multi-modal dataset that contains both air and ground objects is constructed in different illumination and foggy weather scenarios. Finally, a lidar-camera fusion framework with multi-resolution voxelization and depth map learning is proposed for data-driven object detection. Experiments on the constructed dataset show the failure of existing lidar-based backbones in learning extremely sparse points, as a comparison, the fusion framework is outstanding in distinguishing air and ground objects, meanwhile, enabling resilient detection in various lighting and clearance conditions.
  • Thumbnail Image
    ItemOpen Access
    Radar-camera fusion for ground-based perception of small UAV in urban air mobility
    (IEEE, 2023-07-27) Huang, Cheng; Petrunin, Ivan; Tsourdos, Antonios
    The resilient surveillance of cooperative and non-cooperative aerial targets is critical for the safety and security of urban air mobility operations. Accurate detection, tracking, and trajectory prediction are essential to the subsequent tasks, e.g. tactical conflict prediction and resolution. Meanwhile, the combination of radar and camera is a classic option to provide perception services in different challenging environments. In this paper, a deep semantic association network is proposed for building relationships between the image detections and raw radar points, which then contributes to subsequent tasks, e.g. detecting, tracking, and predicting the small UAV with networked radar and camera systems. Various flight trials are conducted for collecting multi-sensor data, finally, training and testing results on this dataset demonstrate the outstanding performance of the proposed fusion workflow in comparison to single-sensor performance. At the same time, the 2D predictions in the sensor network are reconstructed to 3D trajectories for comparison and also reveal the improvements of the radar-camera fusion approach.
  • Thumbnail Image
    ItemOpen Access
    Strategic conflict management for performance-based urban air mobility operations with multi-agent reinforcement learning
    (IEEE, 2022-07-26) Huang, Cheng; Petrunin, Ivan; Tsourdos, Antonios
    With the urban air mobility (UAM) quickly evolving, the great demand for public airborne transit and deliveries, besides creating a big market, will result in a series of technical, operational, and safety problems. This paper addresses the strategic conflict issue in low-altitude UAM operations with multi-agent reinforcement learning (MARL). Considering the difference in flight characteristics, the aircraft performance is fully integrated into the design process of strategic deconfliction components. With this concept, the multi-resolution structure for the low-altitude airspace organization, Gaussian Mixture Model (GMM) for the speed profile generation, and dynamic separation minima enable efficient UAM operations. To resolve the demand and capacity balancing (DCB) issue and the separation conflict at the strategic stage, the multi-agent asynchronous advantage actor-critic (MAA3C) framework is built with mask recurrent neural networks (RNNs). Meanwhile, variable agent number, dynamic environments, heterogeneous aircraft performance, and action selection between speed adjustment and ground delay can be well handled. Experiments conducted on a developed prototype and various scenarios indicate the obvious advantages of the constructed MAA3C in minimizing the delay cost and refining speed profiles. And the effectiveness, scalability, and stabilization of the MARL solution are ultimately demonstrated.
  • Thumbnail Image
    ItemOpen Access
    Strategic conflict management using recurrent multi-agent reinforcement learning for urban air mobility operations considering uncertainties
    (Springer, 2023-01-26) Huang, Cheng; Petrunin, Ivan; Tsourdos, Antonios
    The rapidly evolving urban air mobility (UAM) develops the heavy demand for public air transport tasks and poses great challenges to safe and efficient operation in low-altitude urban airspace. In this paper, the operation conflict is managed in the strategic phase with multi-agent reinforcement learning (MARL) in dynamic environments. To enable efficient operation, the aircraft flight performance is integrated into the process of multi-resolution airspace design, trajectory generation, conflict management, and MARL learning. The demand and capacity balancing (DCB) issue, separation conflict, and block unavailability introduced by wind turbulence are resolved by the proposed the multi-agent asynchronous advantage actor-critic (MAA3C) framework, in which the recurrent actor-critic networks allow the automatic action selection between ground delay, speed adjustment, and flight cancellation. The learned parameters in MAA3C are replaced with random values to compare the performance of trained models. Simulated training and test experiments performed on a small urban prototype and various combined use cases suggest the superiority of the MAA3C solution in resolving conflicts with complicated wind fields. And the generalization, scalability, and stability of the model are also demonstrated while applying the model to complex environments.