Harnessing the capabilities of Unmanned Aerial Vehicles (UAVs) equipped with Light Detection and Ranging (LiDAR) sensors, this project aims to revolutionize real-time information gathering and sharing from the sky. The work spans intelligent transportation, aerial traffic monitoring, and advanced aerial networking, blending computational intelligence, 3D data analytics, and next-generation communication technologies to open new frontiers for smart cities and beyond.
Intelligent 3D Road Coverage
Optimizing the 3D placement of UAVs with integrated LiDAR technology presents a complex challenge, especially when aiming for maximal road coverage. The task is further complicated by constraints related to the LiDAR’s field-of-view, point-cloud density, and geographic and road segment priorities. To address these challenges, we formulated an optimization problem and employed a particle swarm optimization-based algorithm. This approach allows us to maximize road coverage efficiency, thereby enhancing the capabilities of Intelligent Transportation Systems.
Object Detection and Tracking from the Sky
The second part of the project addresses the lack of real-world aerial LiDAR data for traffic monitoring. By generating 3D point cloud data through UAV-mounted LiDAR sensors, the project enables object detection and tracking from the sky. To this end, a 3D simulator is used to generate data sets on which the PointVoxel-RCNN (PV-RCNN) algorithm is implemented for road user detection. An Unscented Kalman filter is also used to predict the state of detected objects, providing an 8% performance improvement over other 3D point cloud solutions.
High-Capacity Flying Network Infrastructure
The final part of the project tackles the challenge of enabling high-capacity data communication between flying UAVs. To achieve this, we integrate Flying Ad-hoc Networks (FANET) with mmWave technology. One of the significant challenges in implementing such a dynamic network is efficiently steering the antennas of the UAVs toward their peers for optimal communication. To quickly find the best network topology, we use a Generative Adversarial Network (GAN)-based algorithm specifically tailored for FANET topology optimization. This approach employs a WaveGAN model followed by a beam search. The model learns from a supervised dataset to generate optimized network topologies, while the beam search adjusts these topologies to meet the structural requirements of the mmWave-based FANET. Simulation results demonstrate the effectiveness of this approach, finding near-optimal FANET topologies quickly for different network sizes.
Leveraging Deep Learning for Enhanced Aerial Sensing
In crafting this multifaceted project, a blend of sophisticated algorithms ensures that each component operates at peak efficiency. The placement of UAVs with LiDAR for the most effective traffic data collection is elegantly solved using Particle Swarm Optimization (PSO). This algorithm maximizes road coverage while adhering to a range of constraints, from the LiDAR’s field-of-view to the specific road segment priorities, setting the stage for unparalleled traffic analysis.
When it comes to detecting and tracking road users, the PointVoxel-RCNN (PV-RCNN) comes into play. This deep learning model excels in the utilization of 3D point cloud data for highly accurate object detection. The algorithm’s keen eye for detail enhances the capabilities of our UAVs, making them not just aerial vehicles but intelligent data gatherers.
Finally, for the seamless and high-speed exchange of this invaluable data, WaveGAN plays a crucial role in Flying Ad-hoc Networks (FANET) for data transfer. The Generative Adversarial Network-based algorithm optimizes network topology in real-time, ensuring the most efficient paths for data exchange. This ensures that the network remains robust and can adapt to various conditions, completing the trifecta of algorithms that make this project a milestone in UAV and LiDAR applications.
Selected Publications
- B. Cherif, H. Ghazzai, A. Alsharoa, H. Besbes and Y. Massoud, “Aerial LiDAR-based 3D Object Detection and Tracking for Traffic Monitoring,” 2023 IEEE International Symposium on Circuits and Systems (ISCAS), Monterey, CA, USA, 2023, pp. 1–5.
- Z. Osterwisch, O. Rinchi, A. Alsharoa, H. Ghazzai, and Y. Massoud “Multiple UAV-LiDAR Placement Optimization Under Road Priority and Resolution Requirements”, in Proc. of the IEEE International Conference on Communication (ICC), Rome, Italy, May 2023.