Â鶹ŮÀÉ

Led by Georgios Sklivanitis, Ph.D.

REU Scholar: Alyssa Falcon

REU Scholar Home Institution: Lehman College

REU Mentor: Georgios Sklivanitis, Ph.D.

Radio propagation digital twin to evaluate pedestrian localization in smart streetscapes

The localization of outdoor wireless signals is a significant research focus at Â鶹ŮÀÉ's Institute for Sensing and Embedded Network Systems Engineering (I-SENSE). By capturing wireless emissions in outdoor spaces, it is possible to track the source of these emissions. However, measuring the Received Signal Strength Indicator (RSSI) for localization methods has its drawbacks, including potential inaccuracies and incomplete datasets due to manual collection. To address these issues, a digital twin of Clematis Street was implemented to eliminate the challenges associated with manual data collection. This project utilized the Sionna open-source framework, leveraging its ray tracing capabilities to simulate channel propagation on a recreation of Clematis Street created with Open Street Maps (OSM). The dataset of RSSI values collected in this digital twin was then tested using various RSSI localization methods, including Trilateration, Convolutional Neural Network (CNN) Classification, Maximum Likelihood Estimation, and Fingerprinting. The results showed that the digital twin behaves the same as the real dataset, allowing researchers to test different localization algorithms that can be seamlessly transferred to real-world testing. This digital twin can be used for further research regarding Clematis Street, as it allows for changes to inputs such as material types of objects, repositioning objects, and applying other localization methods in this environment

REU Scholar: Jeremy Kim

REU Scholar Home Institution: The University of Alabama

REU Mentor: Georgios Sklivanitis, Ph.D.

Towards development of a digital twin for simulation of underwater robotics

This thesis investigates HoloOcean, a simulator designed to enhance the accuracy of simulation data and realistic image generation for underwater environments. The study focuses on HoloOcean's capabilities in rendering precise maps and 3D environments that closely mimic real-life conditions, emphasizing caustic effects and underwater physics. Given the challenges and high costs associated with underwater field trials, high-accuracy simulators like HoloOcean are essential for testing and developing algorithms. HoloOcean, an open-source simulator built on Unreal Engine 4 (UE4), addresses these needs effectively. It features multi-agent support, various sensor implementations typical of underwater environments, and simulated communication capabilities. Notably, it introduces a unique sonar sensor model that utilizes an octree representation of the environment to provide efficient and realistic sonar imagery. My research addresses technical challenges in developing simulation setups and proposes strategies for seamless transitions and compatibility. While many comparative simulators rely on Gazebo due to its effective integration with ROS 1 (Robot Operating System) and its comprehensive set of built-in features, there is a need to evaluate how HoloOcean measures up against these established systems. Thanks to its foundation on UE4, HoloOcean supports the easy addition of new environments, simplifying extensions. Its control is further streamlined through a Python interface, facilitating installation via pip and requiring minimal code to run simulations.