Vehicle-to-Everything Communication Using a Roadside Unit for Over-the-Horizon Object Awareness and Trajectory Planning
Part 4/5
Explore the code »
Table of Contents
Meet the National Science Foundation (NSF) Lawrence Technological University (LTU) and Michigan State University (MSU) Research Experience for Undergraduates (REU) Summer 2023 Cohort!
Our Students:
Michael Khalfin, Rice University, [email protected]
Jack Volgren, Pennsylvania State University, [email protected]
Luke LeGoullon, Louisiana State University, [email protected]
Brendan Franz, Harvard University, [email protected]
Shilpi Shah, Rutgers University, [email protected]
Travis Forgach, University of Michigan, [email protected]
Matthew Jones, Willamette University, [email protected]
Milan Jostes, Lawrence Technological University, [email protected]
Our Principal Investigators:
Chan-Jin Chung, Lawrence Technological University, [email protected]
Joshua Siegel, Michigan State University, [email protected]
Our Teaching Assistants:
Ryan Kaddis, Lawrence Technological University, [email protected]
Justin Dombecki, Lawrence Technological University, [email protected]
Self-driving and highly automated vehicles rely on a comprehensive understanding of their surroundings to operate effectively. While some information can be directly perceived through sensors, there are certain instances where information remains hidden from the original vehicle. To address this, vehicles can report information to each other, enabling the creation of over-the-horizon awareness. We created a Robot Operating System simulation of vehicle-to-everything communication. Then, using two electric vehicles equipped with global positioning systems and cameras and a roadside unit, we aggregated time, position, and navigation information. Also, we trained a deep learning model to detect cones in real-time video streams. Next, we created a Web-based Graphical User Interface that automatically updates when we drive the vehicles and encounter objects. Finally, we used the bicycle model and occupancy grids to predict vehicle trajectories and visualize potential collisions between vehicles or among vehicles and objects. In the future, we hope to recognize other objects, and incorporate time relevance in our long-term object permanence.
We would like to thank our professors and Teaching Assistants for all their time and dedication this summer. We would also like to thank our mentors, Joe DeRose and Nick Paul for their help.