Advancing Autonomous Vehicles in Urban Spaces

Exploring the bustling dynamics of urban landscapes, the “Self-Driving Cars in Urban Environments with Traffic” project team, part of the Mountaintop Summer Experience, is making substantial contributions to the self-driving car community. The team is working to enhance the decision-making capabilities of autonomous vehicles, navigating through the complexities of urban landscapes.

The team's approach focuses on ensuring safety and adherence to road rules for autonomous vehicles. It investigates how these vehicles behave in challenging scenarios where it's not always possible to strictly enforce all road rules. Metrics such as response risk, certainty and safety are fundamental in the analytical procedures, supporting the team’s commitment to advancing research in this field.

The team achieves this by crafting innovative algorithms and embracing the transformative capabilities of machine learning coupled with formal synthesis methods that ensure safety and adherence to rules.

Central to the team’s research is an intricately designed testbed, located in the Autonomous and Intelligent Robotics (AIR) Lab. The testbed mirrors the unpredictable nature of urban environments. Featuring roundabouts, single-lane roads, and S-curves, along with simulated urban elements such as pedestrians and traffic signs, it provides a thorough testing ground for advancements.

Led by Cristian-Ioan Vasile, an assistant professor of mechanical engineering and mechanics, the project is poised to leverage the transformative powers of artificial intelligence and machine learning. The team is committed to developing innovative algorithms that will empower autonomous vehicles with unparalleled navigational skills and responsive capabilities.

Vasile emphasized the rigorous verification and testing of diverse scenarios to contribute to the ever-evolving landscape of research in this area.

"Exploration and innovation are at the very heart of our endeavors," Vasile said. "Our work transcends the mere development of autonomous vehicles; we are challenging the existing limits of what can be achieved in urban settings.

“By compiling a comprehensive database of scenarios and vehicle responses, we are providing invaluable data to the broader self-driving car community, thereby playing a crucial role in a collaborative movement towards safer, more efficient, and reliable autonomous transportation.”

With this commitment to innovation, the vehicles are being prepared to navigate through an array of unpredictable and unconventional scenarios. This ranges from managing obscured traffic signs and hazardous driving conditions to responding adeptly to the unpredictable behaviors of other road users.

The team—comprising Lucas Koranda ‘26 (IBE ISE/finance major), Junan Mei ‘24 (mechanical engineering major), Michelle Li ‘26 (IDEAS major), Heidi Cobb ‘25 (mechanical engineering major), Ryan Kong ‘24 (mechanical engineering major), and three Master of Science in mechanical engineering students Disha Kamele, Gustavo Cardona and Kaier Ling—excels through a collaborative approach that leverages each member's specialized skills.

To date, the team has constructed five fully-assembled autonomous vehicles. The machines are equipped with onboard computers, laser sensors, cameras and upgraded drivetrains for enhanced steering accuracy. Within the testbed's intricate setting, the vehicles navigate with precision, showcasing the practical implementations of the team's comprehensive research.

“Our dedication is propelling these vehicles towards complete autonomy,” said Koranda, a Mountaintop fellow majoring in IBE industrial and systems engineering and finance. “By meticulously implementing various autonomy stages, we’re ensuring a seamless integration with their surroundings.”

He added, “Our methodical, progressive approach guarantees the vehicles’ intelligent interaction with their environment, enhancing their capabilities at each development stage. This underscores Lehigh University’s innovative spirit and our commitment to advancing this field.”

On the software front, the team has made significant strides, integrating sensors with onboard computers and incorporating a detection model, ready for extensive testing.

Reflecting their dedication to the broader research community, the team members are committed to offering open-source documentation and code upon the project's completion. They aim to contribute further by submitting a comprehensive research paper to the International Conference of Intelligent Robots and Systems in 2024.

The team aims to first leverage overhead cameras to establish a foundational GPS-mimicking system for autonomous navigation. This will soon give way to the vehicle's own onboard sensors and cameras, which will navigate various traffic situations—including running red lights and navigating complex right-of-way situations.

Story by Haidan Hu