AVT
Autonomous Vehicle Team at PSU
1. Joining the Penn State Advanced Vehicle Team (AVT)
In August 2024, I joined the Penn State Advanced Vehicle Team (AVT). The Penn State AVT’s mission is to offer an unparalleled learning experience for students by developing algorithms for autonomous driving, re-engineering electric vehicles to enhance driving range, and improving vehicle architecture while maintaining consumer demand.
Since 1988, the Penn State AVT has competed in the Advanced Vehicle Technology Competitions, focusing on building a fuel-efficient vehicle without compromising performance or safety.
I was excited to join a team that emphasizes real-world applications of cutting-edge engineering. From day one, it was clear that collaboration and hands-on learning are central to AVT’s culture.
We regularly partner with industry sponsors and academic departments, ensuring that each member gains exposure to multi-disciplinary projects. It’s a fast-paced environment, but the support from faculty and alumni helps keep our goals on track.
FA 2024 AVT – I'm standing at the far left.
2. Building the “Sensor Decider”
I developed a core pipeline component in ROS2 using Python to integrate data from HDMaps, LiDAR, and Camera. This module publishes data to the Control State Machine, generating real-time vehicle trajectories.
By synthesizing multiple sensor inputs simultaneously, the system can more reliably account for dynamic objects on the road and update trajectories at high frequency.
Additionally, we introduced a robust error-handling routine to ensure the pipeline gracefully recovers from unexpected sensor dropouts or corrupted data, which was crucial for real-world viability.
(This was the first time we tested real sensor inputs from the Perception team directly on the car.)
First sensor test: captured the initial sensor integration on the car.
3. Prototype Demonstration and Design Showcase
Ultimately, the car was able to operate in full autonomous mode by recognizing dynamic objects and traffic signals. Our demonstration validated real-time data generation within an 80 cm error range, highlighting the system’s precision.
The design showcase provided an opportunity for industry professionals and fellow researchers to offer feedback. We gained insights into refining the user interface for remote monitoring and ensuring robust performance under inclement weather conditions.
Moving forward, we plan to expand our testing environments to different terrains and incorporate more advanced sensor fusion algorithms. Each iteration brings us closer to a truly autonomous solution capable of handling diverse real-world scenarios.
Testing video: when the blue light is not blinking, the vehicle is in full autonomous mode.