The CAR Lab

Congratulations to Weisong Shi, chair of Computer & Info. Sciences at the Univ. of Delaware, on being named a University of Delaware Alumni Distinguished Professor. Today, he delivered the lecture, "Vehicle Computing: From Traditional Transportation to Computing on the Wheels," and brought one the The CAR Lab's autonomous vehicles onto campus. HydraU and HydraD are the only two autonomous vehicles in Delaware. Below, Jamie Phillips gives Dr. Shi a medal to recognize this honor. Dean Phillips and Dr. Shi pose with HydraU on The Green. #Hengineers #AutonomousVehicles

✨ Exciting News! ✨ I will present our recent work "An Efficient Data Transmission Framework for Connected Vehicles" at the upcoming ACM/IEEE SEC’24 conference in Rome, Italy! #SEC! 🎉 A huge thank you to my amazing collaborators: Yongtao Yao, Junzhou Chen, Prof. Sidi Lu, and Prof. Weisong Shi. Thanks to The CAR Lab. 🛡️The Challenges: Connected vehicles (CVs) face complex, evolving challenges in maintaining efficient and reliable data transmission for real-time decision-making due to: 1️⃣ High transmission bandwidth costs from continuous big data streaming. 2️⃣ Latency-sensitive, resource-intensive application demands on CVs. 3️⃣ Dynamic driving conditions (e.g., high-speed vs. low-speed) limit the effectiveness of non-adaptive temporal compressive sensing (TCS). 4️⃣ Limited access to real-world hardware platforms for testing. 🌍Key Highlights:  ✨ Adaptive TCS by RL policy: Maintains higher reconstruction quality and average detection rate compared to non-adaptive TCS, reducing bandwidth requirements.  ⏱️ Real-time application with low latency: High detection accuracy on compressed data with an average inference speed of 42 microseconds in the vehicle, catering to latency-sensitive requirements 🛠️ Offloading to EdgeServer: The VEC framework offloads significant CPU and memory pressure from the vehicle to the EdgeServer, satisfying resource-intensive application demands. 🚀 Testing on industry-grade hardware to ensure real-world performance and reliability. 📊Our Results: Our experiments show that the VEC framework achieves an 18x reduction in bandwidth at 320KB/s while maintaining high detection accuracy and reconstruction quality, positioning the framework as a promising solution for real-world CV applications. You can find more details here: 📄 Paper preprint: https://lnkd.in/eqrWhsHk #EdgeComputing #AutonomousDriving #ConnectedVehicle #Compression #Bandwidth #DataTransmission #Robotics #ReinforcementLearning

It is a huge honor to deliver the inaugural lecture of the named professorship at Tuesday, October 29, 2024 | 4–6 p.m. The talk will be live streamed and open to public. Please register to attend if you are interested. https://lnkd.in/edxsgXZa #VehicleComputing #Autonomousvehicles #connectedvehicles