Predictable DNN Inference for Autonomous Driving
Autonomous vehicles (AVs) depend on sensors and deep neural networks (DNNs) to perceive their surroundings and make real-time driving decisions. However, ensuring the predictability of AV perception in both temporal and functional aspects is challenging. Variations in timing and accuracy during DNN inference can occur under different driving conditions. This project aims to profile and optimize these variations, making the autonomous driving pipeline more predictable and robust.
Faculty
Research Fellows
