Can we generate physics-based adversarial samples to understand the robustness of sensors under various environmental conditions?

How to calibrate infrastructure sensor nodes to accurately provide uncertainty information to perception and communication modules?

How can we leverage naturalistic driving data to enhance simulations?

For object detection, is there context bias occurring during training and can we evaluate context bias in different domains?

Can we analytically estimate graph matching in problems with considerable measurement noise on edge weights without presence of node labels? 

How can we leverage graph based optimization methods for LiDAR data analysis?