Haitham Khedr

Papers

Total Citations

139

H-Index

About

Haitham Khedr is a leading researcher at the intersection of formal methods, robotics, and artificial intelligence, with a primary focus on the safety and verifiability of autonomous systems. His work addresses a critical challenge: how to rigorously guarantee the safe operation of robots controlled by neural networks (NNs), especially when these systems process high-dimensional sensor data like LiDAR images. Khedr’s most influential contribution, his 2019 paper on the formal verification of NN-controlled autonomous systems (137 citations), pioneered a method to compute the exact reachable set of a robot navigating a workspace with polytopic obstacles, accounting for the complex, non-linear behavior of the neural network controller. This work provides a provable safety guarantee, a stark departure from purely empirical testing. In his 2020 follow-up, he advanced this framework by exploiting the geometry of a network’s linear regions to make verification more computationally tractable. By bridging the gap between deep learning and formal verification, Khedr’s research is foundational for deploying trustworthy autonomous vehicles and robots in safety-critical environments, establishing him as a key voice in the quest for reliable AI.