Toward Detecting Cyber-Physical Attacks in Additive Manufacturing Using Multi-View Visual Odometry

Abstract This paper presents a method using multi-view visual odometry as an independent tool to reconstruct deposition trajectories in additive manufacturing processes. A physical testbed is presented including camera and encoder retrofits to a Lulzbot TAZ 6. The system, including added sensors, is interfaced using the Wattsworth decentralized IoT framework for data acquisition, preliminary processing, and storage. The proposed visual odometry method is presented, and preliminary testbed results show reliable encoder feedback and camera calibration for use as ground truth in future validation.