Robust and Efficient Visual Odometry for Autonomous Robots in Degraded Environments

This M.S. research project examines how visual odometry can remain robust and computationally efficient when autonomous robots operate in degraded visual environments.

The work begins by characterizing when a stereo visual-odometry baseline loses reliability—for example, under low or dynamic illumination, obscurants, texture-poor scenes, or fast robot motion. Complementary signals such as thermal imagery, inertial measurements, or learned depth priors will be evaluated only where they offer a measurable benefit.

The goal is an uncertainty-aware approach to odometry that improves robustness while remaining feasible for onboard autonomy. The research is being carried out in Carnegie Mellon’s AirLab, under the advisement of Sebastian Scherer.