Robotics and Autonomy

(under construction)

• (Underwater) SLAM

  • Semantic scene understanding, prediction of how scene will react
  • Closed-loop or active perception: Detecting errors, deficiencies or inaccuracies in your world model and taking action to remedy them
  • Long-term SLAM, lifelong learning: Allowing SLAM maps to adapt to a changing world, and changing scene appearances (e.g. moving lights, changing seasons)
  • Running SLAM for a long time, and/or running SLAM over many repeated missions

• Multi-vehicle

  • Cooperative vehicle teams with and without high-bandwidth comms (tether versus acoustic)
  • Sharing perceptual information and decision making for active perception
  • Sensing given heterogeneous capabilities (some have lights, some have cameras, some have stereo, some have sonar?)
  • Closed-loop sensing given hetergeneous capabilities.

• Power-aware planning

• Mapping computational algorithms onto real hardware (e.g. finite RAM)