Perception

The perception group research focuses on the development of state estimation for quadruped robots. The typical state of a robot includes its position, orientation, and velocity. However, the state can also include many diverse quantities such as simultaneous localization and mapping (SLAM) and ground reaction forces. State estimation is an essential step to any control problem and its difficulty is often underestimated.

 

Current perception sensors include:

  • Lidars - for large scale navigation and obstacle avoidance
  • Colour Cameras - for path planning localization and mapping in indoor/outdoor operations, and semantic mapping.
  • Depth Cameras - for foothold planning, localization and mapping in indoor conditions
  • Force/Torque Sensors – for measuring joint effort and ground reaction forces
  • Encoders – for measuring joint angle and speed

The main tasks of the perception system are:

  • State estimation
  • Simultaneous Localization And Mapping (SLAM)
  • Visual inertial SLAM
  • Vision based terrain Modeling for locomotion with foothold planning (accurate 3D model, traversability map including the terrain type)
  • Terrain/obstacle classification for semantic mapping
  • Object recognition and manipulation
  • Visual servoing
  • Automatic sensor calibration