Things I suppose the robot should be able to detect/do when required
(even if not robustly):
- edges
- regions
- depth order
- corners
- convex hull (envelope)
- fit lines, conic sections, and maybe others to be learned like sin, cos, log, exp.
- principal components
- symmetry axes (where symmetries are simple transformations that allow to reproduce part of an image by appliying them to another one).
- reflexion
- rotation
- translation
Material I would love my robot to be able to handle and distinguish
(this may still be a dream though):
- Solid non-deformable objects
- Sand
- Water
- Viscous fluids (shampoo)