Inanimate Occupants of Space

In general understanding things in the environment includes being able to predict consequences of actions, or of currently observed states (including motion). It may also be used in explaning how things got as they are, and in designing new objects to be assembled, or routes to be followed, etc.

The various spaces mentioned here can be occupied by many different sorts of objects that a domestic robot will need to know about, some of them inanimate and inactive, others active, or at least mobile or in some cases not merely active but also animate or intelligent.

The objects will vary in many different 'orthogonal' dimensions about which the robot will have to learn, though maybe some subset can be innate. (For kitty probably everything will have to be innate, so that we can understand the nature of the requirements for learning as a result of trying to program direcly the consequences that learning would produce.)

For a more extended discussion of the orthogonal dimensions (or the 'basis functions' for generating a variety of physical situations and processes) see the discussion in this presentation on vision: http://www.cs.bham.ac.uk/research/projects/cosy/papers/#pr0505

• size
• shape
• weight
• function
• origins
• material of which they are made (kind of stuff)
• their causal powers and constraints, which depend on both the kinds of material (e.g. whether rigid or flexible, elastic or plastic, etc.), the shapes of the objects, and in the case of composite objects, how they are put together (e.g. a triangle is inherently rigid whereas a quadrilateral is not, though adding a diagonal can make it rigid in a plane though not in 3-D, adding two diagonals can make it rigid in 3-D, (some examples are in http://www.cs.bham.ac.uk/research/projects/cosy/papers/#pr0506
• the spatial context which can alter their visibility, the constraints on possible motion, etc., including
  • restricting otherwise possible motions, e.g. a knife-holder restricting most translations and rotations, apart from sliding out of the holder,
  • 'de-activating' some kinds of flexibility, e.g. holding a sheet of cloth between two flat plates can prevent wrinkling, folding; stretching a piece of string between two clasp or tie points can prevent it losing straightness, etc.
• whether the robot ever needs to move them or whether they are merely part of the environment (e.g. large built-in chimneys, cupboards, cookers, etc.)
• There will also be larger objects such as the whole building, the hill on which it is located on a hillside, the town, and many objects in the town such as the roads, the river, the bridge over the river, and many buildings with different functions etc. etc.

Note: for Kitty the variation in physical contexts will be very constrained, as indicated here.