Second layer - integrating the perception of gravity and the floor into the geometry of action

Note: under construction: please check older version(s) at:

• Position and location in the world: annotated Table of Content
• Gravity and the floor: annotated Table of Content

Additional working definitions:

• Action consequences are the 'real world' results of an action sequence.
• A plan is an information structure that involves the prediction of action-consequences associated with the hypothetical action-sequence resulting from the invocation of one or more skills.
• A task is an information structure that describes a desired action-consequence along with a plan.

Layer 2 addresses the problems uncovered in layer 1 for the representation of skills using only joint angles. We found that such a skill representation should allow us to keep up the normal walking motions even when falling or standing on our head. In that layer we introduced the perception of gravity and of direction, but did not integrate them into the geometric representation of skills.

We can use the perception of gravity and the feedback from standing on the floor to add internally perceived but world-based axes of gravity and level.

• For action, we add gravity from the centre of gravity to the floor, and a level floor (at right angle to gravity.
• For perception, we add gravity and a level (like a horizon).

We propose an additional set of internal polar coordinates that is appropriate for limbs rather than joints and that is more relevant to the orientation of the actor in the world, as defined by gravity and the floor.

• We shall assume that the 'level' axis applies to each of the joints, and an axis of 'direction' can be taken from the direction the actor is facing.
• We need a third dimension we shall call 'twist' to account for being able to rotate the hand relative to the elbow, for example.

We can use this coordinate system to describe the position of limbs.

From skilled action to the action sequence in the world and observable by an audience: different geometries.

The actor as an observer - a gravity-anchored standardized geometry for the inner results of perception.

Mimicry based on the gravity-anchored inner geometry - translation into joint-angle geometries.

Layer 2 examples where a skill integrates perception and action:

• Chasing prey: A simple example of integrating the act of visual perception with prediction followed by action
• Climbing stairs: Combining visually derived estimates with managing gravity to control a simple action
• Imitating action: visually perceiving the instructor's limbs to estimate angles relative to gravity - a second approach to modelling mimicry

Information representation and information processing for basic innate skills:

• managing geometries: information representation and transformations - skills components
• managing time: synchronizing movements to coordinate the action resulting from a skill

An information-processing feasibility model is successful if it can simulate the phenomena to be described, within the simplifying assumptions that have been made explicit, and without violating major constraints that should also have been made explicit (such as not requiring a racial consciousness, and not requiring extra-sensory perception).