Mon Jan 11, 2021 10:43 am
- Does it make sense to define a funnel without stating what it does to the object?
- The exit of the funnel is defined as the set of possible final hand-object configurations.
- Some funnels require a specific hand-object configuration so that they can start. For this reason, also the opening of a funnel must incorporate the object pose.
- A screwdriver can be defined by the screws it can be applied to (e.g. only x-heads)
- A programmed function is defined by the inputs it accepts (e.g a double followed by an integer) and the return value it produces (e.g. a float).
- The spatial arrangement of the compliant, actuated constraints implicitly encodes the set of initial, intermediate, and final hand-object configurations.
- Imagine a compliant common funnel (made out of silicone). If it is applied to an object that is slightly larger than the funnel exit, the final funnel-object configuration will be a rubber ring that adapted its morphology to the shape of the object. But the object does not define the exit of the funnel. The morphology of the rubber ring (the funnel exit) implicitly encodes the final funnel-object configuration. True - depending on the object, the final funnel-object configuration will be different. But also depending on the number, the output of the square function will be different. But the algorithm that the function is based on and the prescriptive synergy the funnel is based on, do not change.
- The screwdriver still accepts many different screws, albeit only the ones with a x-head.
- The programmed function is not defined by the exact input, but only by the type of input. It still accepts many different inputs. And the output-type of the function also does not define a function. Perhaps, the definition of a manipulation funnel should include the type of objects it can be applied to.
- The funnel should generalize across different objects. If the object would be part of the funnel definition, this would no longer be the case.
- Morphological Computation Argument
The descriptive synergy is completely defined by the prescriptive synergy and the deformation.
The deformation is completely defined by the prescriptive synergy and the interactions of the morphology with the object.
Therefore, given a morphology and a prescriptive synergy, the descriptive synergy is a function of the object.
In the funnel view: The hand-object configuration is a function of the funnel morphology (its spatial arrangement of compliant, actuated constraints), its prescriptive synergy (actuation of the constraints), and the object. Given that for a specific funnel, the spatial arrangement of compliant, actuated constraints and the actuation pattern are fixed, the hand-object configuration is a function of the object.
Given the idea that a funnel is a mechanical function applied to an object, the definition of a funnel is therefore independent of the object.
- The object should be part of the funnel description, however, not the definition.
For example one could say: the wall funnel aligns a flat object surface with the flat wall surface and therefore, reduces uncertainty in orientation. Thus, this funnel only works for objects that have a flat surface.
But: Defining a funnel by the description (as the example above) leaves the funnel as a blackbox and nobody knows exactly how the alignment is achieved.
Permissions in this forum:You cannot reply to topics in this forum