The research presented in this thesis focuses on the development of tactile event signature sensors and their application, especially in reactive behaviour-based robotic assembly systems. In pursuit of practical and economic sensors for detecting part contact, the application of PVDF (polyvinylidene fluoride) film, a mechanical vibration sensitive piezo material, is investigated. A <I>Clunk Sensor</I> is developed which remotely detects impact vibrations, and a <I>Push Sensor</I> is developed which senses small changes in the deformation of a compliant finger surface. The <I>Push Sensor</I> is further developed to provide some force direction and force pattern sensing capability. By being able to detect changes of state in an assembly, such as a change of contact force, an assembly robot can be well informed of current conditions. The complex structure of assembly tasks provides a rich context within which to interpret changes of state, so simple binary sensors can conveniently supply a lot more information than in the domain of mobile robots. Guarded motions, for example, which require sensing a change of state, have long been recognised as very useful in part mating tasks. Guarded motions are particularly well suited to be components of assembly behavioural modules. In behaviour-based robotic assembly systems, the high level planner is endowed with as little complexity as possible while the low level planning execution agent deals with actual sensing and action. Highly reactive execution agents can provide advantages by encapsulating low level sensing and action, hiding the details of sensori-motor complexity from the higher levels. Because behaviour-based assembly systems emphasise the utility of this kind of qualitative state-change sensor (as opposed to sensors which measure physical quantities), the robustness and utility of the <I>Push Sensor</I> was tested in an experimental behaviour-based system. An experimental task of pushing a ring along a convoluted stiff wire is chosen, in which the tactile sensors developed here are aided by vision. Three different methods of combining these different sensors within the general behaviour-based paradigm are implemented and compared. This exercise confirms the robustness and utility of the PVDF-based tactile sensors. We argue that the comparison suggests that for behaviour-based assembly systems using multiple concurrent sensor systems, bottom-level motor control in terms of force or velocity would be more appropriate than positional control. Behaviour-based systems have traditionally tried to avoid symbolic knowledge. Considering this in the light of the above work, it was found useful to develop a taxonomy of type of knowledge and refine the prohibition.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:653404 |
| Date | January 1996 |
| Creators | Kim, T. |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0019 seconds