Basic set of behaviours for programming assembly robots

• Main Author: Pettinaro, Giovanni Cosimo
• Published: University of Edinburgh 1996
• Subjects: 006.3
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.660566

We know from the well established Church-Turing thesis that any computer programming language needs just a limited set of commands in order to perform any computable process. However, programming in these terms is so very inconvenient that larger sets of machine codes need to be introduced and on top of these higher programming languages are erected. In Assembly Robotics we could theoretically formulate any assembly tasks, in terms of moves. Nevertheless, it is as tedious and error prone to program assemblies at this low level as it would be to program a computer by using just Turing's commands. An interesting survey conducted by Kondoleon in the late seventies showed that the most common assembly tasks in manufacturing industry are just 12. Since the research conducted in this thesis is developed within the behaviour-based assembly paradigm which views every tasks as a behaviour, we wonder whether there exists a limited set of them convenient to be used by human operators or by automatic planners. In order to investigate such a problem, we set a project in which, taking into account the statistics of Kondoleon's survey, we analyze the experimental behavioural decomposition of three significant assembly tasks (two similar benchmarks, the STRASS assembly, and a family of torches). From these three we establish a basic set of such behaviours. The three test assemblies with which we run the experiments can not possibly exhaust all the manufacturing assembly tasks occurring in industry, nor can the results gathered or the speculations made present a theoretical proof of the existence of the basic set. They simply show that it is possible to formulate different assembly tasks in terms of a small set of about 10 behaviours. Comparing this set with Kondoleon's one and with Blalch's general-purpose robot routines, we observed that ours was general enough to represent 90% of manufacturing assembly tasks and convenient enough to be easily used by human operators or automatic planners. A final discussion shows that it would be possible to base an assembly programming language on this kind of set of basic behaviours.