Empirical assembly planning: A learning approach

An assembly structure is a physical as well as a conceptual structure. Physically, an assembly structure is composed of individual components and their relationships. Conceptually, an assembly structure is hierarchically organized into subassemblies.The focus is to generate an assembly sequence from a physical specification of the assembly structure. A component is modeled iconographically as well as by its geometric features: outer normal of a face, center axis of a hole, and more. Two components mate when their features are connected by a set of fixed geometric and kinematic relationships, called mating conditions. These connections are specified by an assembly designer. An assembly planner generates a sequence of mating operations inferred from the specified mating conditions.Mating conditions specify the assembly structure to be constructed: a goal state for the planner. The goal state is hierarchically structured into subassemblies, forming an assembly hierarchy. Using the assembly hierarchy as an abstraction hierarchy, the planner reduces the planning task because a subassembly should be assembled before its parent assembly structure.Two assembly planning approaches are developed. One is based on detecting spatial interferences between sibling subassemblies. The other is based on planning with prior experiences. The latter approach monitors the execution of an assembly sequence, generated from the previous experience, whether there exist any collision free path when executing an assembly operation among the already assembled components, obstructing a collision free path. When failure is detected, the planning experience is modified so that the planner avoids repeating the same mistake.U of I OnlyETDs are only available to UIUC Users without author permissio