Assembly planning is a search problem whose time complexity rises exponentially as the number of parts increases. Hence, it is very important to increase the efficiency of the assembly planning. To increase the efficiency of the assembly planning, the paper presents a) a method for obtaining non-redundant assembly constraints, b) a method for minimizing search-space in generating assembly sequences, and c) a method for selecting efficient assembly sequences using interactive queries.
Assembly constraints have many redundancies if they are obtained seperately since assembly constraints are caused by such independent sources as geometry of parts and the contact coherence. Redundancies of assembly constraints decrease the efficiency of the assembly planning since each assembly task should be tested if it violates any assembly constraint. Non-redundant assembly constraints are obtained by excluding the constraints a) that are intersected by independent sources and b) that are covered by other constraints using the "superset rule". The superset rule is one inherent property of assembly constraints.
The number of assembly tasks rises exponentially O($n^3$). However, significant parts of assembly tasks are rejected during the selection of desirable ones since they violate assembly constraints or they have unstable subassemblies. To minimize the number of searched assembly tasks, the paper searches for only assembly tasks that do not violate assembly constraints and that do not have unstable subassemblies. The proposed method generates assembly sequences by recursively separating unconstrained mating subassemblies from a target product and its subassemblies. Also, the proposed method generates parallel assembly sequences by searching for each combination of assembly tasks that can be executed concurrently.
To evaluate assembly sequences the paper presentes three evaluation measures: a) the number of direction changes at an assembly line, b) the efficiency of an assembly line, and c) the possibility of assembly fail. These evaluation measures and additional constraints are combined by user queries to select efficient assembly sequences.
