LE: a high level language for specifying vision verification tasks

The author describes, L_E, a visual language for specifying verification tasks in the context of the execution of assembly plans. The instructions of L_E have a predicative form, which, being linked in conjunction, imposes conditions on the application of robot actions. These instructions are related to the identification and localization of assembly parts, the verification of the computed trajectory, the verification of the correctness of the object positions, and the verification of the fact that the geometric relations between objects characterizing the manipulation process have been attained. To automatically transform these vision verification tasks into vision programs, the authors defined several data structures by using exhaustively the geometrical and physical object models supporting the system