Sequence planning for robotic assembly of tetrahedral truss structures

An artificial intelligence approach to planning the robotic assembly of large tetrahedral truss structures is presented. Based on the computational formalism known as production system, the approach exploits the simplicity and uniformity of the shapes of the parts and the regularity of their interconnection to drastically reduce the required geometric reasoning computation. The global database consists of a hexagonal grid representation of the truss structure. This representation captures the multiple hierarchies in tetrahedral truss structures and allows a substantial reduction of the search space without sacrificing completeness. It allows the choice of a hierarchy to be made only when needed, thus allowing a more informed decision. Testing the preconditions of the production rules is computationally inexpensive because the patterned way in which the struts are interconnected is incorporated into the topology of the hexagonal grid representation. A directed graph representation of assembly sequences allows the use of both graph search and backtracking control strategies. The extension of the approach to planning repair sequences is outlined. A prototype planner, named TASP, has been implemented and successfully generated assembly sequences for a structure made of 102 struts