Design of an experience-based assembly sequence planner for mechanical assemblies Arun Swaminathan, Saghir A. Shaikh and K. Suzanne Barber The Laboratory for Intelligent Processes and Systems, Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, Texas 78712 (USA) SUMMARY This paper presents a design of an assembly sequence planner based on a “plan reuse” philosophy. Most of assembly planning research in the past has attempted to completely plan each problem from scratch. This research shows that stored cases of basic assembly configurations can be applied to a given assembly problem. It is observed that the number of such basic assembly configurations is quite small. The planner divides the assembly into a number of constituent configurations, which are called “loops”. These act as subgoals in its search for solutions. Plans retrieved for all subgoals are fused into a set of plans that are consistent with the constraints implied by each plan. Application specific constraints on the assembly are explic- itly handled in the second phase of planning. Mechanisms for assembly representation and implementation details of the planner are also presented. KEYWORDS: Assembly planning; Case-based reasoning; Sub- goals; Computational complexity; Precedence constraints; Precedence graphs. I. INTRODUCTION Assembly planning plays a major role in aiding shop floor control, production system design and scheduling activities. The assembly plan provides vital manufacturing informa- tion, and imposes constraints in the selection of production equipment and alternative routings. The planning process followed by a human planner is iterative. Humans examine the part and assembly drawings to create a tentative plan based on intuitive geometric reasoning. They subsequently revise the plans based on mechanical properties, assembly practices, managerial considerations, design notes, etc. 1 The approach commonly used in planning is to subdivide the task into smaller and smaller problems until the simplest problems are obtained. More information on the modeling and reasoning issues in assembly planning can be found in literature. 1–11 An insight gained from our research is that planning in the assembly domain is best served by considering many of the constraints at the local level before the global level. That is, combinations of locally feasible solutions are combined to give globally feasible solutions. The advantage is that the combinatorial explosion of plan options is controlled at a lower level without allowing them to propagate to higher levels. This task needs to be supported by an appropriate identification of “localities”. Localities are a suitable subdivision of the problem in which planning for each subgoal can be performed relatively independent of each other. In this paper, the Assembly Planner using Experience (APE) is proposed. APE demonstrates that the Case-Based Reasoning (CBR) paradigm can be quite effective in the assembly domain. APE utilizes case-based planning to store, retrieve, and modify existing cases (experience) to develop assembly sequence plans. This paradigm is very powerful for complex domains but has not been applied to the assembly domain in a significant way. The focus of this research is to explore application of a planning paradigm, to combine both geometric and application constraints and to leverage past experience to resolve these constraints. The contributions of this research are the following:  To demonstrate some aspects of an efficient representa- tion of assembly information to support the re-use of past experience.  To demonstrate how a large variety of possible assembly parts configurations can be effectively reduced to make the search and retrieval problem of past experiences (cases) computationally feasible for practical sized assemblies.  To demonstrate the ability of APE to plan directly when the need arises, thereby providing a hybrid methodology which allows; (i) re-use of stored cases if they are found in the database, and (ii) for situations not found in the database, “primitive connections”* between parts are employed to form larger plans. Assembly modeling, which involves the representation of both the assembly and the plan, is an important aspect of the planner design. In APE, graphs are used to model the assembly. Graphs provide a formal, efficient and flexible representation. The assembly is represented by three kinds of graphs: Connection Graph, Mating Direction Graph, and Obstacle Fact Graph. The plan is represented by an Assembly Precedence Graph. Formal definitions of these graphs and their properties are defined in this paper. This paper is organized in six sections. The introduction is followed by Section II, which discusses the recent work in assembly planning and the case-based model of reasoning. This section also explains why CBR is useful for assembly planning. Section III describes the assembly modeling using graphs and the concept of “goals” in assembly. Section IV * A “primitive connection” simply shows how two parts may be connected (i.e. the first part could be placed in the assembly and then the second could be connected to the first or vice versa). Robotica (1998) volume 16, pp. 265–283. Printed in the United Kingdom © 1998 Cambridge University Press