Plastic products such as toys with articulated arms, legs, and heads are traditionally produced by first molding individual components separately, and then assembling them together. A recent alternative, referred to as in-mold assembly process, performs molding and assembly steps concurrently inside the mold itself. The most common technique of performing in-mold assembly is through multi-stage molding, in which the various components of an assembly are injected in a sequence of molding stages to produce the final assembly. Multi-stage molding produces better-quality articulated products at a lower cost. It however, gives rise to new mold design challenges that are absent from traditional molding. We need to develop a molding plan that determines the mold design parameters and sequence of molding stages. There are currently no software tools available to generate molding plans. It is difficult to perform the planning manually because it involves evaluating large number of combinations and solving complex geometric reasoning problems.

This dissertation investigates the problem of generating multi-stage molding plans for articulated assemblies. The multi-stage molding process is studied and the underlying governing principles and constraints are identified. A hybrid planning framework that combines elements from generative and variant techniques is developed. A molding plan representation is developed to build a library of feasible molding plans for basic joints. These molding plans for individual joints are reused to generate plans for new assemblies. As part of this overall planning framework, we need to solve the following geometric subproblems -- finding assembly configuration that is both feasible and optimal, finding mold-piece regions, and constructing an optimal shutoff surface. Algorithms to solve these subproblems are developed and characterized.

This dissertation makes the following contributions. The representation for molding plans provides a common platform for sharing feasible and efficient molding plans for joints. It investigates the multi-stage mold design problem from the planning perspective. The new hybrid planning framework and geometric reasoning algorithms will increase the level of automation and reduce chances of design mistakes. This will in turn reduce the cost and lead-time associated with the deployment of multi-stage molding process.