We show that the genetic algorithm that Leu described to plan component placement sequences and feeder assignments for pick-and-place printed circuit board assembly tasks is a special case of a more general model that supports multiple placement-nozzles and independent feeder and board link (chromosome) evaluation methods. We also show that independent link evaluation can be used to offset a reduction in the parent link sample space and that these results are better than what can be achieved through link-pair evaluation. These generalizations extend the capabilities of the genetic algorithm to a broader range of manufacturing scenarios.

We previously proposed a general algorithm for coordinating the motions among multiple machines in a shared assembly environment based on a constant-speed motion model. In this paper, we extend this work to a minimum-jerk polynomial motion model and describe a new speed-planning algorithm to plan automated assembly machines' motions. Machines are planned sequentially, based on their priorities, by mapping the motions of higher-priority machines into forbidden regions in two-dimensional space-time graphs. Collision-free minimum-jerk motions are then planned between the forbidden regions in the graphs. The new speed-planning algorithm is evaluated on a dual-robot surface-mount technology assembly machine in which both robots share a common workspace. Note to Practitioners-Automated assembly processes, especially surface-mount technology manufacturing, require a high degree of precision when placing certain components. This motivated us to find a way of maintaining good positional accuracy by planning smooth motions for the machines that perform these tasks. Since many of these machines have two or more robots, their motions must also be coordinated. We developed an algorithm that combines coordinated motion concepts with a minimum-jerk motion model that can solve these problems. The algorithm plans segmented paths for the robots and then sequentially plans their speeds to prevent collisions between them. The planned speeds ensure position, velocity, and acceleration continuity between path segments. The smooth motions resulting from this method enable high-accuracy component placement. The tradeoff for this improvement is increased cycle time compared to other speed-planning methods

A general, hierarchical model for describing automated assembly systems is developed. This hierarchical model allows any complex assembly system to be decomposed into simpler sub-systems. This decomposition can be applied recursively to generate a multi-level system structure. Planning and control can be performed on each part of this structure and, if necessary, between parts. The model was applied to a simulated dual-robot, surface-mount assembly system to simplify the problem of planning coordinated motions between the robots. For the purpose of illustration, both linear and polynomial motion-planning models were analyzed. Part pick-up and placement delays were also introduced. Furthermore, system performance as a function of the number of placement-nozzles was also quantified. By weighing the benefits of additional placement-nozzles against their costs, the ideal number of placement-nozzles may be determined.

A general, hierarchical method of planning coordinated motions among multiple objects in a dynamic environment is developed. This model consists of a two-phase approach: planning a global path for each object and then locally optimizing along that global path. The applicability of this model to surface-mount manufacturing systems is studied in different scenarios. By changing the priorities of the machines, different results may be obtained. By comparing these results, the best one may then be determined. Simulation results using the two-phase algorithm show an improvement over other planning and control methods.