A swarm intelligence approach to path synthesis of mechanism

Mechanism path synthesis can be viewed as a pattern-matching problem in essence. Some key techniques related to its implementation such as the generation of the training path curves, the digital description of mechanism path and the clustering of path curves are discussed in detail. After generating amounts of training curves, the moment invariants of them are computed, which are defined as the attributes of the curves. By modifying the distance metric in LF model, these moment invariants are clustered based on ant clustering method. When the mechanism corresponding to the expected curve is needed, all the curves are discretized and the problem of pattern matching is solved. All these operations are integrated into a pattern-matching framework for mechanism path synthesis. A practical example of four-bar linkage path synthesis demonstrates that the ant-clustering algorithm is efficient and the proposed method is effective.