Application of Ordering Genetic Algorithm for a Multi-Criteria Optimization of Robotic Assembly Sequences

This paper attempts to use the Ordering Genetic Algorithm to generate the optimized sequence for robotic assembly with the following five-optimization criteria:(i) Geometrical Constraints (ii) Number of reorientations (iii) Gripper changes (iv) Subassembly formation (v) Manipulability. The assembly sequence along with the direction of insertion, grippers and subassembly is represented as an individual chromosome in this work. The fitness function is obtained as a weighted sum of the penalties assigned for every poor or undesirable assembly sequence. The idea of penalties allows many criteria to be considered simultaneously. Moreover this notion of penalty will allow one to transform a multi objective optimization problem to a single objective optimization problem. Starting from a randomly initialized population the next generation is evolved using the genetic operators like the modified version of partially matched crossover and mutation. The evolution of subsequent generations converges towards a near-optimal solution. The effectiveness of the proposed method is demonstrated using an example.