Plant breeding has been defined as the art and science of producing desired characteristics through artificial selection. Practiced since the beginning of civilizations, plant breeders in the 20th Century made enormous changes to important agronomic traits. In the 21st Century, increasing demands for food, fiber, and energy with less water, land, fuel, and fertilizer will force plant breeding to become more efficient and effective. With the application of operations research, we frame the multi-allelic trait introgression project in plant breeding into an engineering system. We also discuss the major problems encountered and create new metrics or models to improve this process. We design the Predicted Cross Value (PCV) for one pair parental selection and demonstrate its advantages over the conventional metrics. Next, in order to optimize the resource allocation during the introgression, we propose the Markov Decision Process model to dynamically allocate resources. The results show that such approach outperforms the static breeding strategy. Finally, to make the PCV concept more practical to realistic breeding process, we extend the PCV to NPCV for multi-pair parental selection. We present the results and show that the NPCV makes the parental selection more efficient and effective. In general, this dissertation discusses applying operations research into the trait introgression process to improve the efficiency and effectiveness from several perspectives.