Abstract: Nearly all the take-away platforms provide opportunities of part-time delivery service. Compared with full-time staff, part-time deliverers have several features when they participate in crowdsourcing delivery, including "multiple delivery path, limited working time, fixed destination and etc". Considering these characters, a research is made on take-away distribution task selection and task routing problem in crowdsourcing mode and a two-layer algorithm is proposed in order to enhance deliverers' efficiency and raise their revenue. A delivery routing problem model is built in the bottom layer, and solved with improved genetic algorithm. The upper layer adopts greedy algorithm and compares delivery revenue according to the model in the bottom layer to select tasks, so that the crowdsourcing workers' revenue can reach the largest level. A numerical case is given to test and verify the effectiveness of model and algorithm, the solving time increases linearly with the number of potential orders. In practice, this method can achieve a relatively high-quality solution in an acceptable time when the scale of alternate orders has been controlled.

Key words: take-away delivery, crowdsourcing, task selection, routing optimization