Existing cooperative file downloading schemes have some blindness on the selection of collaborative nodes. In this paper we present a cooperative file downloading scheme with genetic algorithm to resolve the collaborative nodes selection problem, especially in the case that we cannot have a prior knowledge about the data rate and position of the mobile nodes. A simple and efficient extended on-demand proxy discovery algorithm will be used to find the potential cooperative nodes. Character of the genetic algorithm (GA) makes it suitable for the selection of cooperative nodes. After this, client uses multiple parallel paths for file downloading. The simulation results show that the cooperative file downloading scheme with genetic algorithm can successfully select collaborative nodes with better performance and effectively reduce file download latency.