Pollinators, such as bees, exploit patchily distributed food resources that replenish over time. The challenge for a forager is analogous to the “Traveling Salesman Problem” in graph theory: finding an efficient route to visit multiple locations once (flowers) and return to the origin (nest), for which there is no simple mathematical solution. Recent laboratory experiments have begun to show how bumblebees with an exclusive access to arrays of artificial flowers develop near-optimal foraging circuits (traplines) maximising nectar intake rates while minimizing overall travel distances as they gain experience. Whether and how these optimisation behaviours are expressed in natural conditions, at large spatial scales when multiple foragers might exploit the same floral resources, is poorly understood. Here I will compare data on bumblebees and honeybees foraging in large arrays of computer controlled flowers equipped with automated tracking systems to simultaneously record the complete foraging history and competitive interactions of multiple individuals over several consecutive hours. I will show how spatial network analyses (in which flight segments are edges and flowers are vertices) can help characterizing and model space use by individuals and colonies, as bees learn the spatial configuration and reward values of available feeding sites.