Meeting Abstract
Through simulation we characterized how visual cues that ants use are extracted, prioritized and stored during navigation. A foraging model simulates navigation in a procedurally generated environment where the visual cues could be precisely characterized. In these environments, our algorithms extracted and stored the visual cues that were available during a single Levy walk foraging event. Following a random foraging event, the success on subsequent foraging bouts using the stored information was examined. When we examined subsequent foraging walks we found the success of the simulated ant in finding the goal location using only a particular cue or a combination of cues depended on two factors – the length of the route and decay rate of information in a memory network. To further explore this, we simulated the foraging event over various sampling points and implemented linear or exponential decay in the networks storing the information. Our data suggests that the optimal strategy is to sample and store around 1000 points along the foraging route, independent of scale, with a network subjected to exponential decay. These parameters resulted in a stored representation that allowed the simulated ant to best find the goal on subsequent foraging bouts. We then produced several novel random foraging walks with the same goal location. The subsequent walks for these foraging events had similar success demonstrating sufficient information was stored and resulted in idiosyncratic foraging routes due to the varied information encountered during the random walk. Additionally, we explored how multiple subsequent walks updated and modified memory to produce more robust walks over time. Lastly, we compared the success of subsequent of the model when foraging in sparse and cluttered environments.
