Lernstrategien und visuelle Information für die Lokalisation eines Ziels beim Navigationsverhalten von Hummeln: Eine kombinierte Analyse auf der Verhaltens-, der neuronalen und der Modellebene
Zusammenfassung der Projektergebnisse
Finding home may be a difficult task, especially in complex cluttered environments. Central place foragers such as ants, honeybees or bumblebees need to solve this task many times during a day, since they need to collect food even large distances away from their nest in order to feed the larvae in their nest. Accordingly, these hymenopterans have evolved route finding and homing mechanisms, which – given the small size of their brains – can be expected to be computationally parsimonious and highly efficient. Understanding these mechanisms, thus, is of great importance scientifically, but also with respect to applications in autonomous small-sized artificial systems. This project uses bumblebees as a model system to focus on the learning strategies and the mechanisms employed to localise a barely visible goal, such as a nest hole or a previously learnt route that connects the nest hole e.g. with a food source, by combining behavioural, computational and robotic approaches. Bumblebees reveal an elaborate loop-like flight structure when leaving their barely visible nest hole the first time being initially completely naïve with respect the visual environment. The bees were concluded to learn just after leaving the nest the relevant environmental information. After this learning phase the bees do not immediately fly away, but were concluded to test by dedicated probing behaviour the usability of the learned information for finding the nest entrance again after a foraging trip. However, the learning and homing behaviour of bumblebees has proven to be more variable than expected on the basis of the performance of previously developed model mechanisms. Moreover, the homing mechanisms of bees are not perfect, as they make also mistakes, when they are confronted not only with the correct nest hole, but additional “dummy holes” that are not connected to the nest. If bumblebees have to cope with different sets of environmental visual cues that, after a learning phase, are brought into conflict by displacing them relative to each other as well as to the nest hole, they search at the virtual nest locations, which correspond either to one or the other set of visual cues and also switch between both virtual holes. To what extent the different cues are taken into account depends on the degree of conflict between their relative positions. Again, this kind of behaviour can hardly be explained on the basis of any of the previously proposed mechanisms of local homing behaviour. These unexpected results make it necessary to develop new models for home finding of bees. Although developing this kind of comprehensive model for local homing was beyond the scope of the current project, we could explain by a variety of modelling approaches relevant aspects of homing mechanisms that will serve as constituent components of such a comprehensive model. Three results are particularly relevant. (1) We could develop an optic flow-based mechanism that allows the animal to navigate to a goal in highly cluttered environments without colliding with obstacles in the way. (2) When assessing the quality of a home location in a cluttered environment in terms of its reachability based on some visual homing mechanism, it is sufficient for an animal to probe the homing success only in the immediate vicinity of the home in order to predict with a good reliability the overall success in returning to it from within a much larger area. This catchment area can be considerably increased by combining different cues in a local homing mechanisms, as these may define complementary areas in the environment within which the home can be reached. (3) If an animal has been displaced in a cluttered environment by some disturbance from a previously learnt habitual route and needs to find this route again to be able to return to its home, the best search strategy to find this route depends on a variety of behaviourally relevant parameters, such as the distance the animal has been displaced, the time the animal has for search (e.g. as a consequence of a limited energy reservoir) and the knowledge it may have, for instance, about the direction of its displacement. Part of these navigational mechanisms have not only been modelled to show, to what extent they can account for our experimental results, but could already be implemented on a biomorphic hardware chip and tested on robotic platforms, illustrating that bioinspired computationally parsimonious principles of navigation can be useful, at different levels of abstraction, when developing autonomous artificial systems.
Projektbezogene Publikationen (Auswahl)
- 2015. A Bio-inspired Collision Avoidance Model Based on Spatial Information Derived from Motion Detectors Leads to Common Routes. PLoS Comput Biol 11: e1004339
Bertrand OJ, Lindemann JP, Egelhaaf M
(Siehe online unter https://doi.org/10.1371/journal.pcbi.1004339) - International Conference on Event-based Control, Communication, and Signal Processing (EBCCSP) 2015, Krakow, 2015: 1-7. IEEE Xplore Digital Library
Milde MB, Bertrand OJN, Benosman R, Egelhaaf M, Chicca E
(Siehe online unter https://doi.org/10.1109/EBCCSP.2015.7300673) - (2016). A bio-inspired model for visual collision avoidance on a hexapod walking robot. In: Insect-Inspired Visual Navigation for Flying Robots, Volume 9793, N.F. Lepora et al., eds. (Edinburgh: Springer International Publishing Switzerland), pp. 167–178
Meyer, H.G., Bertrand, O.J., Paskarbeit, J., Lindemann, J.P., Schneider, A., and Egelhaaf, M.
(Siehe online unter https://doi.org/10.1007/978-3-319-42417-0_16) - (2018). Inferring temporal structure from predictability in bumblebee learning flight. In International Conference on Intelligent Data Engineering and Automated Learning (IDEAL) (Madrid, Spain), pp. 12
Meyer, S., Bertrand, O.J., Egelhaaf, M., and Hammer, B.
(Siehe online unter https://doi.org/10.1007/978-3-030-03493-1_53) - 2018. Spiking elementary motion detector in neuromorphic systems. Neural computation: 1-34
Milde MB, Bertrand O, Ramachandran H, Egelhaaf M, Chicca E
(Siehe online unter https://doi.org/10.1162/neco_a_01112) - 2018. Taking a goal-centred dynamic snapshot as a possibility for local homing in initially naïve bumblebees. Journal of Experimental Biology 221: jeb168674
Lobecke A, Kern R, Egelhaaf M
(Siehe online unter https://doi.org/10.1242/jeb.168674) - 2018. The problem of home choice in skyline-based homing. PLoS ONE 13: e0194070-e70
Müller M, Bertrand O, Differt D, Egelhaaf M
(Siehe online unter https://doi.org/10.1371/journal.pone.0194070)
