Predicting ants

This is an intriguing bit of ant behaviour. Ants are trained to find food at a location that moves systematically on each learning trial. For one group the food moves further from the next with each trial. In another group the food moves further to the right with each trial. After training ants are tested with no food present and show a preference for searching at locations that follow the training rule (getting further from the nest or moving further to the right). It is fascinating to consider the possible explanations for this behaviour. My personal view is that the natural tendency of foragers to increase distance with experience may have been coopted in this experiment, rather than any conceptually impressive cognitive prediction mechansisms. But maybe … …

Cammaerts and Cammaerts (2016) Spatial expectation of food location in an ant on basis of previous food locations (Hymenoptera, Formicidae). Journal of Ethology, 1-9.

Familiarity for UAVs

In recent years the idea that visual route navigation might be implemented by agents making simple familiarity judgements has become popular. It allows for navigation with a simple unstructured visual memory that could be implemented in the mushroom body of an insect. Given the simplicity of the idea it was inevitable that it would be tested in UAV platforms. This paper reports tentatively successful trials of such an algorithm.

van Dalen, G. J., McGuire, K. N., & de Croon, G. C. (2016) Visual Homing for Micro Aerial Vehicles using Scene Familiarity. Proceedings International Micro Air Vehicles, Conferences and Competitions 2016.

Rapid landscape learning in bees

We all know that bees have specific flight patterns that are undertaken at the beginning of their foraging careers. Our understanding of these so-called learning flights has been massively advanced in recent years by the use of harmonic radar to track bees. This has allowed us to understand how the flights develop and how they are influenced by the structure of the environment. In this new paper, we get another insight into these learning flights. Bees are tracked for one flight only and then tested by being released within the extent of the learning flight or in a novel area. The results show that bees are capable of one-trial learning.

Degen, J., Kirbach, A., Reiter, L., Lehmann, K., Norton, P., Storms, M., … & Chamkhi, H. (2016). Honeybees Learn Landscape Features during Exploratory Orientation Flights. Current Biology, 26(20), 2800-2804.

Insect-inspired visual sensors

Obviously we know that insects, such as ants, are expert visual navigators but we have a limited understanding of how (if at all) the sensory physiology of ants is tuned to navigation. One approach to that question is to investigate how differently tuned sensors would perform at a navigational task, such as identifying stable environmental features. Many ants have UV and green photoreceptors which led to the suggestion that a UV-green contrast might be a good way to identify environmental features in the world. Here, Differt and Möller show that actually using the UV channel alone can be very useful for navigation, when the UV signal is simply thresholded useful a locally adaptive threshold. This is an interesting approach to thinking about sensor physiology, but also useful for designing sensors for navigating agents.

Dario Differt and Ralf Möller (2016) Spectral Skyline Separation: Extended Landmark Databases and Panoramic Imaging. Sensors, 16(10), 1614; doi:10.3390/s16101614

Navigational toolkit in action

Here is the abstract of a paper looking at the coordinated navigation behaviours of a trail laying species. Looks really interesting.

“Ants use multiple cues for navigating to a food source or nest location. Directional information is derived from pheromone trails or visual landmarks or celestial objects. Some ants use the celestial compass information along with an ‘odometer’ to determine the shortest distance home, a strategy known as path integration. Some trail-following ants utilise visual landmark information whereas few of the solitary-foraging ants rely on both path integration and visual landmark information. However, it is unknown to what degree trail-following ants use path integration and we investigated this in a trunk-trail-following ant, Iridomyrmex purpureus. Trunk-trail ants connect their nests to food sites with pheromone trails that contain long-lasting orientation information. We determined the use of visual landmarks and the ability to path integrate in a trunk-trail forming ant. We found that experienced animals switch to relying on visual landmark information, and naïve individuals rely on odour trails. Ants displaced to unfamiliar locations relied on path integration, but, surprisingly, they did not travel the entire homebound distance. We found that as the homebound distance increased, the distance ants travelled relying on the path integrator reduced.”

Card, A., McDermott, C., & Narendra, A. (2016). Multiple orientation cues in an Australian trunk-trail-forming ant, Iridomyrmex purpureus. Australian Journal of Zoology.

Making your own way home

 

Until recently, there was a very neat story regarding the way aerial and terrestrial insects measured distances. Terrestrial insects such as ants use mechanisms based on step-counting, whereas bees use optic-flow based odometry. This is often used as an example of ecological tuning (or situatedness) where the sensory system most appropriate for behavioural control is tuned to the environment. Flying bees are subject to turbulence and so distance measures based on motor output are likely to be unreliable. This is much less of a problem for ants, who can therefore rely on step-counting.

The lovely paper from Pfeffer and Wittlinger uses a unique behavioural situation to test the generality of step-counting mechanisms in ants. Some ant species are polydomous and workers ensure balanced resources by carrying inexperienced ants between nest sites. When a pair of ants are separated, the carried ant returns to the original nest it was carried from. In this paper it is shown that such carried ants do know the distance back to the nest they came from, and don’t use step-counting for odometry when trying to get back to their original nest. In fact they use optic flow. This shows that ants have two odometery mechanisms. More significantly the information from optic-flow is not available to step-counting systems. This is the opposite pattern to the compass system in ants, where information can be transferred between modalities.

Pfeffer, S. E. & Wittlinger, M. 2016 Optic flow odometry operates independently of stride integration in carried ants. Science 353, 1155-1157.

Lifelong radar tracking

This paper almost slipped by without notice, which would have been a real shame because it is technically stunning and shows what might be possible in future navigation studies. Woodgate et al have been able to track the entire foraging history of individual bumblebees. They find that individuals vary in the ratio of exploration to exploitation in their flights. Of course this is interesting from a behavioural ecology perspective and the way in which a colony uses their environment. It is also very interesting in terms of navigation and the potential for some bees to have widespread knowledge of the environment, whereas others may restrict themselves to route like corridors.

Woodgate JL, Makinson JC, Lim KS, Reynolds AM, Chittka L (2016) Life-Long Radar Tracking of Bumblebees. PLoS ONE 11(8): e0160333. doi:10.1371/journal.pone.0160333

Traplining in honeybees

Trapline foraging is observed in many insects, whereby an individual visits a series of locations in a repeatable and consistent order. As a natural behaviour this has been observed in solitary bees visiting sequences of orchids and in parasitic wasps monitoring a series of potential hosts. Experimentally, traplining has been well studied in bumblebees whose foragers are solitary. In this paper, it is shown that honeybee foragers also develop trapline routes in experimental situations, which is interesting for a species with social recruitment. That traplining seems to be a general emergent property of routes visiting multiple locations, reinforces the prevailing idea that the major navigation mode for insect is the development and maintainance of habitual idiosyncratic routes.

Buatois, A., & Lihoreau, M. (2016). Evidence of trapline foraging in honeybees. Journal of Experimental Biology, 219(16), 2426-2429.

Making sense of complex scenes

We know that insects can guide routes using their memories of visual scenes. However we don’t know much about the perceptual processes that might be employed to derive directional information from a visual scene. Whilst it seems unlikely that insects parse scenes into labelled objects in a way that is similar to object recognition in humans, insects may segment panoramas into regions or even ‘shapes’. Buehlmann et al look at the rules that ants might use to derive directional information from panoramas. Using panoramas made up of multiple shapes, they found that ants derive directional information independently from each shape and recognise shapes using their relative position as well as their appearance.

Buehlmann, C., Woodgate, J. L., & Collett, T. S. (2016). On the Encoding of Panoramic Visual Scenes in Navigating Wood Ants. Current Biology, 26(15), 2022-2027.

Learning walks

Since the seminal experiments of Tinbergen, we have known that insects utilise choreographed movements to learn about the appearance of their nest surroundings. These movements are known as learning walks or learning flights. Experiments with bees have shown how a small number of learning/orientation flights can suffice for an individual to successfully home from a wide area. Fleischmann et al. add significantly to our understanding of these behaviours by looking in detail at the ontogeny of learning walks in desert ants. They show that ants complete a significant number of learning walks before foraging commences. Interestingly, they take ants who are in the the middle of their sequence of learning walks and test how much they have learnt about the visual surroundings of the nest. The precision of ants in their search for the fictive nest is quite poor initially. This suggests that visual learning is slow in these ants, or perhaps that early learning walks have a more general purpose, rather than simple visual learning.

Fleischmann, P. N., Christian, M., Müller, V. L., Rössler, W., & Wehner, R. (2016). Ontogeny of learning walks and the acquisition of landmark information in desert ants, Cataglyphis fortis. Journal of Experimental Biology, jeb-140459.