How bumblebees coordinate path integration and body orientation at the start of their first learning flight.

The start of a bumblebee's first learning flight from its nest provides an opportunity to examine the bee's learning behaviour during its initial view of the nest's unfamiliar surroundings. Like many other hymenopterans, bumblebees store views of their nest surroundings while facing their nest. We found that a bumblebee's first fixation of the nest is a coordinated manoeuvre in which the insect faces the nest with its body oriented towards a particular visual feature within its surroundings. This conjunction of nest fixation and body orientation is preceded and reached by means of a translational scan during which the bee flies perpendicularly to its preferred body orientation. The utility of the coordinated manoeuvre is apparent during the bees' first return flight after foraging. Bees then adopt a similar preferred body orientation when close to the nest. How does a bee, unacquainted with its surroundings, know when it is facing its nest? A likely answer is through path integration, which gives bees continuously updated information about the current direction of their nest. Path integration also gives bees the possibility to fixate the nest when their body points in a desired direction. The three components of this coordinated manoeuvre are discussed in relation to current understanding of the central complex in the insect brain, noting that nest fixation is egocentric, whereas the preferred body orientation and flight direction that the bee adopts within the visual surroundings of the nest are geocentric.

Wood ants learn the magnetic direction of a route but express uncertainty because of competing directional cues.

Wood ants were trained indoors to follow a magnetically specified route that went from the centre of an arena to a drop of sucrose at the edge. The arena, placed in a white cylinder, was in the centre of a 3D coil system generating an inclined Earth-strength magnetic field in any horizontal direction. The specified direction was rotated between each trial. The ants' knowledge of the route was tested in trials without food. Tests given early in the day, before any training, show that ants remember the magnetic route direction overnight. During the first 2 s of a test, ants mostly faced in the specified direction, but thereafter were often misdirected, with a tendency to face briefly in the opposite direction. Uncertainty about the correct path to take may stem in part from competing directional cues linked to the room. In addition to facing along the route, there is evidence that ants develop magnetically directed home and food vectors dependent upon path integration. A second experiment asked whether ants can use magnetic information contextually. In contrast to honeybees given a similar task, ants failed this test. Overall, we conclude that magnetic directional cues can be sufficient for route learning.

Evolved Transistor Array Robot Controllers.

For the first time, a field programmable transistor array (FPTA) was used to evolve robot control circuits directly in analog hardware. Controllers were successfully incrementally evolved for a physical robot engaged in a series of visually guided behaviours, including finding a target in a complex environment where the goal was hidden from most locations. Circuits for recognising spoken commands were also evolved and these were used in conjunction with the controllers to enable voice control of the robot, triggering behavioural switching. Poor quality visual sensors were deliberately used to test the ability of evolved analog circuits to deal with noisy uncertain data in realtime. Visual features were coevolved with the controllers to automatically achieve dimensionality reduction and feature extraction and selection in an integrated way. An efficient new method was developed for simulating the robot in its visual environment. This allowed controllers to be evaluated in a simulation connected to the FPTA. The controllers then transferred seamlessly to the real world. The circuit replication issue was also addressed in experiments where circuits were evolved to be able to function correctly in multiple areas of the FPTA. A methodology was developed to analyse the evolved circuits which provided insights into their operation. Comparative experiments demonstrated the superior evolvability of the transistor array medium.

Neural coding in the visual system of Drosophila melanogaster: How do small neural populations support visually guided behaviours?

All organisms wishing to survive and reproduce must be able to respond adaptively to a complex, changing world. Yet the computational power available is constrained by biology and evolution, favouring mechanisms that are parsimonious yet robust. Here we investigate the information carried in small populations of visually responsive neurons in Drosophila melanogaster. These so-called 'ring neurons', projecting to the ellipsoid body of the central complex, are reported to be necessary for complex visual tasks such as pattern recognition and visual navigation. Recently the receptive fields of these neurons have been mapped, allowing us to investigate how well they can support such behaviours. For instance, in a simulation of classic pattern discrimination experiments, we show that the pattern of output from the ring neurons matches observed fly behaviour. However, performance of the neurons (as with flies) is not perfect and can be easily improved with the addition of extra neurons, suggesting the neurons' receptive fields are not optimised for recognising abstract shapes, a conclusion which casts doubt on cognitive explanations of fly behaviour in pattern recognition assays. Using artificial neural networks, we then assess how easy it is to decode more general information about stimulus shape from the ring neuron population codes. We show that these neurons are well suited for encoding information about size, position and orientation, which are more relevant behavioural parameters for a fly than abstract pattern properties. This leads us to suggest that in order to understand the properties of neural systems, one must consider how perceptual circuits put information at the service of behaviour.

Unsupervised Learning in an Ensemble of Spiking Neural Networks Mediated by ITDP.

We propose a biologically plausible architecture for unsupervised ensemble learning in a population of spiking neural network classifiers. A mixture of experts type organisation is shown to be effective, with the individual classifier outputs combined via a gating network whose operation is driven by input timing dependent plasticity (ITDP). The ITDP gating mechanism is based on recent experimental findings. An abstract, analytically tractable model of the ITDP driven ensemble architecture is derived from a logical model based on the probabilities of neural firing events. A detailed analysis of this model provides insights that allow it to be extended into a full, biologically plausible, computational implementation of the architecture which is demonstrated on a visual classification task. The extended model makes use of a style of spiking network, first introduced as a model of cortical microcircuits, that is capable of Bayesian inference, effectively performing expectation maximization. The unsupervised ensemble learning mechanism, based around such spiking expectation maximization (SEM) networks whose combined outputs are mediated by ITDP, is shown to perform the visual classification task well and to generalize to unseen data. The combined ensemble performance is significantly better than that of the individual classifiers, validating the ensemble architecture and learning mechanisms. The properties of the full model are analysed in the light of extensive experiments with the classification task, including an investigation into the influence of different input feature selection schemes and a comparison with a hierarchical STDP based ensemble architecture.

How do field of view and resolution affect the information content of panoramic scenes for visual navigation? A computational investigation.

The visual systems of animals have to provide information to guide behaviour and the informational requirements of an animal's behavioural repertoire are often reflected in its sensory system. For insects, this is often evident in the optical array of the compound eye. One behaviour that insects share with many animals is the use of learnt visual information for navigation. As ants are expert visual navigators it may be that their vision is optimised for navigation. Here we take a computational approach in asking how the details of the optical array influence the informational content of scenes used in simple view matching strategies for orientation. We find that robust orientation is best achieved with low-resolution visual information and a large field of view, similar to the optical properties seen for many ant species. A lower resolution allows for a trade-off between specificity and generalisation for stored views. Additionally, our simulations show that orientation performance increases if different portions of the visual field are considered as discrete visual sensors, each giving an independent directional estimate. This suggests that ants might benefit by processing information from their two eyes independently.

Acquisition and expression of memories of distance and direction in navigating wood ants.

Wood ants, like other central place foragers, rely on route memories to guide them to and from a reliable food source. They use visual memories of the surrounding scene and probably compass information to control their direction. Do they also remember the length of their route and do they link memories of direction and distance? To answer these questions, we trained wood ant (Formica rufa) foragers in a channel to perform either a single short foraging route or two foraging routes in opposite directions. By shifting the starting position of the route within the channel, but keeping the direction and distance fixed, we tried to ensure that the ants would rely upon vector memories rather than visual memories to decide when to stop. The homeward memories that the ants formed were revealed by placing fed or unfed ants directly into a channel and assessing the direction and distance that they walked without prior performance of the food-ward leg of the journey. This procedure prevented the distance and direction walked being affected by a home vector derived from path integration. Ants that were unfed walked in the feeder direction. Fed ants walked in the opposite direction for a distance related to the separation between start and feeder. Vector memories of a return route can thus be primed by the ants' feeding state and expressed even when the ants have not performed the food-ward route. Tests on ants that have acquired two routes indicate that memories of the direction and distance of the return routes are linked, suggesting that they may be encoded by a common neural population within the ant brain.

From Mindless Masses to Small Groups: Conceptualizing Collective Behavior in Crowd Modeling.

Computer simulations are increasingly used to monitor and predict behavior at large crowd events, such as mass gatherings, festivals and evacuations. We critically examine the crowd modeling literature and call for future simulations of crowd behavior to be based more closely on findings from current social psychological research. A systematic review was conducted on the crowd modeling literature (N = 140 articles) to identify the assumptions about crowd behavior that modelers use in their simulations. Articles were coded according to the way in which crowd structure was modeled. It was found that 2 broad types are used: mass approaches and small group approaches. However, neither the mass nor the small group approaches can accurately simulate the large collective behavior that has been found in extensive empirical research on crowd events. We argue that to model crowd behavior realistically, simulations must use methods which allow crowd members to identify with each other, as suggested by self-categorization theory.

Validation of an iPad visual analogue rating system for assessing appetite and satiety.

The study aimed to validate appetite ratings made on a new electronic device, the Apple iPad Mini, against an existing but now obsolete electronic device (Hewlett Packard iPAQ). Healthy volunteers (9 men and 9 women) rated their appetite before and 0, 30, 60, 90 and 120 minutes after consuming both a low energy (LE: 77 kcal) and high energy (HE: 274 kcal) beverage at breakfast on 2 non-consecutive days in counter-balanced order. Rated hunger, desire to eat and how much participants could consume was significantly lower after HE than LE on both devices, although there was better overall differentiation between HE and LE for ratings on iPad. Rated satiation and fullness, and a composite measure combining all five ratings, was significantly higher after HE than LE on both devices. There was also evidence that differences between conditions were more significant when analysed at each time point than using an overall area under the curve (AUC) measure. Overall, these data confirm that appetite ratings made using iPad are at least as sensitive as those on iPAQ, and offer a new platform for researchers to collect appetite data.

Visual scanning behaviours and their role in the navigation of the Australian desert ant Melophorus bagoti.

Ants are excellent navigators, using a combination of innate strategies and learnt information to guide habitual routes. The mechanisms underlying this behaviour are little understood though one avenue of investigation is to explore how innate sensori-motor routines are used to accomplish route navigation. For instance, Australian desert ant foragers are occasionally observed to cease translation and rotate on the spot. Here, we investigate this behaviour using high-speed videography and computational analysis. We find that scanning behaviour is saccadic with pauses separated by fast rotations. Further, we have identified four situations where scanning is typically displayed: (1) by naïve ants on their first departure from the nest; (2) by experienced ants departing from the nest for their first foraging trip of the day; (3) by experienced ants when the familiar visual surround was experimentally modified, in which case frequency and duration of scans were proportional to the degree of modification; (4) when the information from visual cues is at odds with the direction indicated by the ant's path integration system. Taken together, we see a general relationship between scanning behaviours and periods of uncertainty.

Head movements and the optic flow generated during the learning flights of bumblebees.

Insects inform themselves about the 3D structure of their surroundings through motion parallax. During flight, they often simplify this task by minimising rotational image movement. Coordinated head and body movements generate rapid shifts of gaze separated by periods of almost zero rotational movement, during which the distance of objects from the insect can be estimated through pure translational optic flow. This saccadic strategy is less appropriate for assessing the distance between objects. Bees and wasps face this problem when learning the position of their nest-hole relative to objects close to it. They acquire the necessary information during specialised flights performed on leaving the nest. Here, we show that the bumblebee's saccadic strategy differs from other reported cases. In the fixations between saccades, a bumblebee's head continues to turn slowly, generating rotational flow. At specific points in learning flights these imperfect fixations generate a form of 'pivoting parallax', which is centred on the nest and enhances the visibility of features near the nest. Bumblebees may thus utilize an alternative form of motion parallax to that delivered by the standard 'saccade and fixate' strategy in which residual rotational flow plays a role in assessing the distances of objects from a focal point of interest.

Investigating visual navigation using spiking neural network models of the insect mushroom bodies.

Ants are capable of learning long visually guided foraging routes with limited neural resources. The visual scene memory needed for this behaviour is mediated by the mushroom bodies; an insect brain region important for learning and memory. In a visual navigation context, the mushroom bodies are theorised to act as familiarity detectors, guiding ants to views that are similar to those previously learned when first travelling along a foraging route. Evidence from behavioural experiments, computational studies and brain lesions all support this idea. Here we further investigate the role of mushroom bodies in visual navigation with a spiking neural network model learning complex natural scenes. By implementing these networks in GeNN-a library for building GPU accelerated spiking neural networks-we were able to test these models offline on an image database representing navigation through a complex outdoor natural environment, and also online embodied on a robot. The mushroom body model successfully learnt a large series of visual scenes (400 scenes corresponding to a 27 m route) and used these memories to choose accurate heading directions during route recapitulation in both complex environments. Through analysing our model's Kenyon cell (KC) activity, we were able to demonstrate that KC activity is directly related to the respective novelty of input images. Through conducting a parameter search we found that there is a non-linear dependence between optimal KC to visual projection neuron (VPN) connection sparsity and the length of time the model is presented with an image stimulus. The parameter search also showed training the model on lower proportions of a route generally produced better accuracy when testing on the entire route. We embodied the mushroom body model and comparator visual navigation algorithms on a Quanser Q-car robot with all processing running on an Nvidia Jetson TX2. On a 6.5 m route, the mushroom body model had a mean distance to training route (error) of 0.144 ± 0.088 m over 5 trials, which was performance comparable to standard visual-only navigation algorithms. Thus, we have demonstrated that a biologically plausible model of the ant mushroom body can navigate complex environments both in simulation and the real world. Understanding the neural basis of this behaviour will provide insight into how neural circuits are tuned to rapidly learn behaviourally relevant information from complex environments and provide inspiration for creating bio-mimetic computer/robotic systems that can learn rapidly with low energy requirements.

Snapshots in ants? New interpretations of paradigmatic experiments.

Ants can use visual information to guide long idiosyncratic routes and accurately pinpoint locations in complex natural environments. It has often been assumed that the world knowledge of these foragers consists of multiple discrete views that are retrieved sequentially for breaking routes into sections controlling approaches to a goal. Here we challenge this idea using a model of visual navigation that does not store and use discrete views to replicate the results from paradigmatic experiments that have been taken as evidence that ants navigate using such discrete snapshots. Instead of sequentially retrieving views, the proposed architecture gathers information from all experienced views into a single memory network, and uses this network all along the route to determine the most familiar heading at a given location. This algorithm is consistent with the navigation of ants in both laboratory and natural environments, and provides a parsimonious solution to deal with visual information from multiple locations.

Bumblebee calligraphy: the design and control of flight motifs in the learning and return flights of Bombus terrestris.

Many wasps and bees learn the position of their nest relative to nearby visual features during elaborate 'learning' flights that they perform on leaving the nest. Return flights to the nest are thought to be patterned so that insects can reach their nest by matching their current view to views of their surroundings stored during learning flights. To understand how ground-nesting bumblebees might implement such a matching process, we have video-recorded the bees' learning and return flights and analysed the similarities and differences between the principal motifs of their flights. Loops that take bees away from and bring them back towards the nest are common during learning flights and less so in return flights. Zigzags are more prominent on return flights. Both motifs tend to be nest based. Bees often both fly towards and face the nest in the middle of loops and at the turns of zigzags. Before and after flight direction and body orientation are aligned, the two diverge from each other so that the nest is held within the bees' fronto-lateral visual field while flight direction relative to the nest can fluctuate more widely. These and other parallels between loops and zigzags suggest that they are stable variations of an underlying pattern, which enable bees to store and reacquire similar nest-focused views during learning and return flights.

Coordinating compass-based and nest-based flight directions during bumblebee learning and return flights.

Bumblebees tend to face their nest over a limited range of compass directions when learning the nest's location on departure and finding it on their approach after foraging. They thus obtain similar views of the nest and its surroundings on their learning and return flights. How do bees coordinate their flights relative to nest-based and compass-based reference frames to get such similar views? We show, first, that learning and return flights contain straight segments that are directed along particular compass bearings, which are independent of the orientation of a bee's body. Bees are thus free within limits to adjust their viewing direction relative to the nest, without disturbing flight direction. Second, we examine the coordination of nest-based and compass-based control during likely information gathering segments of these flights: loops during learning flights and zigzags on return flights. We find that bees tend to start a loop or zigzag when flying within a restricted range of compass directions and to fly towards the nest and face it after a fixed change in compass direction, without continuous interactions between their nest-based and compass-based directions of flight. A preferred trajectory of compass-based flight over the course of a motif, combined with the tendency of the bees to keep their body oriented towards the nest automatically narrows the range of compass directions over which bees view the nest. Additionally, the absence of interactions between the two reference frames allows loops and zigzags to have a stereotyped form that can generate informative visual feedback.

A model of ant route navigation driven by scene familiarity.

In this paper we propose a model of visually guided route navigation in ants that captures the known properties of real behaviour whilst retaining mechanistic simplicity and thus biological plausibility. For an ant, the coupling of movement and viewing direction means that a familiar view specifies a familiar direction of movement. Since the views experienced along a habitual route will be more familiar, route navigation can be re-cast as a search for familiar views. This search can be performed with a simple scanning routine, a behaviour that ants have been observed to perform. We test this proposed route navigation strategy in simulation, by learning a series of routes through visually cluttered environments consisting of objects that are only distinguishable as silhouettes against the sky. In the first instance we determine view familiarity by exhaustive comparison with the set of views experienced during training. In further experiments we train an artificial neural network to perform familiarity discrimination using the training views. Our results indicate that, not only is the approach successful, but also that the routes that are learnt show many of the characteristics of the routes of desert ants. As such, we believe the model represents the only detailed and complete model of insect route guidance to date. What is more, the model provides a general demonstration that visually guided routes can be produced with parsimonious mechanisms that do not specify when or what to learn, nor separate routes into sequences of waypoints.

DoPI: The Database of Pollinator Interactions.

Despite the importance of pollinating insects to natural environments and agriculture, there have been few attempts to unite the existing plant-pollinator interaction datasets into a single depository using a common format. Accordingly, we have created one of the world's first online, open-access, and searchable pollinator-plant interaction databases. DoPI (The Database of Pollinator Interactions) was built from a systematic review of the scientific literature and unpublished datasets requested from researchers and organizations. We collated records of interactions between British plant and insect flower-visitor species (or genera), together with associated metadata (date, location, habitat, source publication) when available. The dataset currently (December 2021) contains 101,539 records, detailing over 320,000 interactions. The number of interactions (i.e., the number of times a pairwise species interaction was recorded per occasion) varies considerably among records, averaging 3.6. These include records from 1888 pollinator species and 1241 plant species, totaling >17,000 pairwise species interactions. By combining a large volume of information in a single repository, DoPI can be used to answer fundamental ecological questions on the dynamics of pollination interactions in space and time, as well as applied questions in conservation practice. We hope this dynamic database will be a useful tool not only for researchers, but also for conservationists, funding agencies, governmental departments, beekeepers, agronomists, and gardeners. We request that this paper is cited when using the data in publications and individual studies when appropriate. Researchers and organizations are encouraged to add further data in the future. The database can be accessed at: https://www.dopi.org.uk/.

How might ants use panoramic views for route navigation?

Studies of insect navigation have demonstrated that insects possess an interesting and sophisticated repertoire of visual navigation behaviours. Ongoing research seeks to help us understand how these behaviours are controlled in natural complex environments. A necessary complement to behavioural studies is an understanding of the sensory ecology within which an animal behaves. To this end we have analysed ants'-perspective views of a habitat within which desert ant navigation is well studied. Results from our analysis suggest that: parsimonious visual strategies for homing and route guidance are effective over behaviourally useful distances even in cluttered environments; that these strategies can function effectively using only the skyline heights as input; and that the simplicity and efficacy of using stored views as a visual compass makes it a viable and robust mechanism for route guidance.

Dual coding with STDP in a spiking recurrent neural network model of the hippocampus.

The firing rate of single neurons in the mammalian hippocampus has been demonstrated to encode for a range of spatial and non-spatial stimuli. It has also been demonstrated that phase of firing, with respect to the theta oscillation that dominates the hippocampal EEG during stereotype learning behaviour, correlates with an animal's spatial location. These findings have led to the hypothesis that the hippocampus operates using a dual (rate and temporal) coding system. To investigate the phenomenon of dual coding in the hippocampus, we examine a spiking recurrent network model with theta coded neural dynamics and an STDP rule that mediates rate-coded Hebbian learning when pre- and post-synaptic firing is stochastic. We demonstrate that this plasticity rule can generate both symmetric and asymmetric connections between neurons that fire at concurrent or successive theta phase, respectively, and subsequently produce both pattern completion and sequence prediction from partial cues. This unifies previously disparate auto- and hetero-associative network models of hippocampal function and provides them with a firmer basis in modern neurobiology. Furthermore, the encoding and reactivation of activity in mutually exciting Hebbian cell assemblies demonstrated here is believed to represent a fundamental mechanism of cognitive processing in the brain.