Insect visual homing strategies in a robot with analog processing

 The visual homing abilities of insects can be explained by the snapshot hypothesis. It asserts that an animal is guided to a previously visited location by comparing the current view with a snapshot taken at that location. The average landmark vector (ALV) model is a parsimonious navigation model based on the snapshot hypothesis. According to this model, the target location is unambiguously characterized by a signature vector extracted from the snapshot image. This article provides threefold support for the ALV model by synthetic modeling. First, it was shown that a mobile robot using the ALV model returns to the target location with only small position errors. Second, the behavior of the robot resembled the behavior of bees in some experiments. And third, the ALV model was implemented on the robot in analog hardware. This adds validity to the ALV model, since analog electronic circuits share a number of information-processing principles with biological nervous systems; the analog implementation therefore provides suggestions for how visual homing abilities might be implemented in the insect's brain. Received: 15 June 1999 / Accepted in revised form: 20 March 2000     

Modelling place memory in crickets

Insects can remember and return to a place of interest using the surrounding visual cues. In previous experiments, we showed that crickets could home to an invisible cool spot in a hot environment. They did so most effectively with a natural scene surround, though they were also able to home with distinct landmarks or blank walls. Homing was not successful, however, when visual cues were removed through a dark control. Here, we compare six different models of visual homing using the same visual environments. Only models deemed biologically plausible for use by insects were implemented. The average landmark vector model and first order differential optic flow are unable to home better than chance in at least one of the visual environments. Second order differential optic flow and GradDescent on image differences can home better than chance in all visual environments, and best in the natural scene environment, but do not quantitatively match the distributions of the cricket data. Two models—centre of mass average landmark vector and RunDown on image differences—could produce the same pattern of results as observed for crickets. Both the models performed best using simple binary images and were robust to changes in resolution and image smoothing.     

Simulated visual homing in desert ant natural environments: efficiency of skyline cues

Desert ants, foraging in cluttered semiarid environments, are thought to be visually guided along individual, habitual routes. While other navigational mechanisms (e.g. path integration) are well studied, the question of how ants extract reliable visual features from a complex visual scene is still largely open. This paper explores the assumption that the upper outline of ground objects formed against the sky, i.e. the skyline, provides sufficient information for visual navigation. We constructed a virtual model of the ant’s environment. In the virtual environment, panoramic images were recorded and adapted to the resolution of the desert ant’s complex eye. From these images either a skyline code or a pixel-based intensity code were extracted. Further, two homing algorithms were implemented, a modified version of the average landmark vector (ALV) model (Lambrinos et al. Robot Auton Syst 30:39–64, 2000) and a gradient ascent method. Results show less spatial aliasing for skyline coding and best homing performance for ALV homing based on skyline codes. This supports the assumption of skyline coding in visual homing of desert ants and allows novel approaches to technical outdoor navigation.     

Sex-inducing effect of a hydrophilic fraction on reproductive switching in the planarian Dugesia ryukyuensis (Seriata, Tricladida)

Background

Insects are known to rely on terrestrial landmarks for navigation. Landmarks are used to chart a route or pinpoint a goal. The distant panorama, however, is often thought not to guide navigation directly during a familiar journey, but to act as a contextual cue that primes the correct memory of the landmarks.

Results

We provided Melophorus bagoti ants with a huge artificial landmark located right near the nest entrance to find out whether navigating ants focus on such a prominent visual landmark for homing guidance. When the landmark was displaced by small or large distances, ant routes were affected differently. Certain behaviours appeared inconsistent with the hypothesis that guidance was based on the landmark only. Instead, comparisons of panoramic images recorded on the field, encompassing both landmark and distal panorama, could explain most aspects of the ant behaviours.

Conclusion

Ants navigating along a familiar route do not focus on obvious landmarks or filter out distal panoramic cues, but appear to be guided by cues covering a large area of their panoramic visual field, including both landmarks and distal panorama. Using panoramic views seems an appropriate strategy to cope with the complexity of natural scenes and the poor resolution of insects' eyes. The ability to isolate landmarks from the rest of a scene may be beyond the capacity of animals that do not possess a dedicated object-perception visual stream like primates.     

Some psychophysics of the pigeon's use of landmarks

1. Three pigeons (Columba livia) were trained to find hidden food in a sunken well (3.3 cm in diameter) at a constant place within an (160 cm x 160 cm) experimental box (Fig. 1). After learning the location, the animals were tested occasionally with the well and food absent. Landmarks in the experimental box might be transformed on such tests. 2. Changing the height or width of a nearby landmark had no systematic influence on the position of peak search. Translating a nearby landmark, however, led to a shift in peak search position. All three birds then searched most somewhere between the original goal location, as defined by the unmoved landmarks, and the goal location as defined by the shifted landmark. Within a limited range of landmark shift, the peak shift as a function of landmark shift is linear (Fig. 3). 3. To explain the data (Fig. 7), the pigeon records at the location of the goal the algebraic vectors from a number of landmarks to the goal. These vectors have both a direction and a distance component. When searching for the goal again in the experimental box, it computes independently for each landmark a navigation vector. This is arrived at by vector-adding the algebraic vector from the bird's current position to the landmark in question, supplied by perception, to the corresponding landmark-goal vector in its record. The pigeon moves in the direction and distance specified by a weighted average of the independently calculated navigation vectors. For positive vector weights, vector geometry guarantees that the bird would search somewhere between the original goal and the goal according to the shifted landmark. The extent to which it shifts toward the shifted goal reflects the vector weight given to the shifted landmark.     

Where did I take that snapshot? Scene-based homing by image matching

In homing tasks, the goal is often not marked by visible objects but must be inferred from the spatial relation to the visual cues in the surrounding scene. The exact computation of the goal direction would require knowledge about the distances to visible landmarks, information, which is not directly available to passive vision systems. However, if prior assumptions about typical distance distributions are used, a snapshot taken at the goal suffices to compute the goal direction from the current view. We show that most existing approaches to scene-based homing implicitly assume an isotropic landmark distribution. As an alternative, we propose a homing scheme that uses parameterized displacement fields. These are obtained from an approximation that incorporates prior knowledge about perspective distortions of the visual environment. A mathematical analysis proves that both approximations do not prevent the schemes from approaching the goal with arbitrary accuracy, but lead to different errors in the computed goal direction. Mobile robot experiments are used to test the theoretical predictions and to demonstrate the practical feasibility of the new approach. Received: 11 December 1997 / Accepted in revised form: 12 June 1998     

A model of head direction and landmark coding in complex environments

Environmental information is required to stabilize estimates of head direction (HD) based on angular path integration. However, it is unclear how this happens in real-world (visually complex) environments. We present a computational model of how visual feedback can stabilize HD information in environments that contain multiple cues of varying stability and directional specificity. We show how combinations of feature-specific visual inputs can generate a stable unimodal landmark bearing signal, even in the presence of multiple cues and ambiguous directional specificity. This signal is associated with the retrosplenial HD signal (inherited from thalamic HD cells) and conveys feedback to the subcortical HD circuitry. The model predicts neurons with a unimodal encoding of the egocentric orientation of the array of landmarks, rather than any one particular landmark. The relationship between these abstract landmark bearing neurons and head direction cells is reminiscent of the relationship between place cells and grid cells. Their unimodal encoding is formed from visual inputs via a modified version of Oja’s Subspace Algorithm. The rule allows the landmark bearing signal to disconnect from directionally unstable or ephemeral cues, incorporate newly added stable cues, support orientation across many different environments (high memory capacity), and is consistent with recent empirical findings on bidirectional HD firing reported in the retrosplenial cortex. Our account of visual feedback for HD stabilization provides a novel perspective on neural mechanisms of spatial navigation within richer sensory environments, and makes experimentally testable predictions.     

Spatial Orientation in Blattella germanica L. Larvae

The aim of these experiments was to investigate the type of cues used in homing processes by young Blattella germanica L. larvae. Several types of stimuli were tested: path integration with kinesthetic cues and visual orientation with landmark cues. Tests measured the escape direction of larvae from the food box after disturbance. Either type of cue alone, path integration or visual landmarks, was sufficient to allow larvae to orient towards their shelters, but they oriented more precisély when both types of cue were used. When several landmark cues (proximal and distal) were present, their relative angular position seemed important in the orientation process. Macroscopic shapes in the environment appeared to be used as a global image, memorized to reach the shelter.     

Path integration — a network model

Path integration is a primary means of navigation for a number of animals. We present a model which performs path integration with a neural network. This model is based on a neural structure called a sinusoidal array, which allows an efficient representation of vector information with neurons. We show that exact path integration can easily be achieved by a neural network. Thus deviations from the direct home trajectory, found previously in experiments with ants, can not be explained by computational limitations of the nervous system. Instead we suggest that the observed deviations are caused by a strategy to simplify landmark navigation.     

Mapping the navigational knowledge of individually foraging ants,Myrmecia croslandi

Ants are efficient navigators, guided by path integration and visual landmarks. Path integration is the primary strategy in landmark-poor habitats, but landmarks are readily used when available. The landmark panorama provides reliable information about heading direction, routes and specific location. Visual memories for guidance are often acquired along routes or near to significant places. Over what area can such locally acquired memories provide information for reaching a place? This question is unusually approachable in the solitary foraging Australian jack jumper ant, since individual foragers typically travel to one or two nest-specific foraging trees. We find that within 10 m from the nest, ants both with and without home vector information available from path integration return directly to the nest from all compass directions, after briefly scanning the panorama. By reconstructing panoramic views within the successful homing range, we show that in the open woodland habitat of these ants, snapshot memories acquired close to the nest provide sufficient navigational information to determine nest-directed heading direction over a surprisingly large area, including areas that animals may have not visited previously.     

基于EMDs阵列和BP网络的视觉运动感知过程

以Ｒｅｉｃｈａｒｄｔ的相关型初级运动检测器阵列和Ｒｕｍｅｌｈａｒｔ的误差反传学习（ｌｅａｒｎｉｎｇｂｙｂａｃｋ－ｐｒｏｐａｇａｔｉｎｇｅｒｒｏｒｓ,ＢＰ）网络相结合构成了一个视觉运动感知神经网络,探讨了视觉运动信息的感知过程。试图从计算神经科学的观点来阐明从一推运动分量的检测到二维模式运动感知的神经原理,从而回答运动矢量在脑内如何表征。计算机仿真表明,在有监督学习的条件下,网络可以学会解决局城运动检测所带来的多义性问题,给出模式的真实朝向、运动方向和运动速度。     

Neural mechanisms of tactile motion integration in somatosensory cortex

How are local motion signals integrated to form a global motion percept? We investigate the neural mechanisms of tactile motion integration by presenting tactile gratings and plaids to the fingertips of monkeys, using the tactile analogue of a visual monitor and recording the responses evoked in somatosensory cortical neurons. The perceived directions of the gratings and plaids are measured in parallel psychophysical experiments. We identify a population of somatosensory neurons that exhibit integration properties comparable to those induced by analogous visual stimuli in area MT and find that these neural responses account for the perceived direction of the stimuli across all stimulus conditions tested. The preferred direction of the neurons and the perceived direction of the stimuli can be predicted from the weighted average of the directions of the individual stimulus features, highlighting that the somatosensory system implements a vector average mechanism to compute tactile motion direction that bears striking similarities to its visual counterpart.     

Using artificial evolution and selection to model insect navigation.

BACKGROUND: An animal's behavioral strategies are often constrained by its evolutionary history and the resources available to it. Artificial evolution allows one to manipulate such constraints and explore how they influence evolved strategies. Here we compare the navigational strategies of flying insects with those of artificially evolved "animats" endowed with various motor architectures. Using evolutionary algorithms, we generated artificial neural networks that controlled a virtual animat's navigation within a 2D, simulated world. Like a flying insect, the animat possessed motors that generated thrust and torque, a compass, and visual sensors. Some animats were limited to forward motion, while others could also move sideways. Animats were selected for the precision with which they reached a target specified by a visual landmark. RESULTS: Animats given sideways motors could alter flight direction without changing body orientation and evolved strategies similar to those of flying bees or wasps performing the same task. Both animats and insects first aimed at the landmark. In the last phase, both adopted a fixed body orientation and adjusted their position to keep the landmark at a fixed retinal location. Animats unable to uncouple flight direction and body orientation evolved subtly different strategies and performed less robustly. CONCLUSIONS: This convergence between the navigational strategies of animals and animats suggests that the insect's strategies are primarily an adaptation to the demands of using visual information and compass direction to reach a position in space and that they are not significantly compromised by the insect's evolutionary history.     

Ground-nesting bees determine the location of their nest relative to a landmark by other than angular size cues

Bees and wasps acquire a visual representation of their nest's environment and use it to locate their nest when they return from foraging trips. This representation contains among other features cues to the distance of near-by landmarks. We worked with two species of ground-nesting bees, Lasioglossum malachurum (Hymenoptera: Halictidae), Dasypoda hirtipes (Hymenoptera: Melittidae) and asked which cues to landmark distance they use during homing. Bees learned to associate a single cylindrical landmark with their nest's location. We subsequently tested returning bees with landmarks of different sizes and thus introduced large discrepancies between the angular size of the landmark as seen from the nest during training and its distance from the nest. The bees' search behaviour and their choice of dummy nest entrances show that both species of ground-nesting bees consistently search for their nest at the learned distance from landmarks. The influence of the apparent size of landmarks on the bees' search and choice behaviour is comparatively weak. We suggest that the bees exploit cues derived from the apparent speed of the landmark's image at their retina for distance evaluation.     

An inference upon the neural network finding binocular correspondence

Previously, the authors proposed a model of neural network extracting binocular parallax (Hirai and Fukushima, 1975). It is a multilayered network whose final layers consist of neural elements corresponding to binocular depth neurons found in monkey's visual cortex. The binocular depth neuron is selectively sensitive to a binocular stimulus with a specific amount of binocular parallax and does not respond to a monocular one. As described in the last chapter of the previous article (Hirai and Fukushima, 1975), when a binocular pair of input patterns consist of, for example, many vertical bars placed very closely to each other, the binocular depth neurons might respond not only to correct binocular pairs, but also to incorrect ones. Our present study is concentrated upon how the visual system finds correct binocular pairs or binocular correspondence. It is assumed that some neural network is cascaded after the binocular depth neurons and finds out correct binocular correspondence by eliminating the incorrect binocular pairs. In this article a model of such neural network is proposed. The performance of the model has been simulated on a digital computer. The results of the computer simulation show that this model finds binocular correspondence satisfactorily. It has been demonstrated by the computer simulation that this model also explains the mechanism of the hysteresis in the binocular depth perception reported by Fender and Julesz (1967)This work has been done in the NHK Broadcasting Science Research Laboratories     

A unified mechanism for innate and learned visual landmark guidance in the insect central complex

Insects can navigate efficiently in both novel and familiar environments, and this requires flexiblity in how they are guided by sensory cues. A prominent landmark, for example, can elicit strong innate behaviours (attraction or menotaxis) but can also be used, after learning, as a specific directional cue as part of a navigation memory. However, the mechanisms that allow both pathways to co-exist, interact or override each other are largely unknown. Here we propose a model for the behavioural integration of innate and learned guidance based on the neuroanatomy of the central complex (CX), adapted to control landmark guided behaviours. We consider a reward signal provided either by an innate attraction to landmarks or a long-term visual memory in the mushroom bodies (MB) that modulates the formation of a local vector memory in the CX. Using an operant strategy for a simulated agent exploring a simple world containing a single visual cue, we show how the generated short-term memory can support both innate and learned steering behaviour. In addition, we show how this architecture is consistent with the observed effects of unilateral MB lesions in ants that cause a reversion to innate behaviour. We suggest the formation of a directional memory in the CX can be interpreted as transforming rewarding (positive or negative) sensory signals into a mapping of the environment that describes the geometrical attractiveness (or repulsion). We discuss how this scheme might represent an ideal way to combine multisensory information gathered during the exploration of an environment and support optimal cue integration.     

Path integration in desert ants, Cataglyphis: how to make a homing ant run away from home

Path integration is an ant's lifeline on any of its foraging journeys. It results in a homebound global vector that continually informs the animal about its position relative to its starting point. Here, we use a particular (repeated training and displacement) paradigm, in which homebound ants are made to follow a familiar landmark route repeatedly from the feeder to the nest, even after they have arrived at the nest. The results show that during the repeated landmark-guided home runs the ant's path integrator runs continually, so that the current state of the homebound vector increasingly exceeds the reference state. The dramatic result is that the homing ants run away from home. This finding implies that the ants do not rely on cartographic information about the locations of nest and feeder (e.g. that the nest is always south of the feeder), but just behave according to what the state of their egocentric path integrator tells them.     

The ant’s estimation of distance travelled: experiments with desert ants,<Emphasis Type="Italic"> Cataglyphis fortis</Emphasis>

Foraging desert ants, Cataglyphis fortis, monitor their position relative to the nest by path integration. They continually update the direction and distance to the nest by employing a celestial compass and an odometer. In the present account we addressed the question of how the precision of the ants estimate of its homing distance depends on the distance travelled. We trained ants to forage at different distances in linear channels comprising a nest entrance and a feeder. For testing we caught ants at the feeder and released them in a parallel channel. The results show that ants tend to underestimate their distances travelled. This underestimation is the more pronounced, the larger the foraging distance gets. The quantitative relationship between training distance and the ants estimate of this distance can be described by a logarithmic and an exponential model. The ants odometric undershooting could be adaptive during natural foraging trips insofar as it leads the homing ant to concentrate the major part of its nest-search behaviour on the base of its individual foraging sector, i.e. on its familiar landmark corridor.     

Landmark cues can change the motivational state of desert ant foragers

Desert ants of the genus Cataglyphis rely on path integration vectors to return to the nest (inbound runs) and back to frequently visited feeding sites (outbound runs). If disturbed, e.g., experimentally displaced on their inbound runs, they continue to run off their home-bound vector, but if disturbed in the same way on their outbound runs, they do not continue their feeder-based vector, but immediately switch on the home-bound state of their path integration vector and return to the nest. Here we show that familiar landmarks encountered by the ants during their run towards the feeder can change the ants’ motivational state insofar that the ants even if disturbed continue to run in the nest-to-feeder direction rather than reverse their courses, as they do in landmark-free situations. Hence, landmark cues can cause the ants to change their motivational state from homing to foraging.     

Landmark orientation during the approach to the nest in the stingless beeTrigona (Tetragonisca) angustula (Apidae,Meliponinae)

Summary We displaced a small nest box containing stingless bees (Trigona (Tetragonisca)angustula) over distances of up to 1.6 meters in different directions and counted the numbers of returning foragers to measure the effects of this manipulation on the homing ability of bees. Bees find it hard to locate the nest box when it was displaced more than about 1 m backwards, forwards or sideways relative to the direction into which the nest entrance pointed. They do not find the nest when its height above ground is changed. The bees use landmarks in the vicinity of the nest to locate it: When the nest box is displaced and landmark positions are changed so that their angular position at the new nest site is the same as at the normal nest position their homing ability is less impaired than it is without changes in landmark positions. Our results show that the bees do not use the nest box itself as a landmark until they have approached the nest position to within about 1 meter with the aid of surrounding landmarks.