Selforganizing memory: active learning of landmarks used for navigation

We propose a memory architecture that is suited to solve a specific task, namely homing, that is finding a not directly visible home place by using visually accessible landmarks. We show that an agent equipped with such a memory structure can autonomously learn the situation and can later use its memory to accomplish homing behaviour. The architecture is based on neuronal structures and grows in a self-organized way depending on experience. The basic architecture consists of three parts, (i) a pre-processor, (ii) a simple, one-layered feed-forward network, called distributor net, and (iii) a full recurrently connected net for representing the situation models to be stored. Apart from Hebbian learning and a local version of the delta-rule, explorative learning is applied that is not based on passive detection of correlations, but is actively searching for interesting hypotheses. Hypotheses are spontaneously introduced and are verified or falsified depending on how well the network representing the hypothesis approaches an internal error of zero. The stability of this approach is successfully tested by removal of one landmark or shifting the position of one or several landmarks showing results comparable to those found in biological experiments. Furthermore, we applied noise in two ways. The trained network was either due to sensory noise or to noise applied to the bias weights describing the memory content. Finally, we tested to what extent learning of the weights is affected by noisy input given to the sensor data. The architecture proposed is discussed to have some at least superficial similarity to the mushroom bodies of insects.     

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.     

How do insects represent familiar terrain?

We argue here that ants and bees have a piecemeal representation of familiar terrain. These insects remember no more than what is needed to sustain the separate and parallel strategies that they employ when travelling between their nest and foraging sites. One major strategy is path integration. The insect keeps a running tally of its distance and direction from the nest and so can always return home. This global path integration is enhanced by long-term memories of significant sites that insects store in terms of the coordinates (direction and distance) of these sites relative to the nest. With these memories insects can plan routes that are steered by path integration to such sites. Quite distinct from global path integration are memories associated with familiar routes. Route memories include stored views of landmarks along the route with, in some cases, local vectors linked to them. Local vectors by encoding the direction and/or distance from one landmark to the next, or from one landmark to a goal, help an insect keep to a defined route. We review experiments showing that although local vectors can be recalled by recognising landmarks, the global path integration system is independent of landmark information and that landmarks do not have positional coordinates associated with them. The major function of route landmarks is thus procedural, telling an insect what action to perform next, rather than its location relative to the nest.     

Landmark learning and visuo-spatial memories in gerbils

The aim of this study is to understand what a rodent (Meriones unguiculatus) learns about the geometrical relations between a goal and nearby visual landmarks and how it uses this information to reach a goal. Gerbils were trained to find sunflower seeds on the floor of a light-tight, black painted room illuminated by a single light bulb hung from the ceiling. The position of the seed on the floor was specified by an array of one or more landmarks. Once training was complete, we recorded where the gerbils searched when landmarks were present but the seed was absent. In such tests, gerbils were confronted either with the array of landmarks to which they were accustomed or with a transformation of this array. Animals searched in the appropriate spot when trained to find seeds placed in a constant direction and at a constant distance from a single cylindrical landmark. Since gerbils look in one spot and not in a circle centred on the landmark, the direction between landmark and goal must be supplied by cues external to the landmark array. Distance, on the other hand, must be measured with respect to the landmark. Tests in which the size of the landmark was altered from that used in training suggest that distance is not learned solely in terms of the apparent size of the landmark as seen from the goal. Gerbils can still reach a goal defined by an array of landmarks when the room light is extinguished during their approach. This ability implies that they have already planned a trajectory to the goal before the room is darkened. In order to compute such a trajectory, their internal representation of landmarks and goal needs to contain information about the distances and bearings between landmarks and goal. For planning trajectories, each landmark of an array can be used separately from the others. Gerbils trained to a goal specified by an array of several landmarks were tested with one or more of the landmarks removed or with the array expanded. They then searched as though they had computed an independent trajectory for each landmark. For instance, gerbils trained with an array of two landmarks were tested with the distance between two landmarks doubled. The animals then searched for seeds in two positions, which were at the correct distance and in the right direction from each landmark.(ABSTRACT TRUNCATED AT 400 WORDS)     

Route and landmark learning by rats searching for food

Many foraging animals rely on visual landmarks and/or habitual paths to locate important resources. We examined the degree to which rats rely on these cues when they predicted conflicting food locations. In this foraging task, rats were required to find food which could be located using either a fixed route or a nearby visual landmark. In tests, we found that their subsequent search-based estimates of the food location were the same when animals had acquired a long-term memory of the route, the landmark, or both. We show that the degree to which animals rely on the cues depends not only on the discrepancy between the two cues, but also on whether animals can match the testing “view” with a learned “view” that has been acquired during training.     

Small-scale spatial cognition in pigeons

Roberts and Van Veldhuizen's [Roberts, W.A., Van Veldhuizen, N., 1985. Spatial memory in pigeons on the radial maze. J. Exp. Psychol.: Anim. Behav. Proc. 11, 241-260] study on pigeons in the radial maze sparked research on landmark use by pigeons in lab-based tasks as well as variants of the radial-maze task. Pigeons perform well on open-field versions of the radial maze, with feeders scattered on the laboratory floor. Pigeons can also be trained to search precisely for buried food. The search can be based on multiple landmarks, but is sometimes controlled by just one or two landmarks, with the preferred landmarks varying across individuals. Findings are similar in landmark-based searching on a computer monitor and on a lab floor, despite many differences between the two kinds of tasks. A number of general learning principles are found in landmark-based searching, such as cue competition, generalization and peak shift, and selective attention. Pigeons also learn the geometry of the environment in which they are searching. Neurophysiological studies have implicated the hippocampal formation (HF) in avian spatial cognition, with the right hippocampus hypothesized to play a more important role in the spatial recognition of goal locations. Most recently, single-cell recording from the pigeon's hippocampal formation has revealed cells with different properties from the classic 'place' cells of rats, as well as differences in the two sides of the hippocampus.     

View-based navigation in Hymenoptera: multiple strategies of landmark guidance in the approach to a feeder

In order to analyse how landmarks guide the last stages of an insect's approach to a goal, we recorded many flights of individual wasps and honeybees as they flew to an inconspicuous feeder on the ground that was marked by one or by two nearby landmarks. An individual tends to approach the feeder from a constant direction, flying close to the ground. Its body is oriented in roughly the same horizontal direction during the approach so that the feeder and landmarks are viewed over a narrow range of directions. Consequently, when the insect arrives at the feeder, the landmarks take up a standard position on the retina. Three navigational strategies govern the final approach. The insect first aims at a landmark, treating it as a beacon. Secondly, bees learn the appearance of a landmark with frontal retina and they associate with this stored view a motor trajectory which brings them from the landmark sufficiently close to the goal that it can be reached by image matching. Insects then move so as to put the landmark in its standard retinal position. Image matching is shown to be accomplished by a control system which has as set points the standard retinal position of the landmark and some parameter related to its retinal size. Accepted: 1 March 1997     

FISH landmarks for barley chromosomes (Hordeum vulgare L.).

Barley metaphase chromosomes (2n = 14) can be identified by fluorescence in situ hybridization (FISH) and digital imaging microscopy using heterologous 18S rDNA and 5S rDNA probe sequences. When these sequences are used together, FISH landmark signals were seen so that all 7 chromosomes were uniquely identified and unambiguously oriented. The chromosomal location of the landmark signals was determined by FISH to a barley trisomic series using the 18S and 5S probes labeled with different fluorophores. The utility of these FISH landmarks for barley physical mapping was also demonstrated when an Amy-2 cDNA clone and a BAC clone were hybridized with the FISH landmark probes.     

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.     

Memory and sun compensation by honey bees

Summary Experiments with two species of honey bees (Apis mellifera andA. cerana) have revealed that bees form a detailed memory of the spatial and temporal pattern of the sun's azimuthal movement, using local landmarks as a reference for the learning. These experiments were performed on overcast days, and consisted of removing a hive from one site in which bees had been trained to find food by flying along a prominent landmark, and displacing it to a similar site in which the landmark was aligned in a different compass direction. On overcast days, bees which flew along the landmark in the new site oriented their waggle dances in the hive as if they had actually flown in the training site. Thus, they confused the two sets of landmarks and set their dance angles according to a memory of the sun's position relative to the original landmarks. Furthermore, the dances changed in correspondence with the sun's azimuthal shift over several hours, even reflecting (approximately) the regular temporal variations in the rate of shift; such features of the sun's course must therefore be stored in memory. The primary mechanism underlying the learning of this pattern is probably similar to that proposed by New and New (1962): bees store in memory several time-linked solar azimuthal positions relative to features of the landscape, and refer to this stored array when they need to determine an unknown azimuth intermediate between two known positions.During the cloudy-day displacement experiments, celestial cues often appeared to bees in the new site, contradicting the stored information on which they had been basing their dances. Although most bees quickly adopted the dance angle reflecting their actual direction of flight relative to the sun, some later reverted to the original dance angle, indicating that the information on which it was based had remained in memory when the new information was being expressed; other bees performed bimodal dances which expressed both sets of information in alternate waggle runs. The separation in memory implied by these behaviors may reflect a neural strategy for updating a previously stored relationship between celestial and terrestrial references with new information presented by seasonal changes in the sun's course or by newly learned landmarks.     

Sensory Systems and Spatial Memory in the Fruit Bat Rousettus aegyptiacus

The megachiropteran fruit bat Rousettus aegyptiacus is able to orient and navigate using both vision and echolocation. These two sensory systems have different environmental constraints however, echolocation being relatively short range when compared with vision. Despite this difference, an experiment testing their memory of a perch location demonstrates that once the location of a perch is learned R. aegyptiacus is not influenced by the movement of local landmark cues in the vicinity of the perch under either light or dark conditions. Thus despite the differing constraints of vision and echolocation, this suggests a place is remembered as a location in space and not by associations with landmarks in the vicinity. A decrease in initial performance when the task was repeated in the dark suggested the possibility that a memory of a location learned using vision does not generalize to echolocation.     

Morphometric analysis of Cartesian coordinates of the human skull

A method for locating the three dimensional coordinates of cranial landmarks with respect to the Frankfort, midsagittal, and coronal planes is presented. Sliding calipers were used to obtain the distances from left and right porion and apex to each landmark, except for a few points where spreading calipers are required. In the present example, 35 landmarks (for a total of 105 measurements) were located for each of 35 Peruvian precolumbian skulls. These distances were entered into a program (SKULL) which calculates the Cartesian coordinates of each landmark. The XYZ coordinates of each landmark contain all the information necessary for calculation of the distances between any two landmarks, and these distances may also be obtained as output from program SKULL, if desired (595 distances if all 35 landmarks are used). Reliability of the location of coordinates was determined by comparing computed distances among selected landmarks from program SKULL with traditional anthropometric measurements. Satisfactory agreements were found. Direct multivariate analysis of the coordinates of the landmarks produced insights not available in traditional multivariate analysis of conventional anthropometric measurements.     

Memory of Location and Site Recognition in the Ant Formica uralensis (Hymenoptera: Formicidae)

Spatial recognition cues used in site fidelity in the ant Formica uralensis Ruzsky were studied using outdoor and laboratory arenas. Ant workers visiting symmetrically spaced feeders were colour-marked corresponding to the initial feeder visited during sampling. The effect of manipulating environmental cues on the mean 'spatial specialization' of the population was measured. Site recognition appears to be based on visual landmark/canopy cues. However, ants maintained some fidelity when shielded from these cues, suggesting the involvement of additional cues. When ridding our experimental device of olfactory deposits and shielding visual cues, site fidelity was lost. Idiothetic and/or geomagnetic cues are thought to provide spatial references to visual or olfactory landmarks. Altering nest position relative to the arena and changing the geomagnetic field within the arena in our study, however, did nothing to the site fidelity of visually deprived and non-deprived foragers.
We conclude that site fidelity is developed in a visually structured environment but supplemented by an olfactory backup system that is probably based on discrete home range markings rather than radial odour trails. We demonstrate furthermore that the visual component involved in site location can be stored in the memory of individual F. uralensis foragers during a 6-month hibernation period.     

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.     

A field study investigating effects of landmarks on territory size and shape

Few studies have examined how landmarks affect territories'' fundamental characteristics. In this field study, we investigated effects of landmarks on territory size, shape and location in a cichlid fish (Amatitlania siquia). We provided cans as breeding sites and used plastic plants as landmarks. During 10 min trials, we recorded locations where residents chased intruders and used those locations to outline and measure the territory. In two experiments, we observed pairs without landmarks and with either a point landmark (one plant) or linear landmark (four plants) placed near the nest can. We alternated which trial occurred first and performed the second trial 24 h after the first. Territories were approximately round without landmarks or with a point landmark but were significantly more elongated when we added a linear landmark. Without landmarks, nests were centrally located; however, with any landmark, pairs set territory boundaries closer to the landmark and thus the nest. Territory size was significantly reduced in the presence of any landmark. This reduction suggests that a smaller territory with well-defined boundaries has greater benefits than a larger territory with less well-defined borders.     

Die Speicheldrüsenchromosomen der StechmückeCulex pipiens

The salivary gland chromosomes of an autogenous strain ofCulex p. pipiens are described and mapped. The regions of the centromeres are the only large morphological landmarks and can be used for a quick and exact identification of the chromosomes. Asynaptic parts cannot be constantly observed. Balbianirings are not seen. The puffs being small and only visible for a short time, cannot be used as landmarks. Therefore the single arms of the chromosomes can only be recognized by their very clear and distinct pattern.     

The retrieval of visuo-spatial memories by honeybees

Summary In order to explore how honeybees manage to retrieve the right landmark-memory in the right place, we trained bees along a short foraging route which consisted of two identical huts 33 m apart. Bees entered each hut to collect a drop of sucrose on the floor. The location of the drop was defined by the same arrangement of four blue and yellow cylindrical landmarks. However, in one hut the drop was between two yellow cylinders and in two other it was to the east of the blue cylinders. On tests with the sucrose missing, bees tended to search in the appropriate area in each hut (Fig. 1), thus showing that they used cues other than the sight of the local landmarks to select the appropriate memory.In a second experiment, the position of the sucrose was specified by yellow cylinders in one hut and by blue triangles in the other. When the arrays were swapped between huts, bees searched in the position specified by the array they encountered (Fig. 2). Thus, memories can be triggered by visual features of local landmarks.Bees were also trained outside to collect food from two platforms 40 m apart. The location of sucrose on one platform was defined by yellow cylinders, and on the other it was defined by blue triangles. When these arrays were exchanged between platforms, bees searched on each platform as though the landmarks had not been swapped. It seems that the more distant surroundings, which fill most of the visual field, may be more potent than the local landmarks in deciding which memory should be retrieved.It is argued that one role of distant landmarks and other contextual cues is to ensure that bees retrieve the correct memory of a constellation of local landmarks while the bees are still some distance away from their goal. Even at a short distance, a bee's current image of local landmarks may differ considerably from its stored representation of those landmarks as seen from the goal. Accurate recall of the appropriate memory will be more certain if it is primed by relatively distant landmarks which present a more constant image as a bee moves in the vicinity of its goal.     

Ant navigation en route to the goal: signature routes facilitate way-finding of Gigantiops destructor

We investigated in laboratory conditions how foragers of the tropical ant Gigantiops destructor develop individually distinctive landmark routes. Way-finding along a familiar route involved the recognition of at least two locations, nest and feeding site, and the representation of spatial relations between these places. Familiar visual landmarks were important both at the beginning and at the end of the foraging journey. A motor routine guided the ants at the start of their foraging path towards the first landmarks, which they learnt to pass consistently on the same side, before taking the next direction. At the last stage of the route, landmark recognition allowed them to pinpoint their preferred feeding site without using distant cues or odometric information. By contrast, ants en route to the goal were not systematically guided by a stereotyped sequence of snapshots recalled at each corresponding stage of the route. Each ant slalomed in an idiosyncratic distinctive way around different midway landmarks from a foraging excursion to the next, which induced a variability of the path shapes in their intermediate parts. By reducing the number of landmark recognition-triggered responses, this economical visuomotor strategy may be helpful in the Amazonian forest where many prominent landmarks are alike.     

Neural systems for landmark-based wayfinding in humans

Humans and animals use landmarks during wayfinding to determine where they are in the world and to guide their way to their destination. To implement this strategy, known as landmark-based piloting, a navigator must be able to: (i) identify individual landmarks, (ii) use these landmarks to determine their current position and heading, (iii) access long-term knowledge about the spatial relationships between locations and (iv) use this knowledge to plan a route to their navigational goal. Here, we review neuroimaging, neuropsychological and neurophysiological data that link the first three of these abilities to specific neural systems in the human brain. This evidence suggests that the parahippocampal place area is critical for landmark recognition, the retrosplenial/medial parietal region is centrally involved in localization and orientation, and both medial temporal lobe and retrosplenial/medial parietal lobe regions support long-term spatial knowledge.