Self-correction mechanism for path integration in a modular navigation system on the basis of an egocentric spatial map.

Classical Computer Science approaches to navigation by autonomous robots continue to make good progress. However, we have only a limited understanding of how navigation is implemented in the neural networks of animals, which still perform very much better in navigational tasks than robots. In this paper we explore the implementation of neural network based navigation in a simple robot. We use a modular navigation system that contains separate representations of visual input and the path integration process. These representations are combined to influence the behavior of a robot. Both representations are encoded within recurrent neuronal networks. The outputs of the representations are vectors of polar values that encode the location of the nearest object, or of a specific place in the environment. The robot manoeuvres in relation to these attended locations, in the context of its egocentric spatial map. During ego-motion towards a goal, the network representation of the goal moves in a counter-movement due to applied motor feedback. The robot's position is continuously compared against its visual input, and mismatches between the visually perceived goal position and its spatial representation are corrected.     

Calibrating space: exploration is important for allothetic and idiothetic navigation

Allothetic and idiothetic navigation strategies rely on very different cues and computational procedures. Allothetic navigation uses the relationships between external cues (visual, auditory, and olfactory) and mapping or geometrical calculations to locate places. Idiothetic navigation relies on cues generated by self-movement (proprioceptive cues or cues from optic, auditory, and olfactory flow, or efference copy of motor commands) and path integration to locate a present location and/or a starting point. Whereas it is theorized that exploratory behavior is used by animals to create a central representation of allothetic cues, it is unclear whether exploration plays a role in idiothetic navigation. Computational models suggest that either a reference frame, calibrated by exploration, or vector addition, without reference to exploration, could support path integration. The present study evaluated the contribution of exploration in these navigation strategies by comparing its contribution to the solution of both allothetic and idiothetic navigation problems. In two experiments, rats were trained to forage on an open table for large food pellets, which they then carried to a refuge to eat. Once trained, they were given probe trials from novel locations in either normal light, which permits the use of allothetic cues, or in infrared light, which requires the use of idiothetic cues. When faced with a new problem in either lighting condition, the rats first explored the foraging table before navigating directly home with the food. That exploration is equally important for allothetic and idiothetic navigation, suggests that both navigation strategies require a calibrated representation of the environment.     

Intact landmark control and angular path integration by head direction cells in the anterodorsal thalamus after lesions of the medial entorhinal cortex

The medial entorhinal cortex (MEC) occupies a central position within neural circuits devoted to the representation of spatial location and orientation. The MEC contains cells that fire as a function of the animal's head direction (HD), as well as grid cells that fire in multiple locations in an environment, forming a repeating hexagonal pattern. The MEC receives inputs from widespread areas of the cortical mantle including the ventral visual stream, which processes object recognition information, as well as information about visual landmarks. The role of the MEC in processing the HD signal or landmark information is unclear. We addressed this issue by neurotoxically damaging the MEC and recording HD cells within the anterodorsal thalamus (ADN). Direction-specific activity was present in the ADN of all animals with MEC lesions. Moreover, the discharge characteristics of ADN HD cells were only mildly affected by MEC lesions, with HD cells exhibiting greater anticipation of future HDs. Tests of landmark control revealed that HD cells in lesioned rats were capable of accurately updating their preferred firing directions in relation to a salient visual cue. Furthermore, cells from lesioned animals maintained stable preferred firing directions when locomoting in darkness and demonstrated stable HD cell tuning when locomoting into a novel enclosure, suggesting that MEC lesions did not disrupt the integration of idiothetic cues, or angular path integration, by HD cells. Collectively, these findings suggest that the MEC plays a limited role in the formation and spatial updating of the HD cell signal.     

Geometric cues influence head direction cells only weakly in nondisoriented rats

The influential hypothesis that environmental geometry is critical for spatial orientation has been extensively tested behaviorally, and yet findings have been conflicting. Head direction (HD) cells, the neural correlate of the sense of direction, offer a window into the processes underlying directional orientation and may help clarify the issue. In the present study, HD cells were recorded as rats foraged in enclosures of varying geometry, with or without simultaneous manipulation of landmarks and self-motion cues (path integration). All geometric enclosures had single-order rotational symmetry and thus completely polarized the environment. They also had unique features, such as corners, which could, in principle, act as landmarks. Despite these strongly polarizing geometric cues, HD cells in nondisoriented rats never rotated with these shapes. In contrast, when a cue card (white or gray) was added to one wall, HD cells readily rotated with the enclosure. When path integration was disrupted by disorienting the rat, HD cells rotated with the enclosure even without the landmark. Collectively, these findings indicate that geometry exerts little or no influence on heading computations in nondisoriented rats, but it can do so in disoriented rats. We suggest that geometric processing is only a weak influence, providing a backup system for heading calculations and recruited only under conditions of disorientation.     

Competitive Hebbian learning and the hippocampal place cell system: modeling the interaction of visual and path integration cues.

The hippocampus has long been thought essential for implementing a cognitive map of the environment. However, almost 30 years since place cells were found in rodent hippocampal field CA1, it is still unclear how such an allocentric representation arises from an ego-centrically perceived world. By means of a competitive Hebbian learning rule responsible for coding visual and path integration cues, our model is able to explain the diversity of place cell responses observed in a large set of electrophysiological experiments with a single fixed set of parameters. Experiments included changes observed in place fields due to exploration of a new environment, darkness, retrosplenial cortex inactivation, and removal, rotation, and permutation of landmarks. To code for visual cues for each landmark, we defined two perceptual schemas representing landmark bearing and distance information over a linear array of cells. The information conveyed by the perceptual schemas is further processed through a network of adaptive layers which ultimately modulate the resulting activity of our simulated place cells. In path integration terms, our system is able to dynamically remap a bump of activity coding for the displacement of the animal in relation to an environmental anchor. We hypothesize that path integration information is computed in the rodent posterior parietal cortex and conveyed to the hippocampus where, together with visual information, it modulates place cell activity. The resulting network yields a more direct treatment of partial remapping of place fields than other models. In so doing, it makes new predictions regarding the nature of the interaction between visual and path integration cues during new learning and when the system is challenged with environmental changes.     

Spatial discrimination of visually similar environments by hippocampal place cells in the presence of remote recalibrating landmarks

Place cells in the rat hippocampus commonly show place-related firing activity in the animal's current environment. Here, we evaluated the capability of the place cell system to discriminate visually identical environments. Place cell activity was first recorded while rats moved freely in a cylinder divided into three connected sectors. Two sectors were visually identical whereas the third sector was made distinctive by the addition of visual and tactile cues. When in a given sector, the rats could not perceive the cues present in the other two sectors. Most cells had distinctive place fields in each sector, including the two identical sectors. To rule out the influence of non-controlled cues, rotations of the cylinder (+/- 120 degrees) were conducted. When successful, cylinder rotations resulted in equivalent field rotation for all cells. These results suggest that the place cell system is able to form a specific spatial representation for all sectors, so that the rat knows, at any time, in which sector it is currently located. Presumably, such discrimination relies on angular path integration in which the computational errors stemming from self-motion cues would be corrected by environmental landmarks provided by the distinctive sector.     

Involvement of the hippocampus and associative parietal cortex in the use of proximal and distal landmarks for navigation

Rats with dorsal hippocampus or associative parietal cortex (APC) lesions and sham-operated controls were trained on variants of the Morris water maze navigation task. In the 'proximal landmark condition', the rats had to localize the hidden platform solely on the basis of three salient object landmarks placed directly in the swimming pool. In the 'distal landmark condition', rats could rely only on distal landmarks (room cues) to locate the platform. In the 'beacon condition', the platform location was signaled by a salient cue directly attached to it. Rats with hippocampal lesions were impaired in the distal and to a less extent in the proximal landmark condition whereas rats with parietal lesions were impaired only in the proximal landmark condition. None of the lesioned groups was impaired in the beacon condition. These results suggest that the processing of information related to proximal, distal landmarks or associated beacon are mediated by different neural systems. The hippocampus would contribute to both proximal and distal landmark processing whereas the APC would be involved in the processing of proximal landmarks only. Navigation relying on a cued-platform would not require participation of the hippocampus nor the APC. Assuming that the processing of proximal landmarks heavily depends on the integration of visuospatial and idiothetic information, these results are consistent with the hypothesis that the APC plays a role in the combination of multiple sensory information and contributes to the formation of an allocentric spatial representation.     

Encoding and retrieval of landmark‐related spatial cues during navigation: An fMRI study

To successfully navigate, humans can use different cues from their surroundings. Learning locations in an environment can be supported by parallel subsystems in the hippocampus and the striatum. We used fMRI to look at differences in the use of object‐related spatial cues while 47 participants actively navigated in an open‐field virtual environment. In each trial, participants navigated toward a target object. During encoding, three positional cues (columns) with directional cues (shadows) were available. During retrieval, the removed target had to be replaced while either two objects without shadows (objects trial) or one object with a shadow (shadow trial) were available. Participants were informed in blocks about which type of retrieval trial was most likely to occur, thereby modulating expectations of having to rely on a single landmark or on a configuration of landmarks. How the spatial learning systems in the hippocampus and caudate nucleus were involved in these landmark‐based encoding and retrieval processes were investigated. Landmark configurations can create a geometry similar to boundaries in an environment. It was found that the hippocampus was involved in encoding when relying on configurations of landmarks, whereas the caudate nucleus was involved in encoding when relying on single landmarks. This might suggest that the observed hippocampal activation for configurations of objects is linked to a spatial representation observed with environmental boundaries. Retrieval based on configurations of landmarks activated regions associated with the spatial updation of object locations for reorientation. When only a single landmark was available during retrieval, regions associated with updating the location of oneself were activated. There was also evidence that good between‐participant performance was predicted by right hippocampal activation. This study therefore sheds light on how the brain deals with changing demands on spatial processing related purely to landmarks. © 2014 Wiley Periodicals, Inc.     

Objects and landmarks: Hippocampal place cells respond differently to manipulations of visual cues depending on size,perspective, and experience

Human navigation studies show that landmarks are used for orientation, whereas objects contribute to the contextual representation of an environment. What constitutes a landmark? Classic rodent studies show that hippocampal place fields are controlled by distal, polarizing cues. Place fields, however, are also influenced by local cues. One difficulty in examining mechanisms by which distal and local cues influence the activity of hippocampal cells is that distant cues are necessarily processed visually, whereas local cues are generally multimodal. Here, we compared the effects of 90° rotations under different cue conditions in which cues were restricted to the visual modality. Three two‐dimensional visual cue conditions were presented in a square open field: a large vertical cue on one wall, a large floor cue in a corner abutting two walls, and a smaller complex floor cue in a corner adjacent to two walls. We show that rotations of large distal cues, whether on the wall or floor, were equally effective in controlling place fields. Rotations of the smaller floor cues were significantly more likely to result in remapping, whether or not animals were also exposed to the distal polarizing cues. Responses of distal and local cues were affected differently by extended experience. Our data provide evidence that hippocampal place cell responses to visual cues are influenced by perspective, salience of the cue, and prior experience. The hippocampus processes visual cues either as stable landmarks useful for orientation and navigation or as nonstationary objects or features of the local environment available for associative learning or binding items in context. © 2014 Wiley Periodicals, Inc.     

Spatial navigation and hippocampal place cell firing: the problem of goal encoding

Place cells are hippocampal neurons whose discharge is strongly related to a rat's location in the environment. The existence of such cells, combined with the reliable impairments seen in spatial tasks after hippocampal damage, has led to the proposal that place cells form part of an integrated neural system dedicated to spatial navigation. This hypothesis is supported by the strong relationships between place cell activity and spatial problem solving, which indicate that the place cell representation must be both functional and in register with the surroundings for the animal to perform correctly in spatial tasks. The place cell system nevertheless requires other essential elements to be competent, such as a component that specifies the overall goal of the animal and computes the path required to take the rat from its current location to the goal. Here, we propose a model of the neural network responsible for spatial navigation that includes goal coding and path selection. In this model, the hippocampal formation allows for place recognition, and stores the set of places that can be accessed from each position in the environment. The prefrontal cortex is responsible for encoding goal location and for route planning. The nucleus accumbens translates paths in neural space into appropriate locomotor activity that moves the animal towards the goal in real space. The complete model assumes that the hippocampal output to nucleus accumbens and prefrontal cortex provides information for generating solutions to spatial problems. In support of this model, we finally present preliminary evidence that the goal representation necessary for path planning might be encoded in the prelimbic/infralimbic region of the medial prefrontal cortex.     

Hippocampectomized rats are impaired in homing by path integration

Theoretical, behavioral, and electrophysiologic evidence suggests that the hippocampal formation may play a role in path integration, a form of spatial navigation in which an animal can return to a starting point by integrating self-movement cues generated on its outward journey. The present study examined whether the hippocampus (Ammon's horn and the dentate gyrus) is involved in this form of spatial behavior. Control rats and rats with selective ibotenic acid lesions of the hippocampus were tested in a foraging task in which they retrieved large food pellets from an open field, which when found, they carried to a refuge for consumption. The experiments measured the rats' homing accuracy, returning to the starting location, under conditions in which visual, surface, and self-movement cues; surface and self-movement cues; or only self-movement cues were available. Although both control rats and rats without a hippocampus could use visual and surface cues, only control rats appeared to be able to use self-movement cues. The finding that hippocampal rats are impaired under conditions requiring the use of self-movement cues suggests that the hippocampus plays an essential role in path integration.     

Direction and distance deficits in path integration after unilateral vestibular loss depend on task complexity

The effects of peripheral vestibular disorders on the direction and distance components of the internal spatial representation were investigated. The ability of Menière's patients to perform path integration was assessed in different situations aimed at differentiating the level of spatial processing (simple versus complex tasks), the available sensory cues (proprioceptive, vestibular, or visual conditions), and the side of the path (towards the healthy versus the lesioned side). After exploring two legs of a triangle, participants were required either to reproduce the exploration path, to follow the reverse path, or to take a shortcut to the starting point of the path (triangle completion). Patients' performances were recorded before unilateral vestibular neurotomy (UVN) and during the time-course of recovery (1 week and 1 month) and were compared to those of matched control subjects tested at similar time intervals. Both the angular and linear path components of the trajectory were impaired for patients compared to controls. However, deficits were restricted to the complex tasks, which required a higher level of spatial processing. Most deficits were maximal 1 week after UVN, and some remained up to the first post-operative month. Spatial representation was differentially impaired according to the available sensory cues: deficits were absent in active locomotor blindfolded condition, appeared in conditions involving visual and vestibular information, and were maximal when visual cues alone were available. Finally, concerning the side of the path, unilateral vestibular loss led to global impairment of the internal spatial representation, yet some asymmetrical spatial performances were observed 1 week after UVN. On the whole, results suggest that the environment experienced by the patients is different after UVN and that a different internal spatial representation is constructed, especially for tasks requiring high levels of spatial processing.     

Distal landmarks and hippocampal place cells: effects of relative translation versus rotation

Hippocampal neurons are selectively active when a rat occupies restricted locations in an environment. These place cells derive their specificity from a multitude of sources, including idiothetic cues and sensory input derived from both distal and local landmarks. Most experiments have attempted to dissociate the relative strengths and roles played by these sources by rotating one set against the other. Few studies have addressed the effects of relative translation of the local cue set versus salient distal landmarks. To address this question, ensembles of place cells were recorded as a rectangular or circular track was moved to different locations in a room with controlled visual landmarks. Place cells primarily maintained their firing fields relative to the track (i.e., occupying new locations relative to the distal landmarks), even though the track could occupy completely nonoverlapping regions of the room. When the distal landmarks were rotated around the circular track, however, the place fields rotated with the landmarks, demonstrating that the cues were perceptible to the rat. These results suggest that, under these conditions, the spatial tuning of place cells may derive from an interaction between local and idiothetic cues, which define the precise firing locations of the cells and the relationships between them, and distal landmarks, which set the orientation of the ensemble representation relative to the external environment.     

Geometric information is required for allothetic navigation in mice

In tasks for allothetic navigation, animals should orientate by means of distal cues. We have previously shown that mice use several forms of information to navigate, among which geometry, i.e. the shape of the environment, seems to play an important role. Here we investigated whether geometric features of the environment are necessary for allothetic navigation in mice. Mice were trained to navigate in a circular water maze by means of four distal landmarks distributed either symmetrically (symmetry group) or asymmetrically (asymmetry group) around the maze. Thus, mice could locate a hidden platform by either differentiating the landmarks based on their intrinsic features (symmetry group) or in addition by geometric information, i.e. based on the relative distances between landmarks (asymmetry group). Data indicated that place learning occurred only in the asymmetry group. The results support the idea that mice navigate by using the relational properties between distal landmarks and that geometric information is required for proper allothetic navigation in this species.     

Geometry and landmark representation by pigeons: evidence for species‐differences in the hemispheric organization of spatial information processing?

In this study, we investigated how pigeons ( Columba livia ) represent environmental geometry and landmark information. Birds learned to locate the centre of a square arena by means of geometric cues alone, or by means of both geometric and landmark cues. By manipulating the type of information available at training and testing, we assessed which cues the birds had encoded, and through the use of monocular occlusion we examined how the information was represented by the two brain hemispheres. Our results show that both brain hemispheres encoded geometric and landmark information. During all viewing conditions, the geometric representation was based mainly on an absolute metric for distance. The relative use of geometry and landmarks was experience dependent. With both brain hemispheres available birds relied, to a greater degree, on geometric information and used it in a more integrated way than with either hemisphere alone. Overall, our findings show a different pattern for the hemispheric encoding of geometric and landmark information by the pigeon than that previously reported for the domestic chick. Our results suggest that the organization of spatial information processing in the left and right brain hemispheres of birds may be more diverse than what is currently known.     

Extinction of drug- and withdrawal-paired cues in animal models: Relevance to the treatment of addiction

Conditioned drug craving and withdrawal elicited by cues paired with drug use or acute withdrawal are among the many factors contributing to compulsive drug taking. Understanding how to stop these cues from having these effects is a major goal of addiction research. Extinction is a form of learning in which associations between cues and the events they predict are weakened by exposure to the cues in the absence of those events. Evidence from animal models suggests that conditioned responses to drug cues can be extinguished, although the degree to which this occurs in humans is controversial. Investigations into the neurobiological substrates of extinction of conditioned drug craving and withdrawal may facilitate the successful use of drug cue extinction within clinical contexts. While this work is still in the early stages, there are indications that extinction of drug- and withdrawal-paired cues shares neural mechanisms with extinction of conditioned fear. Using the fear extinction literature as a template, it is possible to organize the observations on drug cue extinction into a cohesive framework.     

The dynamics of memory consolidation of landmarks

Navigating through space is fundamental to human nature and requires the ability to retrieve relevant information from the remote past. With the passage of time, some memories become generic, capturing only a sense of familiarity. Yet, others maintain precision, even when acquired decades ago. Understanding the dynamics of memory consolidation is a major challenge to neuroscientists. Using functional magnetic resonance imaging, we systematically examined the effects of time and spatial context on the neural representation of landmark recognition memory. An equal number of male and female subjects (males N = 10, total N = 20) watched a route through a large‐scale virtual environment. Landmarks occurred at navigationally relevant and irrelevant locations along the route. Recognition memory for landmarks was tested directly following encoding, 24 h later and 30 days later. Surprisingly, changes over time in the neural representation of navigationally relevant landmarks differed between males and females. In males, relevant landmarks selectively engaged the parahippocampal gyrus (PHG) regardless of the age of the memory. In females, the response to relevant landmarks gradually diminished with time in the PHG but strengthened progressively in the inferior frontal gyrus (IFG). Based on what is known about the functioning of the PHG and IFG, the findings of this study suggest that males maintain access to the initially formed spatial representation of landmarks whereas females become strongly dependent on a verbal representation of landmarks with time. Our findings yield a clear objective for future studies. © 2017 Wiley Periodicals, Inc.     

Extinction of drug- and withdrawal-paired cues in animal models: Relevance to the treatment of addiction

Conditioned drug craving and withdrawal elicited by cues paired with drug use or acute withdrawal are among the many factors contributing to compulsive drug taking. Understanding how to stop these cues from having these effects is a major goal of addiction research. Extinction is a form of learning in which associations between cues and the events they predict are weakened by exposure to the cues in the absence of those events. Evidence from animal models suggests that conditioned responses to drug cues can be extinguished, although the degree to which this occurs in humans is controversial. Investigations into the neurobiological substrates of extinction of conditioned drug craving and withdrawal may facilitate the successful use of drug cue extinction within clinical contexts. While this work is still in the early stages, there are indications that extinction of drug- and withdrawal-paired cues shares neural mechanisms with extinction of conditioned fear. Using the fear extinction literature as a template, it is possible to organize the observations on drug cue extinction into a cohesive framework.     

Hippocampal formation is required for geometric navigation in pigeons

The geometric properties of bounded space have attracted considerable attention as a source of spatial information that can guide goal navigation. Although the use of geometric information to navigate has been observed in every species studied to date, the neural mechanisms that support the representation of geometric information are still debated. With the purpose of investigating this topic, we trained pigeons with lesion to the hippocampal formation to search for food in a rectangular-shaped arena containing one wall of a different color that served as the only distinctive environmental feature. Although lesioned pigeons learned the task even faster than control animals, probe trials showed that they were insensitive to geometric information. Control animals could encode and use both geometric and feature information to locate the goal. By contrast, lesioned pigeons relied exclusively on the feature information provided by the wall of a different color. The results indicate that the avian hippocampal formation is critical for learning the geometric properties of space in homing pigeons.     

Which way and how far? Tracking of translation and rotation information for human path integration

Path integration, the constant updating of the navigator's knowledge of position and orientation during movement, requires both visuospatial knowledge and memory. This study aimed to develop a systems‐level understanding of human path integration by examining the basic building blocks of path integration in humans. To achieve this goal, we used functional imaging to examine the neural mechanisms that support the tracking and memory of translational and rotational components of human path integration. Critically, and in contrast to previous studies, we examined movement in translation and rotation tasks with no defined end‐point or goal. Navigators accumulated translational and rotational information during virtual self‐motion. Activity in hippocampus, retrosplenial cortex (RSC), and parahippocampal cortex (PHC) increased during both translation and rotation encoding, suggesting that these regions track self‐motion information during path integration. These results address current questions regarding distance coding in the human brain. By implementing a modified delayed match to sample paradigm, we also examined the encoding and maintenance of path integration signals in working memory. Hippocampus, PHC, and RSC were recruited during successful encoding and maintenance of path integration information, with RSC selective for tasks that required processing heading rotation changes. These data indicate distinct working memory mechanisms for translation and rotation, which are essential for updating neural representations of current location. The results provide evidence that hippocampus, PHC, and RSC flexibly track task‐relevant translation and rotation signals for path integration and could form the hub of a more distributed network supporting spatial navigation. Hum Brain Mapp 37:3636–3655, 2016. © 2016 Wiley Periodicals, Inc .