Navigation Representation Research Articles

Spatial cognition: Uncovering navigational representations in prefrontal cortices

A new study identifies representations of navigational variables in six prefrontal regions in freely moving macaques, expanding our view of how the brain represents space outside of the broader hippocampal formation.

Using a picture (or a thousand words) for supporting spatial knowledge of a complex virtual environment

External representations powerfully support and augment complex human behavior. When navigating, people often consult external representations to help them find the way to go, but do maps or verbal instructions improve spatial knowledge or support effective wayfinding? Here, we examine spatial knowledge with and without external representations in two studies where participants learn a complex virtual environment. In the first study, we asked participants to generate their own maps or verbal instructions, partway through learning. We found no evidence of improved spatial knowledge in a pointing task requiring participants to infer the direction between two targets, either on the same route or on different routes, and no differences between groups in accurately recreating a map of the target landmarks. However, as a methodological note, pointing was correlated with the accuracy of the maps that participants drew. In the second study, participants had access to an accurate map or set of verbal instructions that they could study while learning the layout of target landmarks. Again, we found no evidence of differentially improved spatial knowledge in the pointing task, although we did find that the map group could recreate a map of the target landmarks more accurately. However, overall improvement was high. There was evidence that the nature of improvement across all conditions was specific to initial navigation ability levels. Our findings add to a mixed literature on the role of external representations for navigation and suggest that more substantial intervention—more scaffolding, explicit training, enhanced visualization, perhaps with personalized sequencing—may be necessary to improve navigation ability.

A review of techniques to extract road network features from global positioning system data for transport modelling

ABSTRACT With the spread of smartphones and mobile internet, Global Positioning System (GPS) data from vehicles has become widely available. This data represents a unique opportunity to automatically extract road network features and generate detailed maps that can be used in the creation of transport network models, while minimising the quantity of resources usually invested in that task. Accurate transport network models can be used in a variety of applications either in transport simulation models or autonomous vehicles navigation. Although two relevant literature reviews were performed during the last decade, they were not systematic and did not explore the road network inference methods from a transport network modelling point of view. The objective of this research is to perform a systematic and reproducible literature review on the use GPS data in transport network modelling and provide limitations and future work to extract a road network representation for transport models and autonomous vehicles navigation. This was done by systematically examining the studies’ different approaches with respect to relevant criteria. Most studies produced a simple representation of the road network, not detailed enough for transport models. Other limitations were the bias introduced by the GPS sample and the reproducibility of the different methods.

Hippocampal spatial view cells, place cells, and concept cells: View representations.

A commentary is provided on issues raised in the Special Issue of Hippocampus (2023) on hippocampal system view representations. First, the evidence for hippocampal and parahippocampal spatial view cells in primates including humans shows that the allocentric representations provided by at least some of these cells are very useful for human memory in that where objects and rewards are seen in the world "out there" is a key component of episodic memory and navigation. Spatial view cell representations provide for memory and navigation to be independent of the place where the individual is currently located and of the egocentric coordinates of the viewed location and the facing direction of the individual. Second, memory and navigation in humans are normally related to the visual cues encoded by spatial view cells that define a location "out there" such as a building, hill, and so forth, not to an unmarked place without local cues and identified only by distant environmental/room cues. Third, "mixed" representations, for example of particular combinations of spatial view and place, can arise if the training has been for only some combinations of place and view, for that is what can then be learned by the hippocampus. Fourth, rodents, with their much less good visual acuity (~1 cycle/° in rats, compared with ~60 cycles/° for the human fovea), and rodents' very wide viewing angle for the world (~270°) might be expected, when using the same computational mechanisms as in primates, to use widely spaced environmental cues to define a place where the rodent is located, supported by inputs about place using local olfactory and tactile cues. Fifth, it is shown how view-point dependent allocentric representations could form a view-point independent allocentric representation for memory and navigation. Sixth, concept cells in humans and primates with connectivity to the hippocampus are compared.

Activation of human visual area V6 during egocentric navigation with and without visual experience

V6 is a retinotopic area located in the dorsal visual stream that integrates eye movements with retinal and visuo-motor signals. Despite the known role of V6 in visual motion, it is unknown whether it is involved in navigation and how sensory experiences shape its functional properties. We explored the involvement of V6 in egocentric navigation in sighted and in congenitally blind (CB) participants navigating via an in-house distance-to-sound sensory substitution device (SSD), the EyeCane. We performed two fMRI experiments on two independent datasets. In the first experiment, CB and sighted participants navigated the same mazes. The sighted performed the mazes via vision, while the CB performed them via audition. The CB performed the mazes before and after a training session, using the EyeCane SSD. In the second experiment, a group of sighted participants performed a motor topography task. Our results show that right V6 (rhV6) is selectively involved in egocentric navigation independently of the sensory modality used. Indeed, after training, rhV6 of CB is selectively recruited for auditory navigation, similarly to rhV6 in the sighted. Moreover, we found activation for body movement in area V6, which can putatively contribute to its involvement in egocentric navigation. Taken together, our findings suggest that area rhV6 is a unique hub that transforms spatially relevant sensory information into an egocentric representation for navigation. While vision is clearly the dominant modality, rhV6 is in fact a supramodal area that can develop its selectivity for navigation in the absence of visual experience.

Rapid learning of spatial representations for goal-directed navigation based on a novel model of hippocampal place fields

The discovery of place cells and other spatially modulated neurons in the hippocampal complex of rodents has been crucial to elucidating the neural basis of spatial cognition. More recently, the replay of neural sequences encoding previously experienced trajectories has been observed during consummatory behavior—potentially with implications for rapid learning, quick memory consolidation, and behavioral planning. Several promising models for robotic navigation and reinforcement learning have been proposed based on these and previous findings. Most of these models, however, use carefully engineered neural networks, and sometimes require long learning periods. In this paper, we present a self-organizing model incorporating place cells and replay, and demonstrate its utility for rapid one-shot learning in non-trivial environments with obstacles.

Understanding 3D vision as a policy network.

It is often assumed that the brain builds 3D coordinate frames, in retinal coordinates (with binocular disparity giving the third dimension), head-centred, body-centred and world-centred coordinates. This paper questions that assumption and begins to sketch an alternative based on, essentially, a set of reflexes. A 'policy network' is a term used in reinforcement learning to describe the set of actions that are generated by an agent depending on its current state. This is an untypical starting point for describing 3D vision, but a policy network can serve as a useful representation both for the 3D layout of a scene and the location of the observer within it. It avoids 3D reconstruction of the type used in computer vision but is similar to recent representations for navigation generated through reinforcement learning. A policy network for saccades (pure rotations of the camera/eye) is a logical starting point for understanding (i) an ego-centric representation of space (e.g. Marr's (Marr 1982 Vision: a computational investigation into the human representation and processing of visual information) 2[Formula: see text]-D sketch) and (ii) a hierarchical, compositional representation for navigation. The potential neural implementation of policy networks is straightforward; a network with a large range of sensory and task-related inputs such as the cerebellum would be capable of implementing this input/output function. This is not the case for 3D coordinate transformations in the brain: no neurally implementable proposals have yet been put forward that could carry out a transformation of a visual scene from retinal to world-based coordinates. Hence, if the representation underlying 3D vision can be described as a policy network (in which the actions are either saccades or head translations), this would be a significant step towards a neurally plausible model of 3D vision. This article is part of the theme issue 'New approaches to 3D vision'.

OpenStreetMap-Based Autonomous Navigation With LiDAR Naive-Valley-Path Obstacle Avoidance

OpenStreetMaps (OSM) is currently studied as the environment representation for autonomous navigation. It provides advantages such as global consistency, a heavy-less map construction process, and a wide variety of road information publicly available. However, the location of this information is usually not very accurate locally. In this paper, we present a complete autonomous navigation pipeline using OSM information as environment representation for global planning. To avoid the flaw of local low-accuracy, we offer the novel LiDAR-based Naive-Valley-Path (NVP) method that exploits the concept of "valley" areas to infer the local path always furthest from obstacles. This behavior allows navigation always through the center of trafficable areas following the road's shape independently of OSM error. Furthermore, NVP is a naive method that is highly sample-time-efficient. This time efficiency also enables obstacle avoidance, even for dynamic objects. We demonstrate the system's robustness in our research platform BLUE, driving autonomously across the University of Alicante Scientific Park for more than 20 km with 0.24 meters of average error against the road's center with a 19.8 ms of average sample time. Our vehicle avoids static obstacles in the road and even dynamic ones, such as vehicles and pedestrians.

3D building interior modelling for navigation in emergency response applications

• Broad perspective review on building interior modelling. • Three-dimensional model representations for navigation in emergency response. • Application requirements on a spatial model: space subdivision, topology representation, semantics and graphic representation. The aim of this paper is to present a broad perspective on applications dealing with 3D building models and indoor navigation during an emergency. This is an immense topic, which has been considerably investigated by researchers from different fields of expertise: construction, cognition, spatial information science, cartography, robotics, mathematics, engineering, etc. It is not possible to provide one, complete approach to the problem, because it largely depends on the application. However, some solutions are prevailing, and they are adopted by others. In this paper there is an attempt made to signalise specific directions and trends related to the topic. They include building model requirements, navigation network generation and representation, and 3D visualisation of models and navigation paths. These issues are important in the process of rapid model generation, which is suitable for decision support and situational awareness systems for first responders. The authors highlight issues that may be investigated by young researchers or may be considered as alternatives by more experienced ones. The paper is therefore the aim of this paper is to summarise the contemporary research conducted in various research teams, as well as to present author’s own observations and experiences.

Contrastive Instruction-Trajectory Learning for Vision-Language Navigation

The vision-language navigation (VLN) task requires an agent to reach a target with the guidance of natural language instruction. Previous works learn to navigate step-by-step following an instruction. However, these works may fail to discriminate the similarities and discrepancies across instruction-trajectory pairs and ignore the temporal continuity of sub-instructions. These problems hinder agents from learning distinctive vision-and-language representations, harming the robustness and generalizability of the navigation policy. In this paper, we propose a Contrastive Instruction-Trajectory Learning (CITL) framework that explores invariance across similar data samples and variance across different ones to learn distinctive representations for robust navigation. Specifically, we propose: (1) a coarse-grained contrastive learning objective to enhance vision-and-language representations by contrasting semantics of full trajectory observations and instructions, respectively; (2) a fine-grained contrastive learning objective to perceive instructions by leveraging the temporal information of the sub-instructions; (3) a pairwise sample-reweighting mechanism for contrastive learning to mine hard samples and hence mitigate the influence of data sampling bias in contrastive learning. Our CITL can be easily integrated with VLN backbones to form a new learning paradigm and achieve better generalizability in unseen environments. Extensive experiments show that the model with CITL surpasses the previous state-of-the-art methods on R2R, R4R, and RxR.

A theory of geometry representations for spatial navigation.

The geometric information of space, such as environment boundaries, is represented heterogeneously across brain regions. The computational mechanisms of encoding the spatial layout of environments remain to be determined. Here, we postulate a conjunctive encoding theory to illustrate the construct of cognitive maps from geometric perception. The theory naturally describes a spectrum of cell types including experimentally observed boundary vector cells, border cells, "annulus" and "bulls-eye" cells as special examples. In a similar way, inspired by the integration of egocentric and allocentric information as found in the postrhinal cortex, the theory also predicts a new cell type, named geometry cell. Geometry cells encode the geometric layout of the local space relative to the environment center, independent of the animal's positions and headings within the local space. The predicted geometry cell provides pure allocentric high-level representations of local scenes to support the quick formation of cognitive map representations capturing the spatial layout of complex environments. The theory sheds new light on the neural mechanisms of spatial cognition and brain-inspired autonomous intelligent systems.

Cross-Modal Contrastive Learning of Representations for Navigation Using Lightweight, Low-Cost Millimeter Wave Radar for Adverse Environmental Conditions

Deep reinforcement learning (RL), where the agent learns from mistakes, has been successfully applied to a variety of tasks. With the aim of learning collision-free policies for unmanned vehicles, deep RL has been used for training with various types of data, such as colored images, depth images, and LiDAR point clouds, without the use of classic map--localize--plan approaches. However, existing methods are limited by their reliance on cameras and LiDAR devices, which have degraded sensing under adverse environmental conditions (e.g., smoky environments). In response, we propose the use of single-chip millimeter-wave (mmWave) radar, which is lightweight and inexpensive, for learning-based autonomous navigation. However, because mmWave radar signals are often noisy and sparse, we propose a cross-modal contrastive learning for representation (CM-CLR) method that maximizes the agreement between mmWave radar data and LiDAR data in the training stage. We evaluated our method in real-world robot compared with 1) a method with two separate networks using cross-modal generative reconstruction and an RL policy and 2) a baseline RL policy without cross-modal representation. Our proposed end-to-end deep RL policy with contrastive learning successfully navigated the robot through smoke-filled maze environments and achieved better performance compared with generative reconstruction methods, in which noisy artifact walls or obstacles were produced. All pretrained models and hardware settings are open access for reproducing this study and can be obtained at https://arg-nctu.github.io/projects/deeprl-mmWave.html

Reinforcement learning approaches to hippocampus-dependent flexible spatial navigation.

Humans and non-human animals show great flexibility in spatial navigation, including the ability to return to specific locations based on as few as one single experience. To study spatial navigation in the laboratory, watermaze tasks, in which rats have to find a hidden platform in a pool of cloudy water surrounded by spatial cues, have long been used. Analogous tasks have been developed for human participants using virtual environments. Spatial learning in the watermaze is facilitated by the hippocampus. In particular, rapid, one-trial, allocentric place learning, as measured in the delayed-matching-to-place variant of the watermaze task, which requires rodents to learn repeatedly new locations in a familiar environment, is hippocampal dependent. In this article, we review some computational principles, embedded within a reinforcement learning framework, that utilise hippocampal spatial representations for navigation in watermaze tasks. We consider which key elements underlie their efficacy, and discuss their limitations in accounting for hippocampus-dependent navigation, both in terms of behavioural performance (i.e. how well do they reproduce behavioural measures of rapid place learning) and neurobiological realism (i.e. how well do they map to neurobiological substrates involved in rapid place learning). We discuss how an actor–critic architecture, enabling simultaneous assessment of the value of the current location and of the optimal direction to follow, can reproduce one-trial place learning performance as shown on watermaze and virtual delayed-matching-to-place tasks by rats and humans, respectively, if complemented with map-like place representations. The contribution of actor–critic mechanisms to delayed-matching-to-place performance is consistent with neurobiological findings implicating the striatum and hippocampo-striatal interaction in delayed-matching-to-place performance, given that the striatum has been associated with actor–critic mechanisms. Moreover, we illustrate that hierarchical computations embedded within an actor–critic architecture may help to account for aspects of flexible spatial navigation. The hierarchical reinforcement learning approach separates trajectory control via a temporal-difference error from goal selection via a goal prediction error and may account for flexible, trial-specific, navigation to familiar goal locations, as required in some arm-maze place memory tasks, although it does not capture one-trial learning of new goal locations, as observed in open field, including watermaze and virtual, delayed-matching-to-place tasks. Future models of one-shot learning of new goal locations, as observed on delayed-matching-to-place tasks, should incorporate hippocampal plasticity mechanisms that integrate new goal information with allocentric place representation, as such mechanisms are supported by substantial empirical evidence.

Learning allocentric representations of space for navigation

The hippocampus of the mammalian brain supports spatial navigation by building cognitive maps of the environments in which the animal explores. Currently, there is little neurocomputational work investigating the encoding and decoding mechanisms of hippocampal neural representations in large-scale environments. We propose a biologically-inspired hierarchical neural network architecture to learn the transformation of egocentric sensorimotor inputs into allocentric spatial representation for navigation. The hierarchical network is composed of two parallel subnetworks mimicking the lateral entorhinal cortex (LEC) and medial entorhinal cortex (MEC), and one convergent subnetwork mimicking the hippocampus. LEC relays time-related visual information and MEC supplies space-related information in the form of multi-resolution grid codes as resulted from integrating movement information. The convergent subnetwork integrates all information from the parallel subnetworks and predicts the position of the agent in the environment. Synaptic weights of the vision-to-place and grid-to-place connections are learned based on the stochastic gradient descent algorithm. Simulations in a large virtual maze demonstrate that hippocampal place units in the model form multiple and irregularly-spaced place fields, similar to those observed in neurobiological experiments. The model is able to accurately decode the positions of the agent from the learned spatial representations. Moreover, the model is capable of adaptation to degraded visual inputs, and therefore is robust against perturbations. When the motion inputs are deprived, the model meets with localization difficulty, suffering from less accuracy in position predictions.

Neuroanatomic Correlates of Distance and Direction Processing During Cognitive Map Retrieval.

Navigating toward a goal and mentally comparing distances and directions to landmarks are processes requiring reading information off the memorized representation of the environment, that is, the cognitive map. Brain structures in the medial temporal lobe, in particular, are known to be involved in the learning, storage, and retrieval of cognitive map information, which is generally assumed to be in allocentric form, whereby pure spatial relations (i.e., distance and direction) connect locations with each other. The authors recorded functional magnetic resonance imaging activity, while participants were submitted to a variant of a neuropsychological test (the Cognitive Map Reading Test; CMRT) originally developed to evaluate the performance of brain-lesioned patients and in which participants have to compare distances and directions in their mental map of their hometown. Our main results indicated posterior parahippocampal, but not hippocampal, activity, consistent with a task involving spatial memory of places learned a long time ago; left parietal and left frontal activity, consistent with the distributed processing of navigational representations; and, unexpectedly, cerebellar activity, possibly related to the role of the cerebellum in the processing of (here, imaginary) self-motion cues. In addition, direction, but not distance, comparisons elicited significant activation in the posterior parahippocampal gyrus.

Learning Cognitive Map Representations for Navigation by Sensory-Motor Integration.

How to transform a mixed flow of sensory and motor information into memory state of self-location and to build map representations of the environment are central questions in the navigation research. Studies in neuroscience have shown that place cells in the hippocampus of the rodent brains form dynamic cognitive representations of locations in the environment. We propose a neural-network model called sensory-motor integration network model (SeMINet) to learn cognitive map representations by integrating sensory and motor information while an agent is exploring a virtual environment. This biologically inspired model consists of a deep neural network representing visual features of the environment, a recurrent network of place units encoding spatial information by sensorimotor integration, and a secondary network to decode the locations of the agent from spatial representations. The recurrent connections between the place units sustain an activity bump in the network without the need of sensory inputs, and the asymmetry in the connections propagates the activity bump in the network, forming a dynamic memory state which matches the motion of the agent. A competitive learning process establishes the association between the sensory representations and the memory state of the place units, and is able to correct the cumulative path-integration errors. The simulation results demonstrate that the network forms neural codes that convey location information of the agent independent of its head direction. The decoding network reliably predicts the location even when the movement is subject to noise. The proposed SeMINet thus provides a brain-inspired neural-network model for cognitive map updated by both self-motion cues and visual cues.

Urban Navigation System With Multiple Sub-Maps and Multiple Sub-Navigators

Deploying an autonomous robot in an urban environment with heterogeneous sub-area is a challenging task that requires different types of navigation and map representation. In a real-world setting, spontaneous change of the environment such as sudden raining or congestion of parts of the navigation area is also a common event that must be handled. This paper presents a real-world navigation system capable of switching among multiple navigation algorithms in a novel manner in order to properly handle the workspace with different characteristics and the challenge of changing condition of the environment. The design of the system is a real-world result from deploying an actual robot in a live system with real-world user and environment.

Spatial representation framework for better indoor navigation by people with visual impairment

PurposeThe purpose of this paper is to validate a framework for spatial representation, aka the spatial representation framework (SRF), to define spaces and building information required by people with visual impairment as a foundation of indoor maps for indoor navigation systems.Design/methodology/approachThe SRF was first created with seven main components by a review of the relevant literature and state-of-the-art technologies shown in the preliminary study. This research comprised of two tasks: investigating problems and behaviors while accessing spaces and buildings by visually impaired people (VIP) and validating the SRF where 45 participants were recruited (30 VIP and 15 experts).FindingsThe findings revealed a list of problems and challenges were used to validate and redefine the spatial representation, which was validated by both VIP and experts. The framework subsequently consisted of 11 components categorized into five layers, each layer of which is responsible for a different function.Research limitations/implicationsThis framework provides essential components required for building standard indoor maps as a foundation for indoor navigations systems for people with visual impairment.Practical implicationsThis framework lays the foundation for a range of indoor-based applications by using this SRF to represent indoor spaces. Example applications include: indoor navigation by people with disabilities, robots and autonomous systems, security and surveillance, and context and spatial awareness.Originality/valueThis paper presents the validated spatial representation for indoor navigation by people with visual impairment with its details and description, methodology, results and findings of the validation of the SRF.

Air-ground cooperative topometric mapping of traversable ground

In this paper, we propose an approach for cooperative mapping of traversable ground from aerial and ground views in structured outdoor and indoor environments. The presented approach achieves a hybrid map building based on traversable ground skeletonization and graph matching. The obtained map is an augmented ground traversability map, represented as a hybrid topological/metric graph from heterogeneous sources. This approach provides a very suitable representation for ground navigation and planning. To validate this approach, the proposed algorithm is applied between aerial views, provided by a UAV flying over an experimental site, and ground maps from ground robots at different exploration stages, in realistic simulation and real-world environments.

Vision-Based Robot Navigation through Combining Unsupervised Learning and Hierarchical Reinforcement Learning.

Extensive studies have shown that many animals’ capability of forming spatial representations for self-localization, path planning, and navigation relies on the functionalities of place and head-direction (HD) cells in the hippocampus. Although there are numerous hippocampal modeling approaches, only a few span the wide functionalities ranging from processing raw sensory signals to planning and action generation. This paper presents a vision-based navigation system that involves generating place and HD cells through learning from visual images, building topological maps based on learned cell representations and performing navigation using hierarchical reinforcement learning. First, place and HD cells are trained from sequences of visual stimuli in an unsupervised learning fashion. A modified Slow Feature Analysis (SFA) algorithm is proposed to learn different cell types in an intentional way by restricting their learning to separate phases of the spatial exploration. Then, to extract the encoded metric information from these unsupervised learning representations, a self-organized learning algorithm is adopted to learn over the emerged cell activities and to generate topological maps that reveal the topology of the environment and information about a robot’s head direction, respectively. This enables the robot to perform self-localization and orientation detection based on the generated maps. Finally, goal-directed navigation is performed using reinforcement learning in continuous state spaces which are represented by the population activities of place cells. In particular, considering that the topological map provides a natural hierarchical representation of the environment, hierarchical reinforcement learning (HRL) is used to exploit this hierarchy to accelerate learning. The HRL works on different spatial scales, where a high-level policy learns to select subgoals and a low-level policy learns over primitive actions to specialize on the selected subgoals. Experimental results demonstrate that our system is able to navigate a robot to the desired position effectively, and the HRL shows a much better learning performance than the standard RL in solving our navigation tasks.