Mobile robot visual mapping and localization: A view-based neurocomputational architecture that emulates hippocampal place learning

Echo View Cells From Bio-Inspired Sonar

Place recognition is naturally informed by the mosaic of sensations we remember from previously visiting a location and general knowledge of our location in the world. Neurons in the mammalian brain (specifically in the hippocampus formation) named “place cells” are thought to reflect this recognition of place and are involved in implementing a spatial map that can be used for path planning and memory recall. In this research, we use bat-inspired sonar to mimic how bats might sense objects in the environment and recognize the views associated with different places. These “echo view cells” may contribute (along with odometry) to the creation of place cell representations observed in bats. Although detailed sensory template matching is straightforward, it is quite unlikely that a flying animal or robot will return to the exact 3-D position and pose where the original memory was captured. Instead, we strive to recognize views over extended regions that are many body lengths in size, reducing the number of places to be remembered for a map. We have successfully demonstrated some of this spatial invariance by training feed-forward neural networks (traditional neural networks and spiking neural networks) to recognize 66 distinct places in a laboratory environment over a limited range of translations and rotations. We further show how the echo view cells respond between known views and how their outputs can be combined over time for continuity.

Encoding Natural Movement as an Agent-Based System: An Investigation into Human Pedestrian Behaviour in the Built Environment

Gibson's ecological theory of perception has received considerable attention within psychology literature, as well as in computer vision and robotics. However, few have applied Gibson's approach to agent-based models of human movement, because the ecological theory requires that individuals have a vision-based mental model of the world, and for large numbers of agents this becomes extremely expensive computationally. Thus, within current pedestrian models, path evaluation is based on calibration from observed data or on sophisticated but deterministic route-choice mechanisms; there is little open-ended behavioural modelling of human-movement patterns. One solution which allows individuals rapid concurrent access to the visual information within an environment is an ‘exosomatic visual architecture’, where the connections between mutually visible locations within a configuration are prestored in a lookup table. Here we demonstrate that, with the aid of an exosomatic visual architecture, it is possible to develop behavioural models in which movement rules originating from Gibson's principle of affordance are utilised. We apply large numbers of agents programmed with these rules to a built-environment example and show that, by varying parameters such as destination selection, field of view, and steps taken between decision points, it is possible to generate aggregate movement levels very similar to those found in an actual building context.

Robustness of Visual Place Cells in Dynamic Indoor and Outdoor Environment

In this paper, a model of visual place cells (PCs) based on precise neurobiological data is presented. The robustness of the model in real indoor and outdoor environments is tested. Results show that the interplay between neurobiological modelling and robotic experiments can promote the understanding of the neural structures and the achievement of robust robot navigation algorithms. Short Term Memory (STM), soft competition and sparse coding are important for both landmark identification and computation of PC activities. The extension of the paradigm to outdoor environments has confirmed the robustness of the vision-based model and pointed to improvements in order to further foster its performance.

BISMARC: a biologically inspired system for map-based autonomous rover control

As the complexity of the missions to planetary surfaces increases, so too does the need for autonomous rover systems. This need is complicated by the power, mass and computer storage restrictions on such systems (Miller, D. P. (1992). Reducing software mass through behaviour control. In Proceedings SPIE conference on cooperative intelligent robotics in space III (Vol. 1829, pp. 472-475, 1992). Boston, MA. To address these problems, we have recently developed a system called BISMARC (Biologically Inspired System for Map-based Autonomous Rover Control) for planetary missions involving multiple small, lightweight surface rovers (Huntsberger, T. L. (1997). Autonomous multirover system for complex planetary retrieval operations. In P. S. Schenker, and G. T. McKee (Eds.), Proceedings SPIE symposium on sensor fusion and decentralized control in autonomous robotic systems (pp. 221-227). Pittsburgh, PA). BISMARC is capable of cooperative planetary surface retrieval operations such as a multiple cache recovery mission to Mars. The system employs autonomous navigation techniques, behavior-based control for surface retrieval operations, and an action selection mechanism based on a modified form of free flow hierarchy (Rosenblatt, J. K. and Payton, D. W. (1989). A fine-grained alternative to the subsumption architecture for mobile robot control. In Proceedings IEEE/INNS joint conference on neural networks (pp. 317-324). Washington, DC). This paper primarily describes the navigation and map-mapping subsystems of BISMARC. They are inspired by some recent studies of London taxi drivers indicating that the right hippocampal region of the brain is activated for path planning but not for landmark identification (Maguire, E. A. et al. (1997). Recalling routes around London: activation of the right hippocampus in taxi drivers. Journal of Neuroscience, 17(18), 7103-7110). We also report the results of some experimental studies of simulated navigation in planetary environments.

Biologically based artificial navigation systems: review and prospects

Diverse theories of animal navigation aim at explaining how to determine and maintain a course from one place to another in the environment, although each presents a particular perspective with its own terminologies. These vocabularies sometimes overlap, but unfortunately with different meanings. This paper attempts to define precisely the existing concepts and terminologies, so as to describe comprehensively the different theories and models within the same unifying framework. We present navigation strategies within a four-level hierarchical framework based upon levels of complexity of required processing (Guidance, Place recognition-triggered Response, Topological navigation, Metric navigation). This classification is based upon what information is perceived, represented and processed. It contrasts with common distinctions based upon the availability of certain sensors or cues and rather stresses the information structure and content of central processors. We then review computational models of animal navigation, i.e. of animats. These are introduced along with the underlying conceptual basis in biological data drawn from behavioral and physiological experiments, with emphasis on theories of "spatial cognitive maps". The goal is to aid in deriving algorithms based upon insights into these processes, algorithms that can be useful both for psychobiologists and roboticists. The main observation is, however, that despite the fact that all reviewed models claim to have biological inspiration and that some of them explicitly use "Cognitive Map"-like mechanisms, they correspond to different levels of our proposed hierarchy and that none of them exhibits the main capabilities of real "Cognitive Maps"--in Tolman's sense--that is, a robust capacity for detour and shortcut behaviors.