Biologically plausible models of place recognition and goal location
Parallel distributed processing: explorations in the microstructure of cognition, vol. 2
1994 Special Issue: A model of hippocampal function
Neural Networks - Special issue: models of neurodynamics and behavior
Spatial representation for navigation in animats
Adaptive Behavior
A mobile robot that learns its place
Neural Computation
Self-orienting with on-line learning of environmental features
Adaptive Behavior - Special issue on biologically inspired models of navigation
Learning View Graphs for Robot Navigation
Autonomous Robots - Special issue on autonomous agents
A model of spatial map formation in the hippocampus of the rat
Neural Computation
Constructing maps for mobile robot navigation based on ultrasonic range data
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
Spatial learning for navigation in dynamic environments
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
Hi-index | 0.00 |
This paper considers the problem of environment-mapping forautonomous mobile agents. Central to this problem is how thestructure of the external environment can be captured and representedso that an agent is able to autonomously navigate in a robust way.This is an issue that is not clear even given a perfect sensor systemthat provides all information that the agent needs. Biological datareveal that the activation of hippocampal place cells in an animal,which is performing navigational tasks, is highly correlated with theanimal‘s location, and the sensory basis of place cell firing is ofmultiple modalities which include landmark detection. As a functionalapproximation to hippocampal place learning, this paper presents adynamic network that can be used by an autonomous agent to maplandmarks, places and the spatial relation between them. The spatialrelation is encoded by embedding an areal coordinate coding principleinto the inter-cell connection structure. For the network to be usedto map a large space, a focusing mechanism is introduced, which notonly constrains the scope of the network in which learning can takeplace, but also limits the computation needed to the part of thenetwork that is currently relevant to the activity of the agent. Thisfocusing mechanism can also be used to direct dynamic route-findingin the network that the agent builds. Simulation results of thenetwork demonstrate its applicability and computationalcharacteristics.