Application of a real neural collision avoidance system based on the locust to AGV navigation

摘要

Abstract: erb aereal performance of flying insects isachieved with comparatively simple neural machinery.Insects react rapidly to changing visual images. Theabilities of insects to perform these computations inreal time has already led to a successful prototypeautonomous guided vehicle with a sensor and controlstructure modelled on the fly eye. Increasingly invisual neuroscience it is possible to isolate thecritical image cues used by identified neurones toachieve a selective response to a feature or group offeatures within the changing visual image. In thispaper we describe a biological neural network based onthe input organization of such an identified motiondetecting neurone, which responds selectively to theimages of an object approaching on a collision coursewith the animal. We compare the response of theartificial neural network with the biological neuralnetwork in the same colliding stimulus. This approachled to a series of testable predictions about theorganization of the biological neural network.!14