Nocturnal Visual Orientation in Flying Insects: A Benchmark for the Design of Vision-Based Sensors in Micro-Aerial Vehicles
Abstract
One of the prerequisites for achieving autonomous flight, in machines as well as organisms, is the maintenance of a stable attitude relative to the vertical direction. In order to obtain attitude information, it is possible to use visual cues, such as the mean direction of illumination, the position of the horizon or optic flow. Recently, visual attitude stabilization has been implemented in a number of designs for micro-aerial vehicles. It is also being used by flying insects, and many of them carry ocelli, a supplementary set of simple lens eyes that appear to be specialized for attitude stabilization. These ocelli are particularly well developed in dragonflies, which are among the most competent fliers in existence. We have explored the dragonfly ocellar system, identifying the spatiotemporal transfer functions of its components, namely the dioptric system, the photoreceptor neurons, and, most importantly, the secondorder neurons which are the final stages in processing visual information. The outputs of a total of 15 neurons form a map of part of the visual world, covering a wide range (200 deg) of azimuth along the horizon. In elevation, the range is narrow (30 deg) and different elevations are not mapped onto different neurons, but the outputs of all neurons contain a component that is sensitive to optic flow in elevation. Thus, it appears that the dragonfly ocellar system uses an approach that distinguishes it from existing technical horizon detection algorithms. Our knowledge about details of the system is sufficient to implement it in hardware and to commence testing in realistic situations.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 09, 2011
- Accession Number
- ADA538637
Entities
People
- Gert Stange
Organizations
- Australian National University