I read this article: "Eight pairs of descending visual neurons in the dragonfly give wing motor centers accurate population vector of prey direction". I found it really important and really interesting.
Do you think it could be used to develop the possibilities of an autonomous UAV?
I also think it's interesting, because the autonomous drone (but not 100%) are now sophisticated enough to make landing maneuvers, but, with an unexpected obstacle and this is the crash.
So I think we need to find ways to teach these drones how to avoid obstacles, with keeping their flexibility of movement
Which control technique you would use to make an obstacle avoidance software?
Excellent question, one I have wondered about for years.
From the article, there are 16 views of a dragonfly's eyes: topview showing left and right sides (8 views) and side view showing vertical and dorsal sides (8 views) in
Fig. 4. The target-descending neurons (TSDNs) code a population vector that reflects the direction of a target with high accuracy and reliability across 360 degrees. The views of a dragonfly's eyes are shown in the attached image.
So it can see in all directions as the Camelion.
To build an autonomous drone that has the characteristics of desdragonfly vision, use the sensors distances (as Sharp IR Sensor) and camera motion capture, and make learning with software image processing.
Answer to Dr Peters: I started my analysis from the article you've cited. Which AI paradigm you think could be used to create a model of this situation?
( You think I could start from the reactive paradigm?)
Answer to Dr Bekkaye: I agree with you, so I could use a neural network to predict the evolution of a dynamical system?
Dear Filioppo,
Before starting a model of a vision system inspired by dragonfly vision, I suggest that a more comprehensive study of dragonfly vision be cosidered first. The better our understanding of the dragonfly, the better our eventual vision system will be.
Dear Filippo,
I am checking on articles to suggest. In the meantime, perhaps you will find the attached paper interesting.
Implementing a drone which has the same characteristics of a dragonfly requires a biological study of dragonfly in order to understand the flight mechanics. and how is the translation of the image (what the dragonfly see ) to a physical signal (the movement of avoided the obstacle for example)
The artificial neural network is the best way to predict the evolution of a dynamical system because it can be used for pattern recognition, eg for optical character recognition (OCR), and automated movement of autonomous mobile robots, and it can be used from: C, C + +, PHP, Python, Delphi and Mathematica.
And as I said at the beginning, the artificial neural networks require real-life case examples used for learning (this is called the training set)
@Filippo : I'm note yet a Dr, I'm electronics ING and Biology Master
Filippo,
A dragonfly has compound eyes. Each eye has up to 28,000 simple eyes (ommatidia) that give a mosaic view of the world. See the attached pdf file.
Filippo,
I search through our university library and found some good articles on dragonfly vision. I start posting them for you and the followers of this thread. The first of these articles is
E. Eguchi, Fine structure and spectral sensitivities of retinal cells in the dorsal sector of compound eyes in the dragonfly Aeschna, Z. vergl. Physiologie 71, 1971, 201-218.
See attached pdf file.
Here is a longitudinal view of an Aeschna cyanea compound eye. The four filled circles represent proximal processes of the crystalline cone.
Thank you very much for the articles you're posting Dr Peters. They are really interesting and helpful for the project I'm trying to start.
I have to read them well, I'm not an expert in the field of Biology.
I think it is really important to understand dragonfly aerodynamics because it is characterized by low Reynolds numbers, a UAV (unmanned aerial vehicle) is not characterized by the same Reynolds numbers ( I'm not considering the idea of designing a MAV, micro aerial vehicle).
I have to work on the characteristics of the dragonfly I can transfer in a UAV, whose dimensions are bigger.
I have been really busy in the last period.
I didn't forget this problem.
I think I will continue working on it during this winter.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems