Advanced searches left 3/3

Insect Flight - Astrophysics Data System

Summarized by Plex Scholar
Last Updated: 16 January 2022

* If you want to update the article please login/register

Adaptive numerical simulations of insect flight using wavelet techniques

The blocks' distribution among MPI processes allows for a fast parallelization of large-scale supercomputers. A centered 4th-order finite difference discretization is mixed with biorthogonal interpolating wavelets as grid refinement indicators. On massively parallel computer architectures, multiple validation scenarios are used to determine the reliability and performance of the open access code WABBIT. The same realistic fly bodies obtained from mirco-CT scans are also used in first computations.

Source link: https://ui.adsabs.harvard.edu/abs/2021APS..DFDP13012E/abstract


Excitable spring-wing dynamics 1: Common dynamics and transitions between synchronous and asynchronous insect flight

Following a stretch that has seen a pair of opposing dSA muscles produce emergent cycle habits, dSA is a delayed rise in force. We unify the flight muscle's regular drive and dSA activities within a single framework here. We then discover that synchronous moth muscle excerpts include dSA, but that it is insufficient to result in emergent self-excitation. We combined the two types of actuation with a resonant spring-wing mechanics model to investigate how discrete oscillatory behavior emerges from a continuum of actuator properties.

Source link: https://ui.adsabs.harvard.edu/abs/2021APS..MARB14002G/abstract


Active vision improves sensory acquisition and coordinates motor control in insect flight

We investigated how fruit flies overcame these obstacles by actively controlling their head to influence the visual inputs that are perceived by the eyes. We discovered that head movements quickly slowed down visual disruptions, thus reducing motion blur and enabling flies to record visual movement speeds that were twice as fast as new insect vision purports. The head's instructions were followed by 30 ms, revealing a temporal order in which the head filters visual data that then directs downstream wing steering efforts.

Source link: https://ui.adsabs.harvard.edu/abs/2021APS..MARB14006C/abstract


Insect-like flying robots

In recent years, interest in building mm-scale flying cars has increased, both for at-scale studies of insect flight and collective behaviour, as well as research into environmental monitoring, structural inspection, search & rescue, and archaeological studies. These developments contributed to the untethered flight of a 90-mg four-wing vehicle carrying 170 grams of electronics and solar cells. This vehicle demonstrated thrust-per-muscle-mass and system thrust-efficiency matching that of traditional biological counterparts such as bees, while still lacking onboard guidance and monitoring, as well as the modeling framework, which shows promise for beating these figures in the future.

Source link: https://ui.adsabs.harvard.edu/abs/2021APS..MARB14005J/abstract


Pruning deep neural networks generates a sparse, bio-inspired nonlinear controller for insect flight

To solve the inverse problem of flight control, we create a framework that incorporates model predictive control of a well-known flight dynamics model and deep neural networks. Since they inherently demonstrate network pruning, we can expect more effective networks for a specific group of tasks, we can be inspired by nature. This bio-inspired strategy helps us to optimally sparsify a DNN database in order to perform flight tasks with as few neural links as possible, but there are limits to sparsification. Given initial random weights of the DNNs, Monte Carlo simulations also investigate the statistical distribution of network weights during pruning. According to statistics, the network can be reduced to save approximately 7% of the original network weights on average, with network load estimates determined at each layer of the network. We show that on average, the network can be reduced to save approximately 7% of the original network weights, with network weights kept at each layer of the network.

Source link: https://ui.adsabs.harvard.edu/abs/2022arXiv220101852Z/abstract

* Please keep in mind that all text is summarized by machine, we do not bear any responsibility, and you should always check original source before taking any actions

* Please keep in mind that all text is summarized by machine, we do not bear any responsibility, and you should always check original source before taking any actions