Descrizione del progetto
Dagli insetti ai robot autonomi
Gli insetti che volano rappresentano il modello migliore per lo studio della vista in volo dato il loro sistema nervoso semplice e l’ottica fissa dei loro occhi composti. Considerato il fatto che il crescente mercato della robotica autonoma si basa sulla guida visiva, lo studio del modo in cui questi insetti orientano il loro sguardo in volo si dimostra utile. Il progetto Vision-In-Flight, finanziato dall’UE, si propone di condurre uno studio pionieristico sui meccanismi neurali della vista degli insetti, concentrando l’attenzione sul controllo diretto dello sguardo nel tracciamento degli oggetti. Avvalendosi di un’acquisizione delle immagini in movimento a elevata precisione, di una telemetria neurale wireless ultraleggera e della realtà virtuale, il progetto codificherà la vista degli insetti in manovre di movimento veloci al fine di ottenere un controllo della qualità senza precedenti nei sistemi autonomi.
Obiettivo
This project investigates how biological vision operates under the fastest and most challenging motion condition: flight. Specifically, we look beyond gaze stabilization and focus on directed gaze control such as object tracking. Flying insects are ideal model for studying vision in flight due to its relatively simple nervous system and the fixed optics of the compound eyes. Insect vision has elucidated fundamental circuitries of vision via psychophysics, electrophysiology, computational modelling, and connectomics. However, we have limited knowledge on how insects use vision in free flight and what visual signals influence motor control during aerial interactions. This study aims to reveal how flying insects direct their gaze in-flight to extract target information for guidance and to facilitate the execution of complex flight manoeuvres. To achieve this objective, we will advance three emerging techniques: 1) high-precision insect scale motion capture; 2) ultralight wireless neural telemetry; 3) virtual reality for freely flying insects. I was involved in developing the first two methods and they both still require significant development to suit this project. The third budded from a successful ERC project, which enabled virtual reality experiments with freely behaving animals, and also requires additional breakthrough in order to accommodate this project. By advancing these techniques together, we can fully control the visual input of a freely flying insect and simultaneously record relevant visual signals. While modern image sensors and image processing can sometimes surpass biological vision, machine vision systems today still cannot utilize some tactical benefits of directed gaze control. Indeed, learning how to look is one of the best lessons a visually guided system can take from biology. This research informs the control of autonomous systems such as self-driving cars, unmanned aerial taxi, and robotic courier which will revolutionize the upcoming era.
Campo scientifico
- natural sciencesbiological sciencesneurobiology
- natural sciencescomputer and information sciencesartificial intelligencecomputer vision
- natural sciencesphysical sciencesoptics
- natural sciencesbiological scienceszoologyentomology
- natural sciencescomputer and information sciencessoftwaresoftware applicationsvirtual reality
Parole chiave
Programma(i)
Argomento(i)
Meccanismo di finanziamento
ERC-STG - Starting GrantIstituzione ospitante
SW7 2AZ LONDON
Regno Unito