Emergent spatial goals in an integrative model of the insect central complex

Abstract

The insect central complex appears to encode and process spatial information through vector manipulation. Here, we draw on recent insights into circuit structure to fuse previous models of sensory-guided navigation, path integration and vector memory. Specifically, we propose that the allocentric encoding of location provided by path integration creates a spatially stable anchor for converging sensory signals that is relevant in multiple behavioural contexts. The allocentric reference frame given by path integration transforms a goal direction into a goal location and we demonstrate through modelling that it can enhance approach of a sensory target in noisy, cluttered environments or with temporally sparse stimuli. We further show the same circuit can improve performance in the more complex navigational task of route following. The model suggests specific functional roles for circuit elements of the central complex that helps explain their high preservation across insect species.

Investigating visual navigation using spiking neural network models of the insect mushroom bodies

“…The central complex plays a key role in integrating information from various sources to provide heading directions for navigation ( Honkanen et al, 2019 ). The Drosophila central complex contains the ellipsoid body which encodes heading directions as bumps of activity with dynamics similar to that of a ring attractor network ( Kim et al, 2017 ), and some predict that MB output could affect an animals heading direction by acting on this region and surrounding areas ( Collett and Collett, 2018 ; Goulard et al, 2023 ).…”

“…Some take an algorithmic approach, where they look at the theorised processing that occurs in the mushroom body, and use analogues found in computer vision techniques to try and recreate navigation behaviours by embodying models in robotic or simulated agents ( Möel and Wystrach, 2020 ; Stankiewicz and Webb, 2021 ). Another approach has been to create (rate-based) artificial neural networks (ANNs) which follow the general neural architecture of the MB and embody this in a robot vehicle or agent based simulation ( Gattaux et al, 2023 ; Yihe et al, 2023 ), or utilise it as part of a larger navigation system simulating other navigation-related brain areas ( Sun et al, 2020 ; Goulard et al, 2023 ). Finally some use spiking neural networks (SNNs) to create navigation models that mimic both the architecture and neuronal dynamics of the MB ( Ardin et al, 2016 ; Müller et al, 2018 ; Zhu et al, 2021 ), or use SNNs as one component of a hybrid SNN/algorithmic navigation system ( Nowak and Stewart, 2019 ).…”