Path integration system of insects: there and back again

Abstract

Navigation is an essential capability of animate organisms and robots. Among animate organisms of particular interest are insects because they are capable of a variety of navigation competencies solving challenging problems with limited resources, thereby providing inspiration for robot navigation.

Ants, bees and other insects are able to return to their nest using a navigation strategy known as path integration. During path integration, the animal maintains a running estimate of the distance and direction to its nest as it travels. This estimate, known as the `home vector', enables the animal to return to its nest.

Path integration was the technique used by sea navigators to cross the open seas in the past. To perform path integration, both sailors and insects need access to two pieces of information, their direction and their speed of motion over time. Neurons encoding the heading and speed have been found to converge on a highly conserved region of the insect brain, the central complex. It is, therefore, believed that the central complex is key to the computations pertaining to path integration.

However, several questions remain about the exact structure of the neuronal circuit that tracks the animal's heading, how it differs between insect species, and how the speed and direction are integrated into a home vector and maintained in memory. In this thesis, I have combined behavioural, anatomical, and physiological data with computational modelling and agent simulations to tackle these questions.

Analysis of the internal compass circuit of two insect species with highly divergent ecologies, the fruit fly Drosophila melanogaster and the desert locust Schistocerca gregaria, revealed that despite 400 million years of evolutionary divergence, both species share a fundamentally common internal compass circuit that keeps track of the animal's heading. However, subtle differences in the neuronal morphologies result in distinct circuit dynamics adapted to the ecology of each species, thereby providing insights into how neural circuits evolved to accommodate species-specific behaviours.

The fast-moving insects need to update their home vector memory continuously as they move, yet they can remember it for several hours. This conjunction of fast updating and long persistence of the home vector does not directly map to current short, mid, and long-term memory accounts. An extensive literature review revealed a lack of available memory models that could support the home vector memory requirements.

A comparison of existing behavioural data with the homing behaviour of simulated robot agents illustrated that the prevalent hypothesis, which posits that the neural substrate of the path integration memory is a bump attractor network, is contradicted by behavioural evidence.

An investigation of the type of memory utilised during path integration revealed that cold-induced anaesthesia disrupts the ability of ants to return to their nest, but it does not eliminate their ability to move in the correct homing direction. Using computational modelling and simulated agents, I argue that the best explanation for this phenomenon is not two separate memories differently affected by temperature but a shared memory that encodes both the direction and distance.

The results presented in this thesis shed some more light on the labyrinth that researchers of animal navigation have been exploring in their attempts to unravel a few more rounds of Ariadne's thread back to its origin. The findings provide valuable insights into the path integration system of insects and inspiration for future memory research, advancing path integration techniques in robotics, and developing novel neuromorphic solutions to computational problems.