Small brains, complex challenges…
The superb navigational skills of ants and bees are proof that small brains can achieve amazingly efficient behaviours in complex environments. Our lab seeks to uncover the link between neural dynamics and behavioural decisions, when these insects navigate in their natural habitat. We focus on ant species that are champion visual navigators, who spontaneously learn long routes in various habitats… and actually, many ant species do so!
We combine behavioural experiments and computational models based on the insect neural circuitry. The models enable to explore the complex relationship between the brain, body, and environment; and (hopefully) produce testable behavioural predictions. The behavioural experiments enable to test the models’ predictions, and (hopefully) unravel unsuspected abilities in these little creatures.
Every year, we try to do one or two trips to the field. April to July is the time to go to Seville to study desert ants (Cataglyphis velox). November to March is the time to go to Canberra to study various Aussie ants such as Jack Jumpers (Myremecia pyriformis) or meat ants (Iridiomyrmex purpureus). Field studies often means very long working days, but also lots of fun, and sometimes great discoveries. A vast amount of experiments can be achieved with just a tube, a pen and paper, although others may require more sophisticated equipment. In any case, it is definitely in the field that our research is the most productive. Observing our animal species in their natural environment is also a profound source of inspiration for future, ecologically relevant research questions. And most importantly, it’s a time to learn to better know each other and live a little adventure as a team.
One of our new gadgets! Developed mainly by Hansjuergen Dahmen. Instead of running after your ant, just grab it, attach it by a thread on the thorax and put it on a ball that is floating in the air. Then, when the ant tries to move, the ball turns, but she stays on the spot. That way, we can decouple the motor action from the expected visual feedback (feedback which is, in such case, static). This system has enabled us to ask novel questions about the ways ants process the world for navigation, some already published, and others to come. We are definitely not finished yet with this tool!
Once you have an ant running on a trackball, you can actually make it navigate in a virtual world. For that, what you need is a very good panoramic visual display all around the ant. One such display is currently in ANU, Canberra, within our collaborator Prof. Jochen Zeil’s group. It is a big display – more than a meter across, and composed of 20,000 LEDs including UV. The advantage in Canberra is that you can pick a wild ant right in front of your door, which is trained in the real world, and test it in the VR. We are also developing a second type of display using video-projectors in the university of Toulouse, in collaboration with Dr. Andrew Straw. This one is tuned for ants raised in the lab, whom we directly train to navigate in the VR. Needless to say, the VR system enables us to do any instantaneous transformation of the visual environment while the ant is navigating, so the number of possible questions to ask is vast!
Our goal is to understand how the insect brain enables these little creatures to display such amazing navigational skills. But how can we formulate a hypothesis about brain mechanisms? Instead of using words, we are using neural models. We equip a simplified agent with one of our modelled brains, toss it in a 3D reconstructed ant world, and see how well it performs! One the one hand, our neural models are purely constrained by our actual knowledge of real insect circuits, and on the other hand, we compare the navigational behaviour outputted by the agent to the behaviour of real ants. That way, we hope to slowly refine our models, and thus understand some fundamental principles of how the insect brain works. Sometimes we use our models to try and explain a well-known behavioural feat, other times, we use it to predict the existence of a yet unobserved ability; which leads us straight back to the field. Here are some examples of published models.