Biorobotics

We have recently begun a series of simulation experiments which are aimed at modeling and comparing very simple "time honored", but relatively untested, hypotheses about the rules that animals may be using to locate distant unseen odor sources. We are pursuing an explicitely iterative approach using realistically designed simulation models to test hypotheses that can then be tested using carefully repeated experiments with odor tracking insects. Thus far, we have created a spatial model (illustrated above) in which the simulated agent's only behavioral choice is to turn toward the side of highest odor concentration. In this video you can see the track of the agent turning toward the lighter circles of odor packets. This model has been successful at finding the source if the environment contains an edge to follow. We have also developed a less successful temporal model. In this model the simulated agent compares the odor information it has just received with that detected at the immediately previous timepoint and either continues forward (if the concentration is the same or higher) or executes a turn (if the concentration decreases). The video shows a typical simulation run in which the agent does not make it to the odor source. Modifications to this strategy based on hypothetical mechanisms underlying odor orientation in animals are ongoing.

Planned experiments will include challenging male cockroaches to track pheromone plumes in our wind tunnel environment which has been manipulated to match the simulation's environments, and experimentally manipulating the cockroach sensors (i.e., surgically removing parts or all of an antennae) in parallel with sensor manipulations to the simulated agents in an attempt to deprive a "bilateral comparison" agent access to spatial information about it's odor environment. The aim being to study the tracking performance of a spatial agent that only has access to temporal information about the odor.

We also plan to use this approach to compare the performance of tracking agents and animals that walk vs. fly by using a 3D simulation environment to support experiments with flying moths and a 2D simulation environment to support our experiments with walking cockroaches. Future directions are to manipulate the speed and size of the simulated agent to more closely match the scale of the insects in their environments and to incorporate a realistic dynamic model plume in the simulation environment.