Scholarship robot idea: sound explorer/collector
I'm a wildlife biologist currently between jobs, so my budget for new cool gear like the iRobot Create is pretty much nil. But I have an idea for using the Create as a prototype platform for a field-usable device.
Previous work has established the utility of robotics in obtaining data concerning wildlife behavior (e.g., Patricelli, G. L., J. A. C. Uy, et al. (2006). "Interactive signaling during mate choice in the Satin Bowerbird: an experimental test using robotic females." Journal of Ornithology 147(5): 12-12.)
My background has included electrical engineering, computer programming, and bioacoustics. What I would be interested in trying my hand at is using the iRobot Create to set up an autonomous robot to localize sound, orient toward it, approach, and obtain sound recordings. I believe that this an achievable project in the contest timeframe. I already have microphones and amplifiers suitable for the project. I can either use three or more in a planar array to use time-of-arrival for sound localization, or use material to make each sensor directional and calculate heading from amplitude differences. In any case, the sound sensors would be used to obtain heading information for the robot. Acoustics boils down to a simple dictum: get a good microphone close to your subject. In this case, a robotic platform stands a much better chance of approaching wildlife without causing the sort of disturbance the presence of a human observer usually makes. While the Create platform itself may not be suitable for field deployment, I think that it would make a good prototype for lab testing of the concept. In that light, I will probably use a voice recorder as a lightweight recording system that will only require two actuators to have the robot operate it.
There are some additional considerations that I will need to explore, such as figuring out how to determine closeness of approach. This would eventually need to be based upon calculations of whether the robot is at the boundary of far and near-fields of an acoustic source at the highest frequency of interest. That would maximize the sound amplitude at the recorder without the interference that is the defining characteristic of the near-field regime.
If the prototype works well, there are further ideas that could go into a field robot system. One would be to have the robot localize a source and conduct a survey, carrying a GPS mapping system while taking sound recordings with a calibrated sound recording system. This would be useful to characterize sound propagation properties of the animal's emitted sound. As in Patricelli's work, the field system could be disguised as a conspecific, or at least a species that is likely to be considered "harmless" by the subject.