CAPSTONE Project: Autonomous Fruit Fly Robot
The CAPSTONE project briefly described here is for the Fall 2019 and Spring 2020 semesters.
There is an increasing need for technologies in which data centers provide general intelligence to widely distributed entities -- drones, for example -- that also need to operate ("think") on their own to make critical decisions in real-time. A drone may receive a great deal of vital information from the Cloud, and yet the drone will confront situations in which its survival depends on near-instantaneous cognition and decision-making that cannot be delivered through wireless communication channels. For example, the drone suddenly may confront a flock of birds or another drone. But if a drone had the on-board intelligence of the lowly fruit fly, it could survive such a situation without relying on centralized guidance.
Fruit flies have a numerically small brain, yet they exhibit sufficiently complex behavior including locomotor responses to sensory inputs, e.g., optomotor avoidance responses, and self-driven tasks such as navigation and searching for food sources. Recently, there has been an enormous amount of progress in mapping fruit fly brain circuits on the cellular, connectome and activity levels. This has led to effective visualization of queryable neural circuits and their mapping into executable circuit diagrams on multiple levels of abstraction.
With these capabilities and the ever-increasing knowledge of the fruit fly brain, the ultimate goal of the project is to build autonomous robotic systems with on-board intelligence designed from the bottom up by emulating the actual, known circuitry of the fruit fly brain -- a marvel of efficient, adaptable, physically compact intelligence.
- Qualified students must be well versed in Python.
- Prior exposure to interactive computing (e.g., JupyterLab) is a plus.
- Exposure to neuroscience/biology not required but is a plus.
How to Apply:
Please send your CV/Resume to firstname.lastname@example.org. Please also include any code samples, Github repositories of related previous work if possible. If you have any question, please send inquiries to the above email address.