URSA-ENGAGE PROGRAM

Faculty Project Proposal Summary

Calendar Year 2014 Program

 

 

Faculty Name & Department/School:  Cindy Grimm, MIME

 

Research Project Title:  Bat ears for robots

 

Project Overview: 

Building a physical sonar system, modeled off of the way bat sonar works, for doing simple distance detection on a robot. The device consists of an ultra-sound generation and receiver, 3D printed ears and noses based on CT scans of actual bats, and a robotic platform. The device will be placed on the robot in order to collect both sonar and distance (from a laser scanner) data simultaneously. Basic machine learning will then be used to “learn” how the sonar patterns map to distances.

Note: We have built a very basic proto-type system, but have not yet tested it on the robot. The existing system and data processing was accomplished with first-year students.

 

What skills will students obtain in this project?

There are several components to this project; students may work on any subset of them:

  1. (Modeling:) 3D software modeling of bat ears/noses, and how to print them using a 3D printer.
  2. (Electrical engineering:) How to wire up a simultaneous ultra-sound transmitter/receiver using, eg, the Arduino tool kit.
  3. (Machine learning/computer science:) How to apply basic machine learning algorithms to a data set (using existing MATLAB or C/C++ implementations)
  4. (Mechanical engineering:) How to mount and configure the device on a physical robot.
  5. (Robots/computer science:) How to navigate a robot and collect data from the laser sensor using the ROS (Robot Operating System) language.
  6. (Shape correspondence/mathematics:) How to build a mathematical model of the shape variation of a set of bat noses/ears.
  7. Learn how bat sonar works

 

Student research tasks:

There are several components to this project; students may work on any subset of them:

  1. Mount the device on the robot and collect data
    1. Mounting the circuits & plastic parts on a robot and synchronizing with the robot's laser input (note: grad student help on the robot available - plastic parts and support platform already made).
    2. Learn to navigate the robot and drive it around to collect data
    3. Physically connect the device’s sensors to the robot’s usb port to simultaneously collect the sonar data
  2. Apply machine learning to the collected data
    1. Process the data into a form readable by existing machine learning programs
    2. Run the data and analyze the results
  3. Building a more complex electrical circuit to create a cyclical output signal at different frequencies (25-40khz) using an existing transducer (a range finder from an old Kodak camera)
    1. Receiver for that signal, time synchronized to the output signal
  4. Building a range of noses/ears based on existing CT scans of real noses and ears
    1. Manual marking of existing data sets to put them in correspondence
    2. Using interactive software to process the data/correspondences
    3. Apply sonar analysis software to determine how changing the shape changes the functionality of the sonar

 

Number of hours per week expected of student: 3-4

 

Preferred Faculty Contact Information (email): cindy.grimm@oregonstate.edu

 

________________________________________________________________________________________

Download student application HERE

Back to Main Page HERE