29. 6. 2015, University Vienna
• The Plan
• The Architecture
• The Process
• Technical Specifications
• Image Capture
• PID Controller
• Trials and Debugging
• Lessons Learned
• Build a robot that detects and
item in a room and picks it up
• Detecting an item
• Discriminating task (blue or red)
• Exploration (Random walk)
• Lifting up device
(embodied, maybe even just a
paper device on a physical level)
NXT Brick GoPro
Send Motor Commands
Get video feed
• Camera Integration
• Algorithm for Object Detection
• Random Walk and Navigation
• Find out wheel speeds according to
approaching of the goal
• Feedback Control
(power values for different motors)
• Mechanical Lifting up Mechanism
• Camera should be centered in
front of the wheels
• Ball lifting mechanism purely
• Ball should disappear behind
Successfully picking up a ball
observable, size not fixed, one
or two red/blue objects
• Building robots is fun.
• It’s way more difficult than expected.
• Installing the framework is more complicated than
running it. These tasks are insanely platform
• It needs a lot of installing of libraries and packages to
actually be able to start the work.
• Difficulty testing with 1 brick and 4 people.
• NXT doesn’t like Yosemite.
• Red fingernails are the same to a robot as a red ball.
Everything yellowish seems to be red, too.
• Each task needs a motor.
• Weight and volume matter big time in objects.
• Pairing with Bluetooth sucks.
• Batteries run out quickly.
• If it isn’t tested, it doesn’t work.
• Lots of trial and error.