Module 9: Image-Driven Behavior

For this module, use the image matcher from Module 8 as follows:

• You may program the robot to do anything you would like, so long as at least some of its behavior is controlled by classification of images.
• You should subscribe to at least two topics beyond the image-labeling topic. Both subscriptions should impact the robot’s behavior in at least some circumstances.
• You may use any concept we have explored in any of the course modules. If you choose to use Q-learning, what you do must go beyond it in some way - perhaps Q-learning is only employed for action selection in certain circumstances.

Evaluation

• Design an experiment (along the lines of those in Modules 7 and 8) to assess the robot’s performance.

In your final document for the module, include the following:

• Discuss your task
  • How does your image database help the robot achieve the task?
    • What was your strategy for collecting and labeling pictures?
  • How do the other two topics you subscribe to help the robot achieve the task?
• How does the robot perform at the task?
  • Discuss the design and results of your experiment to assess it.