Project: Monte Carlo Localization

CS 371 - Introduction to Artificial Intelligence
4th Hour Project: Monte Carlo Localization

Due:

Office hours or scheduled meeting during week 11: Demonstration of improvements over previous group accomplishments and plans for further improvements.

During last class 28: Class demonstration of work.

All archivable documentation due by the end of the final exam.

Note: Work is to be done in pre-assigned groups.

Mobile Robotics Monte Carlo Localization Challenge

Using the Anki Cozmo robotics platform, implement or improve upon an implementation of Monte Carlo Localization with an attempt to solve the Kidnapped Robot Problem. Demonstrate and document your work such that future generations of students could duplicate your results and improve upon them.

Tips:

Set simple goals. Follow the KISS Principle. (You can always set more ambitious goals if you achieve these early.) Example:
- 1D state space:
  - Have a fixed robot that can rotate a camera or other localizing sensor to determine state θ.
  - When you start the system, let initial state θ₀ be the "home state".
  - Have the system rotate and sense to build a mapping from sensor inputs to probable locations.
  - After it has terminated mapping, put it in a new mode seeking to return to home state θ₀when it does not appear to be home.
  - "Kidnap" it by rotating the robot and demonstrate that it can relocalize and return home.
Divide labor: project lead, documentation, version control, sensor model, motor model, etc.
Plan team meeting times in advance. Budget for approximately 28 total hours for each person including team meetings. Log hours.

Deliverables:

(30%) Physical demonstration of project
(30%) Emailed log of each team member's hours along with summary time totals for each team member. Format for each log line:
- Name, date, time range, brief one-line description of activity
(40%) Accessible web documentation of project, including goal(s), necessary instructions to recreate the project (e.g. assembly instructions, software prerequisites and installation instructions, code developed, use instructions, etc.), and assessment of functionality (i.e. how well did it work). NOTE: This should be created within a single folder with relative links such that it can be easily archived within my web space and still work.

Anki Cozmo Localization

Programming tools:
- Cozmo SDK:
- cozmo-tools: https://github.com/touretzkyds/cozmo-tools
Possible projects:
- Create a new project where the robot is restricted to rotational movement only and uses visual camera sensing to localize.
- Find and build upon prior Cozmo 2D localization and mapping work you might find.
Java demonstration of MCL framework for simple 1D MCL problem.
MCL Slides
Recommended text: "Probabilistic Robotics" by S. Thrun, W. Burgard, and D. Fox

Mobile Robotics Monte Carlo Localization Challenge

Anki Cozmo Localization

MCL Past Project Documentation