|
CS 371 - Introduction to Artificial Intelligence
Homework #9 |
Due: Last day of class
Note: Work is to be done in pairs.
Mobile Robotics Monte Carlo Localization Challenge
In the summer of 2016, Zuozhi (George) Yang '17 further developed a Neato Vacuum
Robot Monte Carlo Localization project by Paul
Ruvalo. Our CS 371 Mobile Robotics Monte Carlo Localization Challenge
is to further improve upon the ability of our robot to:
- successfully localize with minimal prior information
(ideally just the map of the environment),
- maintain correct localization robustly in the face of
frequent, complicated maneuvers, and
- ultimately be able to relocalize correctly when it is
suddenly repositioned (i.e. address the Kindapped Robot Problem).
In addition to making improvements, it is our goal to document
our improvements, data, experimental methods, observations, reasoning, and
conclusions with such thoroughness and clarity as to empower future generations
of students to extend improvements beyond our best performance.
To this end, each group will (1) select at least one area of focus, e.g.
- Monte Carlo Localization algorithmic extensions (e.g. AMCL) and/or
parameter tuning (e.g. number of particles, number of sensors used),
- Improvement of our motion model through data collection, parameter
fitting, maximum likelihood estimation, etc.,
- Improvement of our sensor model through data collection, parameter
fitting, maximum likelihood estimation, etc., or
- Any other area of focus I approve of where you are able to propose a
reasonable empirical study that has a good possibility of yielding
improvements within the hardware, software, and algorithmic constraints of
the assignment.
Details, code, and supporting documentation have been supplied by George and
will appear here in the future.
Deliverables:
- (75%) A zip file containing web-browsable documentation of:
- (5%) Your area of focus,
- (10%) Your hypothesis of what improvements might be feasible within
this area of focus,
- (10%) Description of the measure (i.e. metric) you used to evaluate
the performance/success of your approach,
- (10%) Data collected along with a description of how the data should
be interpreted,
- (10%) Any algorithms you developed or modified (including parameter
changes) in pursuit of your improvements,
- (10%) A summary of your conclusions from your empirical research,
and
- (20%) Anything a future student would need to know to (1)
repeat your experiments, and (2) apply their lessons and achieve the
same peformance improvements. If all of this is provided
in previous categories above, this 20% will be awarded automatically.
The test is whether or not I am provided with enough documentation to be
able to fully replicate your results.
(Note: Any
http links beyond your pages should be absolute (i.e. starting with
"http://"), whereas multple pages (if desired) should be linked to
relatively (e.g. starting with "./pageInSameDirectory.html"."
- (25%) An informal presentation summarizing your findings, demonstrating
improvements (if any), and noting the most significant thing you learned and
would pass along to another team performing the same work.
Resources: