Back to Homepage
Week 1 : June 1 - 4, 2011
This week, my first reading was a motion planning paper from Jean-Claude Latombe. It
was a classic paper introducing important key terms in motion planning such
as PSPACE-hard and nonholonomic planning. The beauty in this paper is that
although it does mention robots it stresses that non-robotic applications
are growing and that these things (i.e. medical surgery) will shape the
future of motion planning research.
My other reading assignments Flocks, herds, and schools: A distributed
behavioral model, Chapter 3: Configuration Space, and Chapter 7: Sampling
Based Algorithms have either opened my mind to more applications of motion
planning or have explained a certain aspect of motion planning such as
the differences in C-space from workspace or obstacle from free space.
I have also had the chance to experiment with a few of the simulators Aditya
provided and learned what some of the less intuitive parameter titles entail.
For the next week, I am hoping to see more of the code that makes using the
simulators so easy to use and learn more about how these algorithms work
in terms of theory and code implementation.
Week 2 : June 5 - 11, 2011
This week I was introduced to more in depth explanations on motion planning
techniques that I was introduced to in week one such as OBPRM. I achieved
this by reading: Better Group Behaviors in Complex Environments with Global
Roadmaps, Reactivity and Deliberation: A Survey on Multi-Robot Systems, and
OBPRM: An Obstacle-Based PRM for 3D Workspaces.
Also, this week I was able to polish a little more of my C++ knowledge and
learn more about how the group behavior simulations work by observing some of
the code base and drawing conclusions.
Overall, I am enjoying the discussions talking about the theory behind motion
planning ideas and seeing former ideas implemented in simulations as some sort
of visual recap of their research.
The only problem that I am encountering is that there are so many files in the
code base it makes 10 weeks seem short, but I will keep learning and understanding
till I achieve my project is complete.
Week 3 : June 12 - 18, 2011
This past week, I watched the two PhD defenses. Experiences like these are
important to my future since one day I want to be the person defending my research.
Later that night, we celebrated their achievements over dinner.
Also during this week, I was introduced to the pmpl code and was given a project
involving the Group Behaviors code. The extent of my project involves writing a
class to identify objects in proximity specifically nodes.
Week 4 : June 19 - 25, 2011
This week Aditya, our graduate mentor, was feeling ill. This was also the first
week that Osa and I were given programming projects with due dates. Although
Aditya would not be there to guide us in presence, he offered his aid through email.
At the same time, Dr. Amato was busy with paper reviews. Jory Denny was there to
answer any immediate questions that Osa and I had. The support at Texas A&M help
make this week a successful week. It is always tough getting back into the pro-
gramming spirit. Good thing I was not alone.
In terms of deadlines, I made a class that detects which nodes are in proximity and
discards that have an obstacle between the agent and the nodes that were in proximity
of the agent.
In other news, I visited Texas A&M's gym and played basketball for the first time
this summer. I was also able to see the new X-Men movie while chatting with my
fellow REU friends. This has been a great summer thus far.
Week 5 : June 26 - July 2, 2011
This week Sam Jacobs presented his prelim work for Ph.D. I was able to catch some
sneak peeks while watching his practice runs.
In terms of lab work, I tool my Roadmaps class and added distance, direction, and
CNodeData all in pairs in a vector. The next step is to add error to my simulation.
I have already submitted some ideas that will be posted on my website of how I
plan to use some provided resources to calculate error. Lastly, I had a meeting
with Dr. Amato and Jory where Osa and I detailed what we have done. Dr. Amato
then said that we should look into things that like determining position based
on triangulation methods to take steps in the localization direction where the
agent does not know where it is in the environment.
Week 6 : July 3 - 9, 2011
This week, I implemented a localization method. It involves taking two points and
using their distance with relation to the reference (the robot) and their position
in the x-y plane to solve for where the reference is in the x-y plane. This can be
achieved by looking at it from an algebraic standpoint. If there is an unknown x
and an unknown y, then two equations are needed to solve for the two unknowns. Solve
for one and then use this answer to solve for the other one.
Also this week, I had the chance to hear Jory give a presentation about one of the
papers that was accepted to ICRA. It was about toggle PRM. The big difference in
toggle PRM is that it uses the obstacle space to its advantage instead of discarding
a sample in obstacle space.
That is all for this week. It was a shorter one due to the Fourth of July holiday.
I am still enjoying my time at Texas A&M.
Week 7 : July 10 - 16, 2011
This week was not as eventful as the others. We found an error in our current im-
plementation of the localization method. Since the two closest nodes in the simulation
to the given agent are passed in by distance, the localized position sometimes was
wrong. This is because of the seemingly hard-coded nature of my approach. My approach
either assumed that the closest node was on the right or the left. I tested this by
comparing the triangulated position against the real position. To solve my problem,
I will have to make my approach more general. I have not yet solved the problem, but
I am aiming to have my solution by Monday.
Outside of the lab, Dr. Shell spoke at this week's brownbag. It seemed as though his
lab focuses on cooperative robotics, but he did mention that motion planning techniques
were involved in his research as part of his simulations. That concludes my week 7
journal.
Week 8 : June 17 - 23, 2011
This week we finished the implementation of the triangulation method. Now we have
started brainstorming and first draft writing of our joint final paper. The outline
for our final paper :
Abstract
Introduction
Related work and motivation
-Motion planning crash course
1. Motion planning
2. Multiagent systems
3. Group Behaviors
4. Localization
Existing framework
1. Localization class
2. Vision base class
3. isCollision
Our approach
1. Agent Vs. Agents
2. Agent Vs. Nodes
Experiment and results
Conclusion
This can be edited after discussing more in depth on where we want to see our paper.
Week 9 : July 24 - 30, 2011
This week, I was busy preparing my final paper and poster.
This past Friday, I had the chance to practice my poster presentation in front
of many of the peers from the Parasol lab. I enjoyed the experience and received
some good tips on how to improve my poster and my presentation. On my poster,
I include a similar structure to my paper. I present my abstract, some related
work, existing framework, and present my experiments and results.
Also, there was not a brown bag this Thursday but I still was able to submit my
abstract to Theresa. Overall, this was a good week and preparation for next week
is on schedule.
Week 10 : July 31 – August 6, 2011
During my last week at Parasol, I finished my poster and my final paper.
This was definitely the most even filled week since there were a few more
changes and additions that were necessary for good results that would go into
our paper.
One of the biggest changes that we faced during our last week occurred when
noticed that one of our bugs created huge errors that made our current results
seem promising, but they were completely wrong. When we fixed this error,
the results were not as good, but they were the correct results. Our rightvs.-
left decision method did not account for all cases resulting in the graph of
our data to be sub-linear when the correct results were linear.
On the official poster presentation day, during session A I was a peer judge
judging the other REU’s posters and in session B I presented my results.
Learning a little about many different topics made for a fulfilling day regard-
less of the Texas heat.
As I say goodbye to Parasol, I have to thank Dr. Amato, Aditya, Jory, and
all the others from the lab that took me into the lab and gave me the chance
to learn all that I could in ten weeks. It was a fun time and an amazing
experience.
Back to Homepage