Josh Kline's Robotics Project 1

[ Back to Index. ]

This page describes my project for Harvey Mudd's Robotics (CS154) course. I will be building and programming a robot from the Lego Mindstorms kit.

Problem Statement

Create a robot using Lego Mindstorms and program it to traverse a room while avoiding obstacles. The room will be set up like an obstacle course and the robot's objective will be to make it from one side of the room to the other.

This project is interesting to me because it will give me a chance to explore the Lego platform while delving into a broader robotics topic, that of navigation.

Bonus

While completing this project I will also be designing a curriculum to present this problem to a group of students at Pomona High School. This project is in conjunction with Seminar in Science Education (IE197).

Update [2003-11-16 but known for some time before that data]: It is unclear whether this portion of the project will continue. Implementation of a curriculum is dependent upon agreement from my partners in IE197. Personally I am confident that a two or three hour activity could easily be organized using the out of the box LEGO programming environment and a pre-built platform. There are many details to be decided, including how much instruction to give the students and how much to let them try to decipher on their own, e.g. do I tell them how to make the robot turn, or is that part of the challenge for them to figure out. Many possible competition scenarios could be devised for example fastest time through a maze, ability to complete more and more difficult mazes.

Progress

1. Installed The Robotics Invention System (RIS) the software that comes with the Mindstorms kit.
2. Constructed "Hank the Bumpertank" from the The Unofficial Guide to LEGO MINDSTORMS Robots.

   Front/Side View 70.86 Kb

   Top View 49.29 Kb

   Both Motors 59.19 Kb

   One Motor 53.03 Kb

   Touch Sensors/Bumpers 54.93 Kb

3. Programmed Hank using the program listed in the book.
4. Tweaked program to give more desirable results. Some success.
5. 2003-02-10: Spent about 4 hours trying to get the brickOS installed. First cygwin needed to be installed and I think that was done successfully. There seems to be some trouble with the Hitachi gcc compiler. I tried following some instructions i found on the web but on the last "make install" there was an error that I don't understand. I e-mailed to Joe Mallone to see if he could provide any assistance.
6. Unknown date. Unfortunately Joe Mallone was not able to get the brickOS functioning on my current box. With the help of Matt Livianu I narrowed down the problem, or I should say the symptom, to a link error having to do with memcpy.c. Not being knowledgeable about the workings of Linux I am unable to progress further on this issue.
7. 2003-02-16 After several hours of tinkering (trial and error and following directions of myriad web pages) I got the brickOS installed. The Brown University cs148 course web page was helpfull in this process. I was able to download the firmware onto my RCX block but was unable to download any programs. The dll program seems to just hang. Hopefully I will be find some solution to this problem as well.
8. 2003-02-17: Cosmetic Makeover. I have named my current robot "Perspective" in anticipation of building a second robot for use with BrickOS. Today I transformed Perspective into a mean fighting machine complete with lasers and a rotating, tilting turret. Here are some pictures of the preliminary armored stage.

   77.42 Kb

   73.49 Kb

   The current state is much fiercer having an armored turret and additional degrees of freedom in the guns but quality photos have not been processed yet. Stay tuned...

Algorithms: What's in a brain?

• In its current form the program running on Hank is quite simplistic. It moves forward until it detects a bump with either the right or left bumper. It them backs up and turns in the opposite direction of the hit, e.g. if left bumper hit then make a right turn. The distance that Hank backs up and turns can be changed modifying the program. It took some experimentation to get an optimal combination. I found that a short backup distance and a small degree of turning works better than larger values of either. This means that Hank will usually hit the same obstacle several times before turning enough to get around it. It also lessens the chances that hank will turn farther than is needed. Another drawback is that the robot will get stuck in a corners, repeatedly hitting one bumper and then the other and thus turning into the corner each time. It also has a tendency to follow the same general path over an over again.
• The currently algorithm combines the "wander" and "avoid obstacles" low level behaviors described in the article by Rodney Brooks It still remains to create a planned way to travel that would be similar to the "explore" or "build maps" levels of Brooks' design. The sample BrickOS program described in The Unofficial Guide to LEGO MINDSTORMS Robots uses Brooks' subsumption architecture. If I ever get brickOS to work then I can implement this.
• The current program certainly does not use the sense, plan, act model described in the paper by Erann Gat. With no planning to speak of Hank simply senses and then acts.
• Thanks to Prof Dodds for suggesting these citations.

Worlds: What is an obstacle?

• Currently Hank likes to hang out in arenas constructed out of bricks or books. These arenas typically are enclosed on three sides have an open side, or partially open side, through which Hank can enter or leave. The middle of the arenas can be littered with other barriers.
• A major constraint is that the barriers must be high enough to trigger Hank's bumpers and not slip under them.
• Hank is strong enough to push light obstacles although typically only for about a centimeter before the touch sensors are triggered.
• The objective is for Hank to start somewhere in the arena and be able to exit through the opening.
• Here are some photos of a sample arena.

  Sample Obstacle Course 78.58 Kb

  Sample Obstacle Course 38.93 Kb

Near term goals, aka the next step

• Install legOS or other software that is more powerful then the standard RIS. This goal is revised to install brickOS which is the new name and hence new versions of the legOS.
  2003-02-16: This goal is modified to reflect that the brickOS is installed, but is not functioning fully yet.
• More playing, er, I mean testing.

Long term goals

• Implement an algorithm to find a path through an obstacle course. I don't know what form this will take, I make try to invent an algorithm, or use one that already exists. This goal is still vague, e.g. what is an obstacle course?
• 2003-02-16: Once brickOS is functional device an algorithm that will do more than simply wander. It should remember which direction it wants to go and then keep track of how far it has traveled and turned. This may involve adding some rotation sensors to Hank.
• Determine whether Hank moves in such as way as to be interesting to the cat.

Blatant Paper Citation

In their 2002 paper "Get Back in Shape!" Yoshida et al. describe a type of robot that can reconfigure itself. This robot is made up of small identical pieces, each a robot on its own, that use shape memory alloy to attach to each other and rotate around each other. While LEGOs are reconfigurable to a high degree they are not yet able to reconfigure themselves. (Fortunately they have managed to train humans to do this for them.) By Prof Dodds' axiom that "anything that can be done can be done in LEGO" it is probable that there will eventually be LEGO robots that can assemble or rearrange themselves or each other LEGO creations.