Harvey Mudd College
Computer Science 154 - Robotics
Assignment I

The Cleanup Challenge

Introduction

The algorithms that are often the first to come to mind for mobile platforms are navigation, localizing, and mapping algorithms. However, there is a subfield of mobile-robot algorithms that are slightly different from these: coverage algorithms. The goal of a coverage algorithm is to seek out and process the entire free space of a mobile robot. Process might mean to simply drive over, as with lawn-mowing and floor-cleaning systems, or it might involve additional work, e.g., deploying sensors to detect chemicals in an environment, demining an area, or even applying a coat of paint to the surface of an object with a robot arm. Howie Choset of CMU maintains a line of research on robotic coverage at http://voronoi.sbp.ri.cmu.edu/coverage/coverage.html.

The task and environment

The challenge for this project is to design and implement a small robot that will achieve coverage (as completely as possible) of a desk-like environment. This desk-cleaning system will operate by using the fundamental principle of effective managers: it will delegate most of the actual clean-up work to another robot (or human). That is, it will move around a desk or table in an effort to push all of the lightweight objects onto the floor. There may be heavier objects (e.g., computers, large books) that should be condidered obstacles. And, of course, the robot must be sure not to fall off the edge... . One possible environment is shown below. The blue regions indicate obstacles on the desk's surface; the green band represents a white edge that will be placed around the desk's outer edge in order to help the light sensors detect it. The black regions indicate light items that are cluttering up the desk: the system should try to clear these off as completely as possible.

The hardware

The Mindstorms' RCX brick and the Handyboard are both capable of controlling a robot to handle cleaning up a desk. Both operate on top of a lego chassis that you will need to design. Ample documentation for each is available in the lab.

Warnings These are general cautions that you should keep in mind throughout this lab project.

• The robot should not push or ram any wall/obstacle continuously, i.e., the intent is not to move heavy obstacles around. If a robot seems to be pushing against a wall, it will be reset.
  
  
• In general, keep your robot safe! Each group is responsible for their controller, lego chassis, and sensors (all told, about $300-500).
  
  
• To this end, any runs on a real table must be closely supervised so that the robot never comes close to falling off the edge! Development should be done in an environment without an edge, e.g., on an area of the floor demarked with white tape or paper.

In addition, each team will have at least

• Three lego motors
• Two light sensors
• One infrared sensor
• Three touch sensors

The first thing to do is to read the technical reference for the system you will be working with . This is a must as it explains everything you will need to know about the board. The handyboard reference is available here in PDF format. In particular, be sure you understand how to keep the board charged. You should leave your board in trickle-charge mode when not using it. For safety, I would advise avoiding "zap" charge -- forgetting that a board is "zap" charging can fry it.

From there, the project is wide-open. Some things to consider (and references):

• There are a couple of books in the Research Lab that have some tips and advice on lego construction. In particular, lego is not as strong or stable as you might guess, and you will need to reinforce your chassis to hold the Handyboard securely. Here is a link to a PDF file named The Art of Lego Design.
  
  
• You will need to pay attention to the drive train of your robot. The lego motors are high speed and low torque, so you may want to gear them down.
  
  
• Early on (even before a chassis is built) you should try out your sensors and motors via Interactive C. That way, code development can progress in parallel.
  
  
• You will need to detect obstacles using touch sensors; it is unlikely that IR sensors will be robust enough for this. (Bumping objects is OK, just not pushing them.) You can extend the reach of a touch sensor by resting a lever against it.