hawkw/cs383s2015-lab5

Lab 5 for CS383 at Allegheny College

cs383s2015-lab5

Lab 5 for CS383 at Allegheny College

How to Use:

1. ./sbt assembly will build the jarfile.

• You do not need sbt or Scala installed, the shell script will take care of that
• You DO need to export the variable $EV3_HOME in your runtime configuration (.zshrc, .bashrc and so forth). This should be the path to wherever LeJOS EV3 is installed on your system.
• Because of issues related to managing multi-project SBT builds, it is necessary to edit project/Build.scala if you want to change whether you are building the evasion program (RandomNav.scala) or the tracking program (Search.scala).

2. Connect the brick and run deploy.sh to send the jar file to the robot.

• When it asks you for a password just press return.
• Once again, the file name of the jar is hardcoded in the shell script and will need to be changed if you want to deploy the other program. Eventually this will be fixed.

Implementation Details:

The tracking and evasion strategies we implemented are described in the following sections.

Evasion Strategy:

We kept this strategy as simple as possible. Our program RandomNav uses an infinite stream of random degree and distance values to command the LeJOS DifferentialPilot object to move the robot in a random walk. That is, the robot rotates for a random degree value, then travels for a random distance. It does this until any of the robot's buttons are pressed, at which point it exits the main function.

Tracking Strategy:

Obviously, tracking is the more sophisticated task of the two. To accomplish the difficult task of tracking a moving target using only one distance sensor, we devised a way of extrapolating a future position of the target using the currently visible location of the target and its last seen location. This meant that our robot can turn to face this future position and thus track the target with an acceptable degree of reliability.

Note that we assume our target robot uses a constant speed for its evasion technique. Therefore, the only events that will cause our robot will lose track of its target are the target robot changing its heading or stopping. Whenever our robot loses track of its target, it enters a "panic mode", in which it attempts to relocate the target by rotating in progressively widening sweeps from left to right and right to left. Once it has relocated the target, it resumes its method of predicting the target's future positions.

Project Takeaways

The most challenging portion of this project was almost certainly the implementation of the tracking strategy. The limitation of only one ultrasonic sensor presented a significant obstacle that required more complex tracking techniques to overcome. These techniques employed trigonometric functions that allowed us to calculate the trajectory of the target using limited information.

Other challenges we faced were related to the use of Scala to write our programs. We chose to program in Scala rather than Java, as we preferred its' syntax and the expressivity benefits of functional programming. Since Scala programs run on the Java Virtual Machine, an EV3 robot running LeJOS is able to execute them. While we found that programming in Scala rather than Java had many benefits, we had issues with the use of tools other than the LeJOS Eclipse plugin for managing our programs. It was very difficult to configure the SBT build tool to correctly build jar files suitable for deployment on the EV3 robot, and to configure SBT to generate two separate jar files for our two programs. Despite the divergence from our previous use of Java and Eclipse to write and build our programs, we were able to successfully write, build, and upload Scala programs that ran on our Mindstorms robot.

cs383s2015-lab5's Issues

Robot Evasion

• Random heading stream
• Random distance stream
• Use values from streams as control for robot's motion.

Can We Program The Robot In A Good Language?

I would love to not have to use Java.