Currently the mission-design in ROCK is done in several scripts tied together by Roby and including a lot of configuration and logic.

This is all done in Ruby and allows you to use every Ruby-scripting you want to use. Soon this is very hard to maintain and update, especially if you are testing in real-life-environments or participating to competitions. If there would be an easy (best case graphical) way to visualize and even manipulate the high-level behaviour of the robot it will result in easier and more efficient testing.

Normally algorithms are developed with a lot of testing to improve it. As a result there are test scripts and hopefully even unit tests for most of the library-methods and individual tasks. The complex high-level-behaviour is mostly untested, but offers a lot of pitfalls even more critical to the mission than a single failing task or algorithm. Maybe the most annoying error is a typo in mission-design, because this will result in a failing mission and it would be very easy to fix.

Besides the straight forward visualization of missions (reducing the complexity) it would help a lot to have a way to test the missions without being dependent to hardware, a simulation-environment and even correctly working algorithms. There should be a way to test only the mission-design. A nice side effect of this would be a syntax checking (and even basic type checking) of mission-design before running the mission, because it is executed before and tries to cover all possible paths the mission can take.