Goals

Vision:

• Get Intel code compiling on Ubuntu and into a Mercurial tree. (20 hours)
• Hook up .Net wrapper to Intel vision code. (10 hours)
• Create / find a 2-frame GUI tool which takes a picture and applies various Intel code to it and displays an output image (40 hours)
• Write some code to generate simple analysis of 3-D location other objects. (100 hours)
• Hook output of 3-D world to input of vision recognition (10 hours)
• Provide an API to return analysis to driving engine. (40 hours)

Driving:

• Get Torcs compiling on Ubuntu and in a Mercurial tree (10 hours)
• Port Torcs robot written in Delphi to C# (100 hours) (https://www.turnsharp.com/home.aspx)
• Build ways for us to run tests in one click of a listbox or command-line. (20 hours)
• Port a few maps and scenarios into our test suite. (20 hours)
• Provide an API to create images for vision code (20 hours)
• Call vision code at appropriate times (20 hours)

Debugging features:

• hook the output of the vision recognition into the input of the 3-D engine, by creating ghost images.)
• Ability to log all changes to the world including results of vision analysis, so that we can replay to figure out what happened.

Later Goals

• Add other features like waypoints, stopsigns. (Someone else doing this we should hook up with [1])
• Evaluate better & more professional-looking free 3-D and physics engines which already have C# wrappers, etc.