Progress for week 44 (2014)

From Robin

(Difference between revisions)
Jump to: navigation, search
(Else-Line)
(Else-Line)
Line 28: Line 28:
* Implemented Lamarckian learning
* Implemented Lamarckian learning
* Done some testing
* Done some testing
-
* Fixed fitness issue with learning scenario
+
* Fixed fitness issue in learning scenario
 +
* Fixed licence issue with Visual Studio on laptop
== Stian ==
== Stian ==

Revision as of 10:00, 5 November 2014

Contents

Student template (copy this for your entry)

Budget

  • Todo 1
  • Todo 2

Accounting

  • Done 1
  • Done 2

Eivind

Budget

  • Write communication protocol
  • Create the game area
  • Improve the testing program

Accounting

  • Implemented threaded cross-program communication protocol based on Zeromq.
  • Aquired and prepared all the materials for the game area.
  • General program improvements, progress and research

Else-Line

Budget

  • Implement Lamarckian learning
  • Continue testing parameters
  • Look into good statistical methods for comparing results

Accounting

  • Implemented Lamarckian learning
  • Done some testing
  • Fixed fitness issue in learning scenario
  • Fixed licence issue with Visual Studio on laptop

Stian

Budget

  • Find a workaround for crashes caused by bugs in the tested algorithms
    • Try to find a safer range of parameters or run the algorithms in a seperate process.
  • Start work on including the rSGM algorithm in the framework.

Accounting

  • Added more constraints to avoid some troublesome parameter combinations which lead to occasional crashes.
  • Fixed a bug in the ELAS algorithm.
  • Implemented threaded evaluation using Intel Threading Building Blocks.
  • Looked into further algorithms, but ran into library dependency problems.
  • Read an NSGA-II paper.

Snorre

Budget

  • Finish implementing simple GA
  • Experiment with different population sizes and number of iterations in the GA
  • Experiment and observe behaviour of the robot with the GA using an uneven terrain
  • Make a plan for how to implement the NSGA-II algorithm

Accounting

  • Implemented a simple GA and experimented a bit with differen pop. sizes and no. of iterations
  • Tested out the robot on different terrains, made a suitable one for the task (also using the GA)
  • Started planning and implementing the NSGA-II algorithm

André

Budget

  • Try to find more information of where to by the omnidirectional camera
  • Try to implement a Feature detection algorithm with openCV
  • Borrow a Kinect an set the kinect up ready for testing
    • Use the kinect first as distance measure
    • Look into more usages of the kinect

Rune

Budget

  • Improve the pathfollower
  • Extend with voronoi diagram

Accounting

  • Improved the pathfollower!
  • Switched to probabilistic road map, not yet finished.
Personal tools
Front page