1 / 31

DARPA introduction

DARPA introduction. Joost Westra. Introduction. Darpa & Tekkotsu Project work Vision Color Calibration Tool Detection World state Thread Practical. Why Darpa?. Dutch ARchitecture Project for Aibos Problems with the old (German) code Creating your own code

ethan-chase
Download Presentation

DARPA introduction

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. DARPA introduction Joost Westra

  2. Introduction • Darpa & Tekkotsu • Project work • Vision • Color Calibration Tool • Detection • World state • Thread • Practical

  3. Why Darpa? • Dutch ARchitecture Project for Aibos • Problems with the old (German) code • Creating your own code • Preventing the same problems

  4. Why Tekkotsu? • Great support • Great documentation • Supplied with JAVA user interface • Solutions for low level control • Event driven • Handling concurrency in good way • Not to specific

  5. Project work • People work short time (high turnover) • Long term project • RoboCup & other research • Short startup time

  6. Requirements • Strong modular approach • Future research (extendibility) • Easy to understand (clarity)

  7. No waterfall approach • requirements changes frequently on a large project • implementation step hindered by previous steps • All testing and debugging most be done at the end

  8. Iterative approaches • Rapid prototyping • Identify the initial user requirements. • Develop a prototype. • Use and evaluate the prototype. • Revise the prototype. • Start again with point 3.

  9. Iterative approaches • Extreme programming • Short releases • Keep people happy • Every member contributor to the project • Extensive testing • Pair programming • Refactoring • Milestones

  10. Coding style • Uniform • Tab is 4 spaces wide • Shorter and clear implementation int updateOutputs() { outputs = 1; }

  11. Guard against multiple inclusions // IncludeGuard.h #ifndef INCLUDEGUARD_H #define INCLUDEGUARD_H // Body of header file here... #endif // INCLUDEGUARD_H

  12. Naming • Java naming conventions • class: class FrenchVanilla : public IceCream { • object identifier: FrenchVanilla myIceCreamCone(3); • Function: void eatIceCreamCone();

  13. Documentation • Inline code documentation • Review reports • Small • Choices made during implementation • deviated from specifications • Other remarks • In the code • Theoretical reports • worthy for publication

  14. Testing • Make tests before programming • Test at least daily • Do not commit untested code • At least make it compiling! • Test code on the AIBO • Unit testing

  15. Versioning • Different branches for each team • Always at least one stable version • Maybe different branches for tasks • Don’t need maze solver for soccer • Even better if module could be used and is not hindering

  16. SVN over CVS • Directory versioning • True version history • Renaming and copying are supported • User friendly • TortoiseSVN in windows

  17. Editor • Eclipse • Use your own favorite editor • Do not litter the SVN with project files or binary files • Compiling is done in command line, no special configuration settings needed

  18. Vision explained • More practical approach • Recommended by Tekkotsu site • How to extend for object recognition • Processing on: raw/segmented/jpeg • Vision pipeline uses lazy evaluation

  19. Color Calibration Tool • Default tool very limited • Mistake -> start all over • No reloading off image classification • Uses only U and V channels • Colors change if angle changes • Overlap in different colors

  20. Color Calibration Tool • Ported from software project • Uses multiple images • Uses Y,U,V Channels • Image are classified using polygons • Mistake -> delete mistake • Reloading image classification • Possible to add more images • Automated overlap handling

  21. Detection Generator • Behavior can listen to CMVisionObject • Better to create a filter • Filter listens to CMVisionObject • Behavior can subscribe to filter • GoalDetectionGenerator

  22. Detection Generator • Prevent duplication of code (modular) • Easy to make new behavior • Step-by-step guide

  23. World State • Shared memory region • Direct access: state->yourVariableName • Event throwing • Behaviors reacting to changed world state • Step-by-step guide • Creating and throwing new events • Extending the world state • Listening to events • Demo behavior

  24. World state • No dynamic memory allocation • No vectors ect. • Can be a problem (3APL for example) • No pointers • If no dynamic memory, not a big problem • Only accessible from default threads • Communication possible true main thread

  25. Adding an extra Thread • Standard: Main, Motion, Sound • Higher level (slower) reasoning • Adding the thread not very difficult • The communication is (was)

  26. Adding an extra Thread • Shared Queue • Post Tekkotsu event • Other special behavior places the message in queue • Step-by-Step guide • Simply starting new behavior

  27. Practical • Read documentation! • Use matching ip-adress • Robot safety • Recommend using windows with cygwin • Problems with memory card writers • Installation Guide

  28. Cygwin • Cd /cygdrive/c/ • Ls • Tab completion • rsync package needed

  29. Basic commands and files • Project\make newstick • Project\make update • ftpupdate • Tools\mon\ControllerGUI [ip-adress] • Wlan.txt • Cygwin.bat

  30. Questions?

  31. To the lab! • We are in the middle of moving (sorry)

More Related