Wishlist
- Agent.GPS
- Out of Memory in CameraData line 146. (Seems to be solved in rev 1927)
- Area calculation and other optimalizations (Partly done (area removed from ManifoldImage:RenderPatch) by merge of Agent\Map with 2009\competition).
- No mobility.tif exist in the file system warning (solved in rev 1930).
- GetGeo info's (already reduced in rev 1927).
- 2009 branch (made)
- Missing Matrix functionality. Chapter 4 of MatLab has a nice overview. Matrix2 has Transpose, Determinant and Invert, Matrix3 not yet. Other basic functions are Identity, Norm, PseodoInverse, EigenValues and Single Value Decomposition.
- The Matrix class contains several functions which are not overloaded yet. Function which are available are (Identity, Zeros, Ones), (Negative, Trace, Equals).
- Improved Matrix-interface. Missing Operation is '\', missing function Minor(row,col).
.
Started Labbook 2012.
12 December 2011
- Succesfully build 2010/assistance, with new ThirdParty/GDAL-1.8. Still some complaint about 64bits reference. Did clean. Build fils on cd ogr/ogr_frmts(should be cd ogr/ogrfs_frmts. Error was in top makefile.vc.
Gdal fails now in frmts;nmake plugindirs.
- UsarClient had old Post-Build event. Should move to Post-Build of gdal. Solution build tries an plugins-install (fails on copying ./data to GDAL_HOME, defined as C:\warmerda\bld in nmake.opt).
8 December 2011
7 December 2011
5 December 2011
- The wayback machine has a snapshot of the warmup.html page of Sebastian Thrun.
- Continued with ogrsf_frmts. All subdirectories compile and lib is called without explicit call for an machine (/MACHINE:(X64|X86). Did a nmake clean. Compilation fails in sudir pds. Adding the directories and frmts/pcidsk solved this issue. Build fails now on directory gcore.
- Did a nmake /f makefile.vc in directory swig\csharp, because that is what I need. Compilation fails on existance of ..\..\gdal_i.lib.
- Tried to build parts of nmake /f makefile.vc target-lib: nmake /f makefile.vc dll; tries to make gdal.lib (fails on vb6-directory).
- vb6 is easily made, but nmake /f makefile.vc dll fails on existance of x64-files. Added some language directories to swig to allow nmake /f makefile.vc clean to be executed. nmake /f makefile.vc dll first needs ogr\ogrsf_frmts, followed by ogr\ogr.lib; vb6. Adding gcore\Version.res gives 258 unresolved externals. Gcore needed frmts/gtiff. frmts/gtiff needs frmts/jpeg. Both directories solved many unresolved issues (118 left). Added aigrid and ceos (203 left). Added elas and hfa (64bits file). Did a nmake clean again. Added source-code of alg. No complains about 64bits files (103 left). Added raw and gfx (79 left). Added ceos2 and png (77 left). Added dted, mem, jdem, gif (72 left). Added envisat, aaigrid, usgsdem, l1b (68 left). Added fit, vrt, xpm, bmp (64 left). Added rmf, nitf, airsar, rs2, ilwis, msgn, rik, pcraster (52 left). Added leveller, sgi, srtmhgt, idrisi, jaxapalsar, ers, ingr, dimap, gff, terragen, gsg, cosar, adrg (29 left). Added coasp, tsx, blx, til, r, nortwood, saga, xyz, hf2, kmlsuperoverlay (18 left). Added ozi, bsb, zlib (1 left). Adding grib solved the issue. gdal_i.lib exists after executing nmake /f makefile.vc dll.
- Went to swig\csharp. Nmake fails inside subdir ogr. Added code. Now also this command is succesfully executed (6 dll created which are needed inside UsarCommander).
1 December 2011
- Looked if I could reproduce the sonar map from Radish:
Found documentation of the ActivMedia Pioneer 1. Committed changes to 2010/assistance (rev. 2251).
- ThirdParty\GDAL-1.8 was missing (only available at nb-unreal for assistence64bits. Copied frmts\iso8211 from C:\src\gdal-dev-x64\vs2005.
- As suggested in my RoboResc labbook, I changed the nmake command to target-lib. Compiling the port worked, but frmts\iso8211 fails on XCOPY. Commented XCOPY out. Another could command could be ogr_dir. Stil, frmts\sdts seems to be necessary.
- Finished frmts, build complains now about ogr-directory. Strange enough, Visual studio complains that the target is X64 on a 32bits machine. This is not in the nmake.opt, should look in the configuration manager of the solution.
30 November 2011
- Read again chapter 4 of Introduction to Autonomous Mobile Robots, this time the 2nd edition. Image fingerprinting for localization is still described for historical reasons, the modern approach is a bag of features. Also new is a comparison of 6 line extraction techniques.
- Found an early draft (2000) of Probabilistic Robotics.
- Searching for UvA-global-localization.
- Looked at simulated sonar dataset from Radish.
- Converting assistance solution (VC2005) with VC2010 resulted in problems with Iridium, GDAL, Neodym, SharpZip lib.
29 November 2011
- Tried Fox's Aibo localization project in cygwin-octave. Octave complained that the X11 DISPLAY environment variable was not set. Installed cygwin's X11 package.
- Could start Xserver with Xinit. No complains anymore about DISPLAY, yet still 'unknown line property black' in /usr/share/octave/3.4.2/m/plot/line.m.
- Replaced references from 'black' and 'white' to 'k' and 'w'. Still problems with pause. Removed pause (code becomes slow and still crashes). Julian suggest to use a redraw now. Should look if gnuplot works.
- The exercise of Dieter seems to be bearing-only. Bearing only is covered in question 10.6.2, but this is not a very informative question.
- The Student version of MatLab costs 89 dollar (additional toolboxes 29 dollar).
28 November 2011
- Looking for a good introduction of the localization chapters 7 and 8. Problem is that both EKF and Particles are already covered. Possible presentations are h_kalman.pdf from Freiburg '09 or 7-particle-MCL.ppt from New York. In both cases I should skip part of the introduction.
- 09d-ekf-localization.pdf from Freiburg '11 nicely skips most of the previous parts.
- 11-pf-mcl.ppt.pdf from Freiburg '11 is less interesting. New York's presentation is a better choice.
- Could find some video's of ETHZ's Crab robot, but not from Pygmalion.
- Alternative could be Siegwart's Localization slides , but here both EKF and MCL are covered in a single lecture.
- A mini-display port to VGA adapter works fine. Best resolution for beamer in A1.30 is 1280x720, but this means that Wolfram's slides are cropped both vertical as horizontal. With 1280x960 it looks good on the screen (and horizontal stretched on my screen).
27 November 2011
- Read chapter 4 of Introduction to Autonomous Mobile Robots (1st edition). The chapter is quite long, because it includes a short introduction to image processing for robotics. An example is given of image fingerprinting for localization.
21 November 2011
17 November 2011
16 November 2011
- Found a 3D description of ICP (including generalized-ICP) at the Mobile Robot Programming Toolkit.
- Same set contains a number of datasets from Freiburg and Malaga. Downloaded Malaga dataset 2006, 1.9 km of outdoor Odometry and Sick LMS measurements.
- Format is a binary RawLog. Looked if I could decompress the file, but I failed. The textfiles are generated from the RawLog file (without timestamps).
- First result is not that good, should check signs of ODO and LASER:
Commited experimantal rev. 2249.
15 November 2011
- Checked the origin of dead reckoning. I am clearly on Oxford's site: just an adjective in the sense of "complete(ly)" or "absolute(ly)," also found in "dead wrong," "dead ahead," "dead last," etc..
10 November 2011
- The presentation of Wolfram about histograms and grids is rather short. It ends with a video based on the IJCAI paper. There are some movies at Xavier's homepage, but not the one used by Wolfram. Elevator riding is the most interesting (although not for this lecture).
- Dieter Fox his project seems to be a nice project for next period. Made already the assignment.
9 November 2011
- Julian comes with the following hints (for next year) for Assignment 3.8.1:
- Is acceleration known of uncertain? So, should it be modelled as an action or as system noise?
- Iff x=Gauss(m,s) then y = Ax+B and y = Gauss(Ax+B, A.s.A^T)
- Provide a plot_gauss2d.m function.
6 November 2011
- Read chapter 3 of Introduction to Autonomous Mobile Robots (1st edition). Nice example of an omnidirectional robot and a two steer robot (which can drive the same trajectory, but not instantly. It has to stand still at certain position to turn its wheels from one configuration into the next. Also a nice control law to move smoothly from one pose in polar coordinates to another pose.
2 November 2011
6 October 2011
- Hessel found an article about Dynamic Markov Localization, which uses a octtree to get a global localization in Bonn's laboratory (well known from the ProbRob book). Should also work for Freiburg's building 79 (although no sonar). The AIC world has good dimensions and has sonar (although simulated).
28 September 2011
6 January 2011
- Continued with Thrun's warmup assignment.
- Moos' Road path seems to be 90 degrees rotated.
- Hessel has to be called with main(10000,1,100); % way of initialization, 1=interval, 2=random. With 10.000 particles, the algorithm runs out of memory.
Previous Labbooks