/pyhiro-regrasp

python scripts to control the nxo, add fixture regrasp

Primary LanguagePython

pyhiro

python scripts to control the hiro/nxo/hrp5p

The following libraries are used

  1. Panda3D
  2. Networkx
  3. Numpy, Scipy
  4. MySQLdb

The following executable programs are used:

  1. MySQL set (including server, bench, dump, etc.)
    • Import 20170113freegrip.sql as into the DBMS system first.

The program follows these common sense:

  1. A 3d point or nd vector is represented using one row of np.array list
  2. A n-by-3 matrix is represented by n rows of np.array list

##Executing the program for Fixture Regrasp

  • fixture
  1. prepare object,fixture stl model.(mentionthat the fixture model contains its postion,the object model is near (0,0,0))
  2. Compute the grasps using freegrip_air.py (under manipulation/regrasp_onworkcell/grip)
  3. Compute the drop stable pose by dropworkcell.py(under manipulation/regrasp_onworkcell/dropsimulation)
  4. Compute the ik-feasible grasps all over a workcell using dropworkcellgrip.py (under manipulation/regrasp_onworkcell/grip)
  5. Plot and compute motion Sequence by nxtglplot(under manipulation/regrasp_onworkcell/examples), build the regrasp graph is in regriptpp.py (under manipulation/regrasp_onworkcell)
  • table
  1. Compute the grasps using freegrip_air.py (under manipulation/regrasp_onworkcell/grip)(if it has been compute in fixture program,there is no need to compute again)
  2. Compute the stable placements using freetabletopplacement.py (under manipulation/regrasp_onworkcell/grip)
  3. Compute the stable placements and ik-feasible grasps all over a table surface using tableplacements.py (under manipulation/regrasp)
  4. Build the regrasp graph using regriptpp.py (under manipulation/regrasp) Step 4 is integrated in regrasp/examples/nxtsglplot.py, see the final results by executing this file.
  • experiments: compute mant task
  1. generate startgoal first using startgoalgenerate,
  2. then prepare task ik first using taskik.py
  3. then build regrasp map using regriptpp_quick.py,so the build regrasp will be very fast about 0.02s each task.(regripp.py version time is 1.6s)
  4. in examples.the nxtglplot_experiment_quick will use regripp_quick.nxtglplotreal is for real robot,pathplot is for plot regrasp length and task sucess rate.(tbc:fixture table basket camera)
  • real robot: Nextage
  1. table regrasp:nxtsglplotreal(regrasp/examples)
  2. fixture regrasp:nxtsglplotreal(regrasp_onworkcell/examples)

##database

  • pay attention to the database(right and clean)

Delete the grasp data (in MySQL workbench):

SET FOREIGN_KEY_CHECKS=0;
truncate table freegrip.tabletopplacements;
truncate table freegrip.tabletopgrips;
truncate table freegrip.ik;
truncate table freegrip.freeairgrip;
truncate table freegrip.freetabletopplacement;
truncate table freegrip.freetabletopgrip;
SET FOREIGN_KEY_CHECKS=1;

Delete the tabletop regrasp data (in MySQL workbench):

SET FOREIGN_KEY_CHECKS=0;
truncate table tabletopplacements;
truncate table tabletopgrips;
truncate table ik;
SET FOREIGN_KEY_CHECKS=1;

Delete the fixture regrasp data (in MySQL workbench):

SET FOREIGN_KEY_CHECKS=0;
truncate table freegrip.freeairgrip;
truncate table freegrip.ik_drop;
truncate table freegrip.dropfreegrip;
truncate table freegrip.dropstablepos;
truncate table freegrip.dropworkcellgrip;
SET FOREIGN_KEY_CHECKS=1;

models

  • pay attention to objpath,workcellpath, the stl model(check size and position in blender)

parameters:

###in dropsimulation: dropworcell.py

  • experiment time: parameter ntimes,def updateODE limittime,
  • setup drop pos：def addobj setpos,height,
  • setup check stable:def checkrange minxminy，checkdiff,check repeat) ###in example:

nxtsglplot.py,nxtsglplotreal.py

  • determine which path to show : id

regriptpp_quick.py

  • loadgripstobuildgraph, there is choice to control the num of droptablepos to show the time effect results)

startgoalgenerate.py

  • tasks position range: grids-x,y
  • tasks angle range: discretesize in def savetoDB ###in grip: freegrip_air.py
  • reduceradius-lower more airgrips,
  • torqueresist-lower less grip

freetableplacement.py

  • dhover-0.15,0.12 higher less placement