OMPL planner creates only a small number of states in roadmap/graph
Posted: 28 Jan 2019, 23:02
Hi,
I am sharing the following scene (3DoFHolonomicPathPlanning demo modified to use PRMstar with 10 seconds planning time):
https://drive.google.com/open?id=1JFzqd ... 2g5HjfgoW0
I have added some debug information (e.g. number of total collision checks and number of collisions)
The number of created states of 1400 in 10 seconds seems to be low.
In my application, where I plan in 3D+1D rotation for a path of the gripper to a goal config, the planner creates even fewer states:
About 300 states in 10 seconds. It is probably lower, because the collision checking is slower (about 55k in 10 seconds). But in my case only about 2.4% of states are in collision (1337/55k=2.4%).
I am in contact with one of the OMPL lead developer for advice and he finds it strange, that only so few states are created. He suggested to update V-REP to use the latest OMPL version with V-REP (1.4.2). Currently V-REP uses 1.1.0 (I believe).
If I try to compile the V-REP plugin for OMPL 1.4.2, I get the following error. It was able to recompile V-REP plugin with OMPL 1.1.0 (but not 1.4.2):
Thank you.
I am sharing the following scene (3DoFHolonomicPathPlanning demo modified to use PRMstar with 10 seconds planning time):
https://drive.google.com/open?id=1JFzqd ... 2g5HjfgoW0
I have added some debug information (e.g. number of total collision checks and number of collisions)
Code: Select all
Info: PRMstar: Space information setup was not yet called. Calling now.
Info: PRMstar: Optimization objective specified.
Info: PRMstar: Starting planning with 2 states already in datastructure
Info: PRMstar: Created 1394 states
Info: PRMstar: bestCost_: 4.640044
ob::PlannerStatus solved = task->planner->solve(in->maxTime) took 10035ms
Collision check count 166114
In collision check count 13369
In my application, where I plan in 3D+1D rotation for a path of the gripper to a goal config, the planner creates even fewer states:
About 300 states in 10 seconds. It is probably lower, because the collision checking is slower (about 55k in 10 seconds). But in my case only about 2.4% of states are in collision (1337/55k=2.4%).
I am in contact with one of the OMPL lead developer for advice and he finds it strange, that only so few states are created. He suggested to update V-REP to use the latest OMPL version with V-REP (1.4.2). Currently V-REP uses 1.1.0 (I believe).
If I try to compile the V-REP plugin for OMPL 1.4.2, I get the following error. It was able to recompile V-REP plugin with OMPL 1.1.0 (but not 1.4.2):
Code: Select all
Scanning dependencies of target v_repExtOMPL
[ 33%] Building CXX object CMakeFiles/v_repExtOMPL.dir/plugin.cpp.o
In file included from /usr/local/include/ompl/base/StateSpace.h:43:0,
from /usr/local/include/ompl/base/SpaceInformation.h:43,
from /usr/local/include/ompl/base/Goal.h:41,
from /home/uli/Software/V-REP_OMPL_plugin_1.4.2/v_repExtOMPL/plugin.cpp:7:
/usr/local/include/ompl/base/ProjectionEvaluator.h:49:22: fatal error: Eigen/Core: No such file or directory
compilation terminated.