Vrep behaves differently on different computers

[Rodrigo]

Hi,


I have been running simulations using V-Rep 3.1.3 pro edu on my 4 core, Intel Core I5 Laptop which runs Ubuntu 14.04 64 bits.

Simulations run fine and are repeatable, that is every time I run the simulation the robot gets to the same position in space at the same simulation time, however when I run the simulation on another computer, a 32 core AMD Opteron(tm) Processor 6386 SE server running Ubuntu 12.04 64 bits, for the sake of running more simulations in parallel, the simulation behaves completely different, that is for the same simulation time the robot gets to a completely different position compared to the other computer. So as long as I run simulations in only one computer the results are the same but when I change to the other computer simulation results change. I copied the simulator folder and scenes from one computer to another so I am using the same program on both machines.

At first I thought it was because I was using a custom physics timestep for the Bullet engine (If I use the ODE engine I get different results for every run even on the same computer) so I changed back to the default engine settings but the problem persisted. I have also noticed that there are some library dependencies that are not met in the Ubuntu 12.04 server (ldd command):

Code: Select all

        linux-vdso.so.1 =>  (0x00007fffcfc00000)
        liblua5.1.so => not found
        libQt5Core.so.5 => not found
        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f9bc9d70000)
        libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f9bc9a68000)
        libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f9bc9850000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f9bc9490000)
        /lib64/ld-linux-x86-64.so.2 (0x00007f9bc9fb0000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f9bc9190000)

Compared to the Ubuntu 14.04 Laptop:

Code: Select all

	
   linux-vdso.so.1 =>  (0x00007fffe4bfe000)
	liblua5.1.so => /usr/lib/x86_64-linux-gnu/liblua5.1.so (0x00007fb010937000)
	libQt5Core.so.5 => /usr/lib/x86_64-linux-gnu/libQt5Core.so.5 (0x00007fb010290000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fb010072000)
	libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fb00fd6e000)
	libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fb00fb57000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fb00f791000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fb00f48b000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fb00f286000)
	libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fb00f06d000)
	libicui18n.so.52 => /usr/lib/x86_64-linux-gnu/libicui18n.so.52 (0x00007fb00ec65000)
	libicuuc.so.52 => /usr/lib/x86_64-linux-gnu/libicuuc.so.52 (0x00007fb00e8eb000)
	libglib-2.0.so.0 => /lib/x86_64-linux-gnu/libglib-2.0.so.0 (0x00007fb00e5e3000)
	librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fb00e3db000)
	/lib64/ld-linux-x86-64.so.2 (0x00007fb010b8f000)
	libicudata.so.52 => /usr/lib/x86_64-linux-gnu/libicudata.so.52 (0x00007fb00cb6d000)
	libpcre.so.3 => /lib/x86_64-linux-gnu/libpcre.so.3 (0x00007fb00c92f000)

Is this because of the missing libraries or is there any other reason for this strange behaviour?

Thanks in advance!

[coppelia]

Hello,

have first a look at this post.
So it can depend on various factors. Also on how (if) you synchronize some external controllers with V-REP.
What kind of simulation are you running, or what type of robot.
As you may know, very small differences will cumulate during dynamics simulation, and might appear as very large positional differences eventually.

Cheers

[Rodrigo]

Hi

I am doing multi robot simulations in which a group of simple 1DOF, box robots are attached to each other by force sensors, I am setting some parameters with Java at the beginning of the simulation but for the rest of the time V-Rep runs alone.

I have tested the simulation in different computers and I have found that results depend on the operating system used:

I have tested on another laptop running Ubuntu 14.04 and I get exactly the same position values as in my Ubuntu 14.04 Laptop (up to four decimal places, which is what the position tool in Vrep displays, I have not tested for higher precision), I also ran the simulation in a machine running ArchLinux and got exactly the same results as in my Ubuntu 14.04 laptop.

I installed Ubuntu 12.04.5 in my laptop using VirtualBox, the simulation returns exactly the same values as in the 32 core Ubuntu server that also runs Ubuntu 12.04.5

As an additional check I ran the simulation in Windows and got different results compared to the other two cases.

So results are repeatable, at least in my case, as long as I run the simulator under the same version of the operating system, the question is: could there be a way to make V-REP behave in the same way under the two Ubuntu systems, maybe by installing missing libraries or recompiling the program?

Thanks!

[coppelia]

Hello Rodrigo,

can you send us a scene file so that we can test it here?

Cheers