search

xinleipan / VirtualtoReal-RL

BMVC 2017: Virtual to Real Reinforcement Learning for Autonomous Driving
MIT License
44 stars 13 forks source link

Error while installing TORCs #2

Open Slashzer-0 opened 5 years ago

Slashzer-0 commented 5 years ago

Hello,

Thanks for providing guidelines for the code for the paper. I am facing some issues while following your instructions for installing TORCs. I am using Ubuntu 18.04. I faced some library issues which I have resolved. Also, I changed some hard-coded path locations to get it working. At this point, I am getting the following error:

g++ gennmtab.o -L/home/deepayan/Downloads/VirtualtoReal-RL-master/TORCS/torcs-1.3.6/export/lib  -lopenal -lalut -lvorbisfile -L/usr/lib  -lplibjs -lplibssgaux -lplibssg -lplibsm -lplibsl -lplibsg -lplibul -lglut -lGLU -lGL -lpng -lz -ldl -lXrandr -lXrender -lXxf86vm -lXmu -lXi -lXt -lSM -lICE -lXext -lX11 -lm   -o gennmtab

/usr/bin/ld: out of memory allocating 3298534887680 bytes after a total of 290816 bytes

Do you have any suggestions for this? I am wondering whether an older version of Ubuntu might help? Also do you suggest some other source for installing TORCs?

Thanks again.

xinleipan commented 5 years ago

Hi @rtalapatra, we didn't test it on Ubuntu 18.04, it should work on Ubuntu 16.04.

Slashzer-0 commented 5 years ago

I tried running it in Ubuntu 16.04. I am getting the following error on make:

image

Any suggestions?

lingyunli1994 commented 5 years ago

@rtalapatra Hello，I also found this problem. Is this problem solved?

xinleipan commented 5 years ago

@rtalapatra Sorry for the late reply! I hope you already solved this problem, if not, it might be that your make file has a wrong target address.

xinleipan commented 5 years ago

@lingyunli1994 @rtalapatra try this new repo that has a python wrapper, which may save you lots of time if you are not familiar with lua: https://github.com/xinleipan/py_TORCS

lingyunli1994 commented 5 years ago

@xinleipan Thank you for your answer. This new one is no problem.However, virtual images cannot be saved.The paper says that you get 1673 virtual images.

xinleipan commented 5 years ago

@lingyunli1994 you can directly run the environment and save the observation as images. Note that this code only contains segmentations for the Michigan way environment, for other environment, you will need to change some of the torcs part code to get the segmentation.