Use of GPU for path optimization

rst-tu-dortmund / teb_local_planner

An optimal trajectory planner considering distinctive topologies for mobile robots based on Timed-Elastic-Bands (ROS Package)

http://wiki.ros.org/teb_local_planner

BSD 3-Clause "New" or "Revised" License

1.05k stars 551 forks source link

Use of GPU for path optimization #2

Open bausa opened 8 years ago

bausa commented 8 years ago

Would it be feasible to use a GPU to optimize the paths as opposed to multiple threads or do you suspect the overhead would surpass the benefit. From just quickly looking it seems like a good candidate for parallelization. Also, the matrix library used has an option to compile with GPU support. Would this substantially impact the planners performance?

beetleskin commented 8 years ago

I did not see any GPU/CUDA/OpenCL options in the g2o cmake file. But there is a library for Sparse Non-linear Least Squares Optimization on a GPU: https://github.com/OpenOF/OpenOF