-
Notifications
You must be signed in to change notification settings - Fork 419
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Physics benchmarking #227
Physics benchmarking #227
Conversation
…physics interaction demo_runner calls.
Quite useful! Thank you!
Very cool!
Still, physics+RGBD at ~500 FPS is a great starting point! |
Definitely seems like your devfair's load is hurting things. I can ~1550 RGB FPS at 128x128 on mine. |
I am getting closer to ~300 FPS for bullet mode for phys_rgb at 128x128 (and really all the resolutions) :) |
… and type of objects
* Add enable_physics option to benchamrk.py * Add dynamic commands to demo_runner. * Added MotionType enum to physics bindings module * Fix timestep config not making it to PhysicsManager. Changed default timestep.
Motivation and Context
We would like to analyze the impact of non-static object manipulation through the new physics integration introduced by PR #91 .
This PR adds benchmarking for use of kinematic(default) and dynamic(Bullet physics) objects in habitat sim to benchmark.py with rgb and rgbd sensor modes.
Added new benchmark for time inside step function.
Added benchmark script parameters for number and type of object to simulate during benchmark.
To accommodate testing, MotionType has been added to python bindings in a physics module.
Bug fix: timestep was not making from config file to PhysicsManager initialization.
How Has This Been Tested
Local and devfair benchmarking:
Setup:
Results of kinematics benchmark on devfair:
Results of dynamics benchmark on devfair:
With 10 cheezit objects (simplified):
With 10 chefcan objects (complex meshes):
Benchmark notes:
Types of changes
Checklist