This is the distributed simulation environment code accompanying the paper "Fast Sample Collection Through Massive Parallelisation For Accelerated Deep Reinforcement Learning Development"
Please clone the bullet3 fork https://github.com/syslot/bullet3 on branch physx4.0 and build it according to the instructions
Example training processes can be found in src/Train for the Laikago robot
The training for ANYmal, Valkyrie, and the Mujoco environments can be found in src/distributed_train. The script train_all.py specifies the training parameters and training procedures as described in the paper.