Do I need more than an Jetson AGX Xavier Developer Kit with ISSAC to train Deep Reinforcement Learning Networks?

I’ve considered building a Deep Learning workstation comprised of two RTX graphics cards to develop and train Deep Reinforcement Learning and other types of Neural Networks and machine learning algorithms for robotics applications.

Will buying a Jetson AGX Xavier Developer Kit with Issac SDK be enough so I won’t need to build a workstation as well?

I’m also very interested in doing Deep Reinforcement Learning with ISAAC. To my understanding it’s better to have a workstation to do the training in simulation, or it can be done on cloud GPUs. And then you transfer the trained skills to the jetson to run it on the real robot.

Anybody has experience with that ? What kind of workstation is recommended or what cloud GPUs service ?

Isaac Gym is NVIDIA’s platform for Robot’s reinforcement learning. It will be available later this year (late summer). Stay tuned…

Reinforcement learning will be done on a workstation or cluster. Once trained on a workstation/cluster, Jetson + Isaac SDK can be used for inference.