~/projects/reinforcement-learning

Reinforcement Learning

Policy training and evaluation in Genesis and Isaac Sim.

Standard and custom cartpole environments trained across many parallel instances, then evaluated as single rollouts.

genesis
isaac-sim
parallel-envs
policy-eval

Isaac Sim viewport with a grid of parallel training environments

Genesis Sim · typical cartpole

The classic balancing task as a baseline.

Genesis viewer showing many cartpole environments training on a checkered floor

Genesis viewer showing a single cartpole balancing during evaluation

Genesis Sim · custom cartpole

A custom cartpole model in the same training loop.

Genesis viewer with many custom cartpole environments and their coordinate frames

Genesis viewer with the custom cartpole evaluated on its own

Isaac Sim · RL

The same workflow in NVIDIA Isaac Sim.

Isaac Sim with a grid of robots training in parallel

Isaac Sim with a single robot during evaluation