VecEnv
An interface in Isaac Lab for managing parallel environments. It implements batched synchronous updates of environment state through PyTorch tensor operations, storing all environment states in GPU memory and executing in parallel via CUDA kernels.
Timeline (last 90 days)
Isaac Lab's VecEnv interface allows thousands of environments to run in parallel, storing the state of all environments in GPU memory and using CUDA kernels to execute physics stepping and observation computation in parallel.
By wrapping as an SB3-compatible VecEnv, NeSLE allows users to switch the backend from CPU multi-process simulation to GPU parallel simulation by replacing just one line of environment creation code.
All Facts (2)
Isaac Lab's VecEnv interface allows thousands of environments to run in parallel, storing the state of all environments in GPU memory and using CUDA kernels to execute physics stepping and observation computation in parallel.
50%UnverifiedBy wrapping as an SB3-compatible VecEnv, NeSLE allows users to switch the backend from CPU multi-process simulation to GPU parallel simulation by replacing just one line of environment creation code.
50%