What’s New

The upcoming release of Isaac Lab-Arena v0.4 adds three new capabilities to the evaluation framework. Try out the early versions on the main branch!

Agentic Authoring

Use coding agents to author evaluation environments with Arena’s placement relations and composite tasks.

Agentic Evaluation Authoring

Newton-Based Workflows

Evaluate policies on insertion, contact-rich assembly, and deformable-object tasks using the Newton physics backend and Arena’s evaluation tools.

Newton-Based Evaluation Workflows

Multimodal VLM as a Policy

Use a multimodal vision-language model such as Astra as a robot policy, and compare action strategies using Arena’s evaluation tools.

Multimodal VLM as a Policy