Use the Generated Environment

Once you are satisfied with the environment, you can use it to evaluate a policy on the environment.

For example, you can use the policy runner to evaluate a PI policy on the environment. For other policy types, see Running a Real Policy.

Open one terminal and run the following command outside the Arena docker container to launch the PI policy server:

./isaaclab_arena_openpi/docker/run_openpi_server.sh

In the other terminal, run the following command to launch the policy runner. The commands below use the ready-made spec that ships with Arena; to evaluate a spec you generated yourself, point --env_spec at isaaclab_arena_environments/agent_generated/<env_name>.yaml.

Complete the shared Prerequisites before running this command.

python isaaclab_arena/evaluation/policy_runner.py \
   --viz kit \
   --policy_type isaaclab_arena_openpi.policy.pi0_remote_policy.Pi0RemotePolicy \
   --enable_cameras \
   --num_envs 1 \
   --num_episodes 2 \
   --env_spec isaaclab_arena_environments/kitchen_bench/kitchen_bench_lightwheel_open_fridge.yaml

PI controls DROID to reach the fridge and open its door in the agentically generated kitchen environment.