Docker compose fails in AI Challenge: model container not detected

A frustrated participant in the No_Quarter_Robotics AI for Industry Challenge posted that running `docker compose -f docker/docker-compose.yaml up` leads to a deadlock: the evaluation engine never finds the model container. Logs show the model container initializing successfully with the 'Johnny02' policy and Zenoh configuration, but the engine repeatedly checks for a node named 'aic_model' and retries indefinitely. Three trials (trial_1, trial_2, trial_3) are parsed but never executed because the model never appears ready. The user laments that debugging locally is nearly impossible, referencing other unresolved threads about similar errors after submission, slow docker builds, and containers not terminating cleanly.

The issue highlights a persistent challenge in competition environments where Docker-based local evaluation diverges from the remote grading pipeline. The 'No node with name aic_model found' error suggests a networking or timing mismatch between containers, likely related to Docker Compose's service discovery or the Zenoh shared_memory configuration being disabled. With multiple topics on the same challenge covering failed submissions and empty result files, it's clear that many participants are hitting similar walls. This particular case underscores the need for better debugging tooling and standardized container orchestration in AI competitions to prevent wasted hours on infrastructure issues.