Clarification on Collaborative Scenarios and Heuristic Wrapper Support in Gigastep

[Louay-Ben-nessir]

Hello,
The Gigastep paper mentions support for collaborative multi-agent scenarios. However, while exploring available examples, I encountered either fully competitive scenarios where control extends to all agents or random adversarial agents.

I have the following questions:

• Do you explicitly support fully collaborative scenarios? like having a flag for cooperative in the make_scenarion function
• Would it be feasible to implement a heuristic wrapper for the enemy agents similar to the one found in SMAX's heuristic_enemy.py (https://github.com/FLAIROx/JaxMARL/blob/main/jaxmarl/environments/smax/heuristic_enemy.py)? Such a wrapper could aid in setting up collaborative multi-agent environments.

I believe these clarifications would be valuable not only for me but also for the broader community using Gigastep.

Thank you for your time and consideration!