Open Harry67Hu opened 6 months ago
Hi Harry, thank you for your interest in our MOSMAC benchmark. We are currently in the process of finalizing the codebase to ensure it is well-documented and user-friendly. We expect to have it available very soon. Once the code is released, we would be thrilled to explore collaboration opportunities with you and your research team. Please feel free to reach out to us again when the code is available, and we can discuss potential ways to work together!
You have done a very good job! I believe this work can serve as an example of multi-task and long-term planning in swarm systems. We are currently trying to control a MARL-trained swarm system with LLMs. I have a question regarding your code, specifically about the following line: [self.n_actions = self.n_actions_no_attack + self.n_local_enemies ] Does this mean that the agent's attack actions are no longer targeting specific IDs, but rather any agent within the observation radius (for example, sorted by distance)? If so, have you written a function to transform the action accordingly? Wishing you all the best!
Could you please make the code for this environment open source? It would be extremely beneficial for our research work! We could consider collaborating!