Hello there, I have a multi-agent setup where I use a custom algorithm with a Unity environment. I have a Python wrapper to allow the algorithm to interact with the environment, and everything with regards to that works fine. Recently, I made a few changes to the environment's layout, and after building the environment and running training on it, the agent ID's in the DecisionSteps object has changed from being [0,1] to being [2,3]. I have no idea why this suddenly started happening, it used to be fine until I changed the map layout. All well and good, I thought, it revealed that I didn't generalise my wrapper code enough to be able to interact with the environment regardless of agent ID's. Fixed that, tried again, and then the big issue: During training, the agent ID's suddenly change. They are [2,3] for a few seconds, then suddenly become [4,5]. I cannot understand the cause for this, and this is a rather big problem since each agent's actions are assigned to them based on their ID. Bringing me to my real question: Where do the agent ID's get assigned, in the DecisionSteps object? I would like to go have a look there, to see why the ID's suddenly change. Any other pointers as to why this might be happening would also be appreciated.