Question about modifying action space for variable phase selection in sumo-rl

[joyce914qq]

I’m currently working on an extension of sumo-rl for adaptive traffic signal control, and I have a quick question:
Would it be possible to modify the current action structure so that instead of only choosing between “keep current phase” and “switch to next”, the agent can directly select from a list of available signal phases (i.e., variable phase selection)?

[LucasAlegre]

Hi @joyce914qq , the current action structure is already like you mentioned. The action space is composed of the N available green phases, and the agent can choose among them at any decision time.