Increasing number of steps per episode/ iteration

[AizazSharif]

Hi @praveen-palanisamy,

I hope you are doing great.
I am facing one issue whether somehow I cannot cross the limit on number of steps ("max_steps" in scenario.py) per episode. The maximum in some scenarios I have reached is either 1024 or 2048 steps, and after the limit another episode starts.

Increasing the value in scenario.py breaks the code. Any help in this regard will be really appreaciated.

[praveen-palanisamy]

Good to hear from you!
Which scenario specifically are you trying to increase the max_steps for?
Based on the map for the Town where the scenario is situated/setup, it may not be recommended to have a larger value for max_steps since that could change the dynamics/behavior of the agent's learned policy.
The limits set in the scenarios file is based on what a reasonable multi-agent policy would take to drive the agents to their respective goals/destinations plus an extra buffer.

[AizazSharif]

Thanks for the reply @praveen-palanisamy.
In scenarios.py I am referring in general to all of the multi-agent scenarios where max_steps is either 500 or 2048 limit. The default goal destinations definitely are hard to reach using the current step limits. Is there any way to change it within the framework or make it flexible?
Any help would be much appreciated!

[praveen-palanisamy]

Okay!
Yes, you can experiment with different max steps limit by modifying the max_steps values for the scenario(s) in the scenarios.py file.
You mentioned in your first message that you tried that and ran into issues when you increased the values in scenarios.py. Could you share the full log of what went wrong?