Skip to main content

The Sequence in PredPreyGrass MARL

The sequence in Multi Agent Reinforcement Learning is of importance in simulation. Therefore, scheduling of the agents’ activation is particularly important and the activation regime can have a substantial effect on the behavior of a simulation [Comer2014]. Most many-agent algorithms can in practice handle only instances where all learning agents step at the same time, because sequentially stepping makes it compuationally very difficult to solve.

Sources:

  • Comer, Kenneth W. “Who Goes First? An Examination of the Impact of Activation on Outcome Behavior in Agent-Based Models.” George Mason University, 2014.