Executive::processeventinlist() exception during second RL episode with AGV network

[siktodigonto_ipe17]

Hello FlexSim Community,

I am developing a congestion-aware AGV dispatching system by integrating reinforcement learning with FlexSim for my research project. The model contains ten orders, each with a different required quantity and lead time. At the beginning of each order, the reinforcement-learning agent selects:

1. The number of active AGVs
2. One dispatching rule from seven available rules

The objective is to fulfill each order within its specified time while minimizing AGV congestion.

The model runs normally when I execute it directly in FlexSim. I do not receive any exceptions or errors during a regular simulation run. However, when I run the model through my Python Gymnasium environment (flexsim_env_OHT.py), the following problem occurs:

time: 1097.437850 exception: Exception caught in Executive::processEventInList ().

The first episode completes successfully. The model then receives the reset command, and the initial observation for Episode 2 is returned correctly. Episode 2 also begins running and processes some actions normally. However, at a seemingly random simulation time, FlexSim produces the Executive::processEventInList () exception. Python then remains at “Waiting for Observation message” because FlexSim does not reach the next reinforcement-learning decision point.

Important observations:

• The error usually occurs during the second episode, after resetting the model.
• It does not always occur at the same simulation time.
• It does not occur at a specific order boundary.
• It does not consistently occur at the same control point or with the same AGV.
• The event log often shows an AGV PreArrivalEvent or TraversalEvent immediately before the exception.
• The model runs without this problem when it is executed directly in FlexSim.
• I tested the model in both FlexSim 26.0.0 and FlexSim 27.0.0, and the same problem occurs.
• Changing or disabling the dynamic addition and removal of AGVs did not eliminate the problem.
• The reset observation appears to contain cleared order, congestion, backlog, and origin-destination demand values.

I am attaching the FlexSim model, Python code, and screenshots. Could someone please help me identify the cause and explain how to prevent this exception so that the model can continue running across multiple reinforcement-learning episodes?

Thank you.