02‐16‐2024 Constraint Setting with New Single Objective Models

Attendees

• Chi-Hui
• Joe

Discussion

• Updating Normalize Dataset function
• Change the calculation of c1 to reflect rollout averages from excessive speed and queue length policies
• Upper bound U1 is avg reward per step from excessive speed policy
  • When evaluating excessive speed policy according to excessive speed reward
• Lower bound L1 is avg reward per step from queue length policy
  • When evaluating queue length policy according to excessive speed reward
• Constraint c1 is now (U1 - L1) * ratio + L1
• Leave rest of the algorithm the same