TowardstheTransferabilityofRewardsRecovered viaRegularizedInverseReinforcementLearning