Reinforcement Learning in Conflicting Environments for Autonomous Vehicles
This addresses decision-making problems for autonomous vehicles interacting with humans, but it is incremental as it applies existing methods to known dilemmas.
The paper tackled decision dilemmas like Newcomb's Problem and Prisoner's Dilemma in autonomous vehicles using Reinforcement Learning, finding that unmodified RL algorithms yield the maximum expected utility solution.
In this work, we investigate the application of Reinforcement Learning to two well known decision dilemmas, namely Newcomb's Problem and Prisoner's Dilemma. These problems are exemplary for dilemmas that autonomous agents are faced with when interacting with humans. Furthermore, we argue that a Newcomb-like formulation is more adequate in the human-machine interaction case and demonstrate empirically that the unmodified Reinforcement Learning algorithms end up with the well known maximum expected utility solution.