Autonomous vehicle decision-making problems

Abstract

Decision-making in autonomous vehicles (AVs) sits at the intersection of control theory, machine learning, and human-centered design. This article examines core decision problems that challenge AV deployment: perception uncertainty, real-time planning under partial observability, multi-agent interactions in dense traffic, safety versus comfort tradeoffs, and the ethical allocation of risk. We review contemporary methods — rule-based stacking, imitation learning, reinforcement learning (single-agent and multi-agent), model predictive control (MPC) hybrids, and probabilistic decision frameworks — and synthesize recent literature on robustness, interpretability, and regulatory implications. A synthetic research study compares representative approaches across urban and highway scenarios, showing trade-offs between safety incidents, decision latency, and passenger comfort