Multi-objective Optimal Control for Proactive Decision Making with Temporal Logic Models

摘要

Data-driven learning and proactive decision making are key ingredients of modern autonomous systems (AS). Robots, ranging from surgical assistants to autonomous cars, must seamlessly interact with other agents, which requires understanding their intents and behavioral models. While most current strategies used by a robot to understand the plan of a human rely on passive observations, recent work has focused significant attention on proactive intent decoding and decision making. Examples include autonomous cars that gently nudge into adjacent lanes to discern the driving style of nearby drivers for lane-merging or use large signs to proactively signal when pedestrians can safely cross at intersections.