Shared Autonomy

Shared autonomy studies how people and automated systems can jointly control robots, vehicles, and other interactive systems while preserving human intent and agency. Common approaches infer goals from actions or language, blend human commands with learned policies, and use methods such as model-predictive control or control barrier functions to guide actions and enforce safety. Work is evaluated in settings including robotic manipulation, teleoperation, driving, navigation, and drone control, with measures such as task success, workload, safety, and user understanding. A central challenge is deciding when and how much the system should intervene: assistance can improve performance and reduce effort, but poorly timed or opaque actions can conflict with the user or undermine learning and control.