Robotics
Shared autonomy
Definition
Shared autonomy is an interaction mode in which a human and a robot jointly influence the robot's actions. The robot interprets human input and supplies assistance while leaving some control or decision authority with the person.
Also known as: Shared control
Updated
Combine human input with robot assistance
In direct teleoperation, the robot aims to execute the operator's commands. A shared-autonomy system also estimates what the operator is trying to accomplish and modifies, completes, or constrains those commands. Examples include steering a manipulator toward a likely grasp, maintaining clearance from an obstacle, or automating part of a multi-step task while the person chooses the goal.
Dragan and Srinivasa formalise one form of shared control as policy blending. Their system predicts user intent and arbitrates between the user's input and the robot's predicted policy. Their manipulator study also found that useful arbitration depends on context, including confidence in both the robot and the user.
Shared autonomy is a spectrum
The division of authority can change by time, task, or state. A person might command direction while the robot controls grasp alignment, or the robot might act autonomously until uncertainty rises and then request more input. Some systems blend continuous commands, while others allocate whole subtasks.
This differs from full autonomy because the human remains inside the decision or control loop. It also differs from a fixed safety limiter. A safety limiter can reject dangerous commands without estimating the intended goal, whereas shared autonomy often uses an intent model to provide task-directed assistance.
Incorrect assistance can oppose the user
Intent inference is uncertain, especially early in a motion or when several goals are close together. Confidently assisting toward the wrong object can be worse than providing no assistance. Arbitration must also avoid unstable or inescapable states created by incompatible human and robot commands.
Performance measures should include more than task time. Operator effort, reversibility, predictability, accessibility, and the ability to override assistance affect whether control is genuinely shared. People can also change their behaviour in response to the robot, so an intent model trained on unassisted input may become inaccurate after assistance is introduced.
Sources
Related terms
Teleoperation
Teleoperation is the control of a robot by a human operator from a separate location or interface. The operator supplies commands while feedback, such as camera images or the robot's motion, helps them guide the task.
Behavior tree
A behavior tree is a hierarchical control structure that selects and switches between an autonomous system's actions according to the status returned by its child nodes. Robots use behavior trees to compose reactive task logic from reusable actions and conditions.
Reinforcement learning
Reinforcement learning trains an agent to choose actions that maximize expected cumulative reward through experience with an environment. In robotics, the learned policy can select movements or higher-level behaviors from observations.
Active perception
Active perception is perception in which a robot chooses motions or interactions partly to obtain more useful observations. Examples include moving a camera to reveal an occluded object or touching an object to reduce uncertainty about its state.