Hardware & control
Bimanual manipulation
Bimanual manipulation is the use of two hands or robot manipulators to carry out one manipulation task with spatial, temporal, or force coordination between them. The two sides may play symmetric roles, such as lifting a large object together, or asymmetric roles, such as one hand holding an object while the other operates a tool.
Also known as: dual-arm manipulation, dual arm manipulation, two-arm manipulation
Updated
Two arms do not automatically make one bimanual task
The literature does not use one perfectly settled boundary. The dual-arm manipulation survey distinguishes independent actions from coordinated actions and notes that some narrower definitions reserve “bimanual” for two manipulators physically interacting with the same object.
The practical distinction is coordination. Two arms sorting unrelated objects at the same time can be treated as parallel single-arm tasks. Folding a garment, opening a container, passing an object between hands or stabilising a part during insertion requires the two action streams to satisfy shared timing, geometry or contact constraints.
Coordination can be loose or tightly coupled
The bimanual manipulation taxonomy separates uncoordinated actions from coordinated ones. Loose coupling may require the hands to meet at a handover point or reach a state at the same time. Tight coupling adds continuing trajectory and force dependencies, as when both hands carry one rigid object.
Roles also vary. Symmetric actions give both hands similar jobs. In asymmetric actions, one hand can establish a reference frame or stabilise the object while the other performs a more precise movement. A policy and evaluation protocol should preserve that distinction rather than treating the arms as interchangeable.
Bimanual data needs a shared timeline
The ALOHA project demonstrates how bimanual learning depends on synchronised observations and commands from both sides during precise, contact-rich tasks. A dataset should record both arm and hand states, both action streams, camera views, object state, contacts, task phase and outcome on one clock.
Independent normalisation or dropped samples can alter the relative timing that made a demonstration work. For humanoids, torso motion, balance and whole-body constraints may also be part of the task even when the visible interaction happens at the hands.
Sources
Related terms
Hardware & control
Robot manipulation
Robot manipulation is a robot's controlled physical interaction with objects or its environment to change or maintain their state. It includes grasping, carrying, pushing, pulling, inserting, wiping, folding, tool use, and other tasks performed through selective contact. Manipulation can use a gripper, hand, tool, arm, or another part of the robot.
Hardware & control
Dexterous manipulation
Dexterous manipulation is the controlled, skillful reconfiguration of an object through coordinated motion and contact, often using multiple fingers. It can involve changing an object's pose within a hand, regrasping, sliding, rolling, finger gaiting, or making precise contact with the environment. The term describes capability, not a fixed minimum number of fingers or joints.
Data & collection
Teleoperation
Teleoperation is real-time human control of a robot from a remote or mediated interface. For humanoid training-data collection, the operator’s inputs are mapped to robot motions while cameras, proprioception, commands and outcomes are recorded, producing embodied demonstrations in the robot’s own observation and action spaces.
Models & learning
Action chunking
Action chunking is a policy design in which one observation is used to predict a sequence of several future robot actions rather than a single next action. The chunk can capture short-term motion structure and reduce the effective number of sequential predictions, while its execution and replanning schedule determine how quickly the robot can react to new observations.
Hardware & control
Whole-body control
Whole-body control coordinates a humanoid’s full multibody state, actuated joints and environmental contacts to pursue several motion or force objectives while respecting constraints such as balance, joint limits and friction. It is a family of hierarchical or optimisation-based methods, not one algorithm; implementations may output joint positions, accelerations or torques.