Atlas A7

Robotics
Project configuration

Atlas A7 reference
Left / right color concepts; not physical product photographs. Single and dual-arm configurations are separate selections.

Seven-axis lightweight research manipulator

Discover the details

System overview

Built around
your research.

A seven-degree-of-freedom arm for fixed-workstation manipulation and robot learning. Kinematic redundancy, hand-guided teaching and combined force-position control support motion planning, impedance-control and imitation-learning research. Select a left arm, right arm or bimanual setup.

Redundant kinematics
7 axes
Nominal maximum payload
3 kg
Nominal maximum reach
739.45 mm

Configuration, lead time and delivery scope confirmed with your inquiry.

Design & capability

Designed for the task.

More freedom for planning

A seven-axis anthropomorphic arrangement supports end-effector orientation, joint obstacle avoidance and redundancy resolution. Nominal arm mass is 4.3 kg with +/-0.1 mm repeatability.

Hand-guided demonstration

The manual lists gravity compensation, zero-force teaching and trajectory replay at a 100 Hz teaching sample rate. Joint torque-related readings are not direct end-effector contact forces.

Control and research integration

Position, velocity, torque and MIT hybrid control are documented, alongside Python / C++, ROS 2 Humble, MoveIt2, Gazebo and URDF. Confirm interface versions.

Specifications

Configuration details and measurement conditions.

Structure and performance
Degrees of freedom7
Single-arm mass4.3 kg
Maximum payload3 kg
Maximum reach739.45 mm
Repeatability+/-0.1 mm
MaterialsAluminum-alloy frame, plastic enclosure
MountingRigid, level fixed workstation; hole pattern per version drawing
ConfigurationsLeft / right / bimanual

Maximum payload depends on pose, tool mass, center of gravity and dynamics. Repeatability is not absolute positioning accuracy.

Joint specifications
JointRangeMaximum speedRated torquePeak torque
J1-60 to +180 degrees180 deg/s15 Nm30 Nm
J2-180 to +15 degrees180 deg/s15 Nm30 Nm
J3+/-158 degrees180 deg/s15 Nm30 Nm
J4+/-110 degrees180 deg/s15 Nm30 Nm
J5+/-158 degrees180 deg/s3 Nm7 Nm
J6+/-90 degrees180 deg/s3 Nm7 Nm
J7+/-158 degrees180 deg/s3 Nm7 Nm

Coordinates, zero positions and signs depend on arm side and driver version. Peak torque is not a continuous rating.

Power, communication and control
System powerDC 48 V; supply rated for at least 6 A with operating margin
Combined power<300 W
CommunicationCAN 2.0, 1 Mbps; USB-to-CAN adapter
External ports1 power / 1 CAN
Joint encoders21-bit dual encoders
Control modesPosition / velocity / torque / MIT hybrid
TeachingGravity compensation, zero-force recording, trajectory replay
Teaching sample rate100 Hz
Recorded fieldsJoint position, velocity and torque-related data; units and timestamps per SDK

CAN bandwidth and teaching sample rate are not policy inference or end-to-end control frequency. Use the complete-system 48 V supply requirement.

Software and configuration
Recommended environmentUbuntu 22.04 LTS
SDKPython / C++, version-specific APIs and examples
Middleware / planningROS 2 Humble / MoveIt2
SimulationGazebo, URDF and matching controller configuration
ComputeExternal workstation or cloud service; GPU, models and trained skills not included by default
Temperature / humidity5-50 C / 35-80% RH, no frost
Noise<60 dB; source measurement conditions unspecified
AccessoriesUSB-to-CAN, cables, fasteners and tools per bill of materials
Confirm separatelyPower adapter, gripper, tactile end effector, cameras, fixtures, compute and safety equipment

Delivery and validation

Agree on the complete configuration.

Nominal manual specifications, not Actrium measurements. Atlas A7 is a project product name. Force control alone does not establish safe shared human-robot operation. Validate the complete installation and agreed tasks.

  • Freeze arm side, quantity, end effector, mounting, power and USB-to-CAN configuration.
  • Provide agreed drivers, SDK examples, models, version notes and license scope.
  • Verify zero positions, limits, emergency stop and communication-loss handling. Enable the proper mode for hand-guided teaching; do not force an unpowered or braked arm.
  • Accept recording, replay and tasks under defined installation, load and speed conditions.

Training and deployment

Plan compute for this platform.

Plan GPU compute, model training and hosted inference, with control kept on the robot.

Training & Compute →

Related products

Build with Actrium

Start with your task.

Define the devices, data and evaluation your project needs.