1. Expedition A3 Ultra Overview

1. Expedition A3 Ultra Overview

1.1 Component Description

No.NameNo.NameNo.Name
1Head Display Screen9Thigh17Waist Handle
2Left and Right Speakers10Lower Leg18Left and Right Fisheye Cameras
3Interactive Binocular Camera11Dual-Shoulder Touch Sensors19Rear Fisheye Camera
4Power Switch12Dual Battery Compartments20LiDAR
5Microphone13Service Compartment21Chest RGBD Camera
6Upper Arm14Waist22Pelvis RGBD Camera
7Forearm15Foot
8Dexterous Hand16Charging Port

1.2 Robot Specifications

Robot ModelExpedition A3 Ultra
Whole RobotHeight1.74 m
Dimensions174 cm (H) 52 cm (W) 30 cm (L)
Weight60 kg
Active Degrees of Freedom51
Neck Degrees of Freedom2
Single-Arm Degrees of Freedom7
Waist Degrees of Freedom3
Single-Leg Degrees of Freedom6
Dexterous Hand Degrees of Freedom10 (optional, supports up to 20 DOF)
Perception System3D LiDARAvailable
Embodied Manipulation RGB-D CameraAvailable
Terrain Perception RGB-D CameraAvailable
Interactive RGB CameraRGB binocular camera
Surround-View RGB CameraAvailable
GPSAvailable
GPS RTKAvailable
UWB PositioningAvailable
Shoulder Touch SensorAvailable
CommunicationCommunication MethodsWiFi, 4G/5G network, Bluetooth
InteractionMicrophoneOmnidirectional microphone pickup
* Front and rear 8-microphone array
SpeakerAvailable
Interactive ScreenCurved screen
Indicator LightOmnidirectional indicator light
PerformancePeak Knee Joint Torque320 N·m
Movement SpeedMaximum daily walking speed: 1.8 m/s
Daily walking speed: < 1.2 m/s
Laboratory maximum speed: 2.5 m/s
Top running speed: 5 m/s
Single-Arm Payload5 kg
Energy and PowerBattery Capacity1152 W·h
Battery LifeOverall endurance: 8 h
* Actual endurance depends on usage conditions and is affected by walking speed and payload. Standing / walking duration ratio: 1:1
Charging MethodsSupports direct charging, battery swap, and autonomous charging
Charging TimeLess than or equal to 2 h
Charging Input Voltage110V-220V
Computing PowerBasic Compute BoardRK3588*2
Advanced Compute BoardNVIDIA Thor
OtherHandheld Wireless TerminalAvailable
Smart OTA UpgradeAvailable
Secondary DevelopmentSupported

1.3 Robot Joint Motion Range

Joint NameJoint LimitsJoint NameJoint Limits
Arm
J1 (Shoulder pitch): -170° to 170°WaistJ1 (Waist roll): ±20°
J2 (Shoulder roll): -91° to 91°J2 (Waist yaw): ±150°
J3 (Shoulder yaw): ±160°J3 (Waist pitch): ±30°
J4 (Elbow pitch): -60° to 140°Leg
J1 (Hip pitch): -168° to 154°
J5 (Wrist yaw): ±160°J2 (Hip roll): -60° to 100°
J6 (Wrist pitch): ±93°J3 (Hip yaw): ±158°
J7 (Wrist roll): ±33°J4 (Knee pitch): -10° to 148°
HeadPitch joint (Head pitch): 15° down, 25° upJ5 (Ankle pitch): -52° to 30°
Rotation joint (Head yaw): ±90°J6 (Ankle roll): ±20°

1.4 Robot Camera Field of View

Robot ModelPerception ConfigurationPerception Capability
Expedition A3 Ultra
Interactive binocular cameraSupports visual recognition and response in human-robot interaction scenarios
Left and right fisheye camerasSurrounding environment perception, object recognition, and obstacle avoidance
Rear fisheye camera
Chest RGBD cameraDepth and image perception of the operation area
Pelvis RGBD cameraDepth and image perception of the ground area

1.5 Network Topology

The HDU connects to the external network through WiFi and forms the robot's internal core network together with the MDU and ADU via eth_hdu (10.42.10.x). 5G serves as a backup link and can be used to connect to the external network when WiFi is unavailable.

Note: The Ethernet port inside the robot debug compartment is not enabled yet and is currently invalid. The Type-A port is for debug testing only, and the Type-C port is used only for charging the remote controller. Do not connect any other devices.

1.6 Industrial PC Specifications

The robot system consists of three core computing units:

MDU (Motion Domain Unit): Based on the RK3588 industrial control platform, it integrates an SoC with better real-time performance to provide stable computing power for motion-control algorithm modules. It is also the hardware unit closest to the robot actuators.
HDU (Head Domain Unit): Based on the RK3588S2 platform, it integrates a compute SoC, network peripherals, GPS, and abundant sensor interfaces, providing intelligent interaction and control computing capability for head-related software and algorithm modules.
ADU (AGI Domain Unit): Based on the NVIDIA Thor platform, it integrates a high-performance compute SoC, a small number of peripherals, and rich sensor interfaces, mainly providing sufficient computing power for embodied intelligence algorithm modules.
For RK3588 series specifications, refer to the Rockchip official website, and for NVIDIA Thor platform specifications, refer to the NVIDIA official page.

The table below shows the hardware parameters of the three core computing units:

Industrial PCHardware TypeSpecifications
MDUCPU8-core ARM big.LITTLE (4×A76 + 4×A55), up to 2.3 GHz, with L3 cache
Memory16 GB
Disk Capacity256 GB
HDUCPU8-core ARM big.LITTLE architecture (4×A76 + 4×A55), up to 2.3 GHz
Memory16 GB
Disk Capacity128 GB
ADUCPU12-core ARM (single cluster), up to 2.6 GHz, 12 MB L2 cache, supports SVE2
Memory64 GB
Disk Capacity358 GB