ENGINEAI PM01

Physical Size & Weight

Standing height 1380mm

Total weight with lithium battery: 40kg, empty body 32kg

Single arm static maximum payload 15kg

24 independent servo rotational joints total

Each arm 5 DoF, each leg 6 DoF, 320° rotatable waist joint, single pitch neck joint

Integrated planetary reduction brushless servo motor with dual-channel magnetic encoders

Hip/knee/shoulder peak torque 130N·m; wrist/neck continuous torque 50N·m

Built-in torque over-limit current limiting protection, compliant collision safety response

Intel N97 quad-core industrial CPU: real-time microsecond-level biped balance control loop

NVIDIA Jetson Orin NX 16GB GPU: multi-modal large model inference, visual detection, voice recognition, 3D mapping

Fully compatible with ROS 2 Humble robotics operating system, complete open SDK provided

Detachable quick-release lithium battery pack, capacity 10000mAh

Continuous dynamic movement runtime around 2 hours

Charging input 54.6VDC max 4.5A, full charging time 2 hours

Multi-layer battery protection: overcharge, over-discharge, over-current, short-circuit, over-temperature auto power-off

Head dual RGB + depth stereo camera: 3D mapping, obstacle ranging, human face and gesture recognition

Torso built-in 9-axis IMU: real-time body tilt, acceleration, angular velocity data for balance control

Dual wrist force-torque miniature sensors: contact force magnitude and direction detection for safe physical interaction

Waist miniature LiDAR: wide-range obstacle scanning and indoor autonomous navigation path planning

Chest embedded HD touch display + programmable RGB LED eye lights: program control, status feedback, emotional visual expression

Max stable walking speed 3~4km/h, peak running speed 2m/s

Supports forward/backward walking, side stepping, in-place rotation, running, dancing, forward somersault

Max negotiable indoor slope ±12° without tipping

Operation temperature 0°C ~ +40°C indoor clean environment

Storage temperature (battery removed) -20°C ~ +60°C

Humidity 10%~80% non-condensing, main body IP40 protection rating

Fully open-source full software stack without closed proprietary firmware, all motion control, sensor data processing, AI large model deployment code available for user modification and optimization

24 DoF humanoid joint layout + 320° rotating waist enables flexible human-like postures, running, jumping and dance movements; compact high-torque planetary joints balance agility and load capacity

Separate CPU motion control and GPU AI inference hardware avoids computing resource conflict, supports offline local large model operation without cloud server dependency

Multi-modal natural human-robot interaction: voice conversation, facial/gesture visual recognition, physical contact force feedback, touch screen operation, LED emotional light display

Fully modular mechanical design: each joint module can be individually detached, replaced or upgraded; battery pack quick swap within 30 seconds; all sensors adopt unified plug-in interface for easy hardware expansion

Native ROS 2 swarm communication protocol support, multiple PM01 robots can execute synchronized group dance, collaborative object transportation and coordinated assembly tasks for swarm robot research

Full body load-bearing skeleton: aviation lightweight high-strength aluminum alloy CNC machined, high tensile strength with low overall weight to reduce joint load and extend battery life

Outer protective shell: injection molded high-strength ABS composite plastic with matte silver coating, prevents skeleton and joint collision scratch, provides streamlined appearance for demonstration use

Internal electronic hardware: all computing boards, drive mainboards, IMU and power management circuits installed inside sealed aluminum alloy electromagnetic shielding cavities in torso chest, eliminates interference between high-power motor drive circuits and sensitive low-voltage sensor/AI computing circuits

Single joint internal structure: integrated brushless motor, multi-stage planetary gear reduction, dual-channel magnetic position encoder, built-in temperature and current sensing circuit, sealed cylindrical aluminum alloy housing; planetary gears amplify motor torque for heavy limb movement, dual encoders achieve sub-degree joint positioning precision for smooth humanoid motion

Power-on self-test stage

After installing charged battery and powering on, onboard Intel CPU automatically detects all 24 joint actuators, IMU, depth camera, LiDAR, wrist force sensors, touch screen and GPU module hardware connection status.

Load all sensor and motion control parameters into ROS 2 middleware, automatically execute full-body joint zero calibration movement to align encoder reference points. After self-check and calibration finish, robot enters standby state with neutral white eye LED light.

IMU balance data transmits directly to CPU real-time balance control loop with ultra-low microsecond latency.

Depth camera RGB/depth data, LiDAR obstacle ranging data, wrist contact force sensor data sent to GPU for AI model calculation: object detection, gesture recognition, 3D environment mapping, collision force response logic.

Built-in noise-canceling microphone voice audio transmitted to GPU speech recognition large model for voice-to-text conversion and user command semantic analysis.

CPU low-latency motion control pipeline: receives IMU body tilt, joint encoder position, wrist contact force feedback data, calculates target rotation angle for all 24 joints in real time, sends motion commands to each servo joint module to execute smooth humanoid limb movement, dynamically adjust body posture to avoid tipping during walking, running and jumping.

GPU multi-modal AI decision pipeline: processes all preprocessed sensor data to generate high-level autonomous task decisions: voice command task parsing, human interactive response, object grasping coordinate calculation, indoor navigation path planning, multi-robot synchronized movement scheduling. All AI decision results transmitted to CPU motion control pipeline to trigger preset humanoid motion sequences and interactive feedback actions.

Dual computing pipelines run circularly without interruption during robot operation. Multi-layer safety monitoring logic real-time tracks joint operating current, internal joint temperature, wrist contact force values.

When colliding with humans or obstacles, wrist force-torque sensor detects abnormal contact force, triggers joint torque limiting compliant safety response, immediately reduces joint output torque to prevent human injury and mechanical skeleton/joint damage.

If any joint exceeds safe temperature or current threshold, corresponding motor automatically switches to power-limited low-speed operation to avoid permanent actuator burnout failure.

Factory disassembly modular packaging to reduce transportation volume and prevent transit collision damage. On-site assembly must strictly follow official assembly manual in fixed order: install torso aluminum skeleton and internal shielding electronic modules → sequentially mount leg, waist, arm, neck joint modules via standard quick-lock fasteners → install ABS outer protective shell panels → insert fully charged battery pack → power on and execute mandatory full-body joint zero calibration via chest touch screen, calibration complete before any motion test.

Only operate on flat, smooth hard indoor floors (epoxy workshop, ceramic tile lab, solid wood exhibition floor). Uneven carpet, gravel ground, slippery wet floors and outdoor rugged terrain are forbidden without customized balance algorithm calibration, which greatly increases tipping risk and joint mechanical collision damage.

Reserve minimum 3m×3m unobstructed empty space as robot operation zone for walking, running and dynamic movement testing. Remove sharp furniture, heavy metal equipment and fragile glass products from activity space to eliminate collision hazards.

Ambient illumination maintained between 200~1500lux stable indoor light; total darkness or strong spotlight will severely reduce vision sensor accuracy and AI object detection performance.

Daily startup calibration: every time shutdown exceeds 4 hours, full-body joint zero automatic calibration must be executed before running motion programs or autonomous AI interaction tasks, uncalibrated operation leads to inaccurate joint positioning, distorted movement posture and increased tipping risk during dynamic locomotion.

Monthly mechanical backlash compensation calibration: every month of continuous operation, use official PC calibration software to complete joint mechanical gap compensation calibration, eliminate gear wear clearance accumulated by daily movement, maintain sub-degree joint positioning precision and smooth humanoid movement effect for long-term continuous operation.

Quarterly multi-sensor recalibration: every three months re-calibrate stereo depth camera color balance and ranging precision, LiDAR ranging offset error correction, IMU accelerometer and gyroscope drift compensation, eliminate sensor parameter drift caused by long-term operation and temperature change, guarantee stable accurate perception data input for onboard large model inference.

University robotics and artificial intelligence research laboratories for biped balance algorithm, multi-modal vision-language large model embodied deployment, natural language human-robot interaction, multi-robot swarm coordination, humanoid autonomous indoor navigation research

Light industrial human-machine collaborative assembly workstations for consumer electronics, small home appliances, automotive electronic component repetitive picking, placement, screw locking and product visual inspection tasks; wrist force-torque compliant collision safety control enables barrier-free co-work with human operators

Commercial exhibition hall intelligent service demonstration platforms for visitor greeting, guided tour, interactive conversation, choreographed group dance performance

Aerospace bipedal humanoid robot prototype function verification projects

Indoor service robot research platforms for automatic cargo handling, environmental inspection

All pre-operation calibration must be completed before running any motion programs; initial test operation set low movement speed in controlled empty space.

Monitor battery power level during long-time continuous operation, prepare spare quick-swap battery packs to extend working duration.

Forbid operation near stair edges, high drop-off areas to prevent robot falling and severe mechanical damage.

Do not block ventilation heat dissipation openings of onboard computing hardware, avoid overheating of GPU and CPU modules.

Keep all motion control firmware and AI model software versions consistent across all robot modules, document all customized motion algorithms, sensor calibration parameters and task program code for backup.

Carry out regular maintenance on joint modules, connecting cables and sensor hardware; use dry compressed air to clean surface dust, no liquid cleaning agents allowed.

Backup all motion programs, calibration parameters and AI model files before major algorithm or firmware upgrade.