Matlab logo
Manus VR
Matlab logo
Manus VR

Bridging Biological Motion and Multi-DoF Robotic Actuation

Humanoid robots operate with dozens of highly articulated degrees of freedom (DoFs) subject to unstable inverted pendulum dynamics. Onboard joint encoders, foot force sensors, and embedded IMUs suffer from accumulative state estimation drift and dynamic contact shocks.

To synthesize stable bipedal gaits, train neural motor policies, or benchmark push-recovery controllers, engineering teams require an external reference frame. Tracklab’s humanoid robotics mocap platforms deliver sub-millimeter positional ground truth and drift-free angular kinematics, enabling you to isolate algorithmic control errors from physical structural compliance.

Sub-Millimeter End-Effector & Pelvis Tracking (< 0.1 mm)

Accurately measure 6DoF center-of-mass (CoM) drift, pelvis tilt, and swing-foot ground clearance.

High-Rate Joint Angular Velocity Capture (Up to 360+ Hz)

Track high-speed impact transients during foot-strike transitions without motion blur.

Synchronized Kinetic & Kinematic Data Streams

Integrate optical 6DoF markers, IMU attitude, and Bertec/AMTI force plates into a single synchronised clock domain.

Specialised Motion Capture Applications for Bipedal Systems

Swing-Foot Ground Clearance Tracking.

Measure sub-millimeter clearances during swing phases to prevent dynamic trip events on uneven surfaces.

Center of Mass (CoM) & Pelvis State Estimation

Provide independent 6DoF telemetry of the pelvis and trunk to benchmark Kalman-filtered state estimators.

Ground Reaction Force Integration

Synchronise optical foot marker tracking with multi-axis force plates to calculate exact Center of Pressure (CoP) and joint torques.

High-Density Human Kinematics for Neural Locomotion & Manipulation Policies

External Ground-Truth Verification for Reactive Balance Controllers
Sudden mechanical disturbance response, capture point verification, and fall-prevention recovery steps.

Direct Data Streaming into Humanoid Control Workflows

Architecture / FrameworkConnection MethodDelivered Data Structure
ROS & ROS 2 (Humble/Iron)Native NatNet / VRPN nodessensor_msgs/JointState, geometry_msgs/PoseStamped, TF2 tree
Simulation (Isaac Sim / MuJoCo)LiveLink / USD / Python APIRigid-body transforms, full-body skeleton joint angles, collision hulls
MATLAB & SimulinkDirect NatNet SDK & TCP/UDP6DoF kinematic matrices, quaternions, time-stamped analog sync
C++ & Python SDKsHigh-performance client librariesLow-overhead packet streaming (< 2 ms compute overhead)

Direct Data Streaming into Humanoid Control Workflows

Onboard joint encoders only measure relative angle between links—they cannot detect structural flexion, gear backlash, or foot slip against the ground. Embedded IMUs drift over time and struggle during high-frequency foot-strike impacts. External optical motion capture provides an absolute global coordinate system with sub-millimeter precision, allowing you to accurately benchmark state estimators and calculate true center-of-mass positions.

Bipedal platforms create frequent self-occlusions as limbs cross or interact with payloads. We solve this by:

  • Supplementing optical systems with wearable IMUs (Xsens) for continuous kinematic continuity regardless of line-of-sight.
  • Designing redundant multi-camera arrays (typically 8 to 24+ cameras) providing continuous multi-angle coverage.
  • Utilizing Active LED continuous-ID markers, which pulse unique infrared signatures so cameras never lose track of a marker identity even when briefly occluded.

Yes. Tracklab systems provide low-latency Python APIs, C++ SDKs, and ROS 2 bridges that feed directly into physics simulators such as NVIDIA Isaac Sim, MuJoCo, and PyBullet. This allows researchers to capture real human motor trajectories and directly retarget them into virtual humanoid policy networks for reinforcement learning and sim-to-real transfer.

Yes. The software architecture supports simultaneous multi-body tracking. A human operator wearing an optical suit or sensorized gloves can be tracked in the exact same physical and spatial coordinate frame as the humanoid robot, which is ideal for real-time teleoperation, leader-follower trials, and collaborative ergonomics studies.

Build Your Humanoid Robotics Tracking Testbed

Whether you are designing a dynamic walking arena, instrumenting a high-speed perturbation treadmill, or capturing rich human imitation learning datasets, Tracklab’s Melbourne-based engineering team delivers turnkey tracking solutions tailored to your lab’s footprint and dynamic performance requirements.

Talk to an expert today, just call 03 7018 2121 or email us at info@tracklab.com.au.