← Back to case studies

Case Study

Semi-Automated Heavy-Duty Liquid Cargo Handling

Modular ROS 2 control system combining Cartesian motion control and vision-based feedback to improve safety, precision, and operational efficiency for heavy-load manipulation.

Control SystemsState EstimationPerception → ManipulationROS 2Industrial Robotics

At a glance

ROBOT
UR10e
Controlled test environment
CONTROL
Cartesian
PoseStamped targets + error correction
PERCEPTION
RGB-D
RealSense D435 + YOLO
FOCUS
Stability
Heavy-load semi-automation

Problem & outcome

The challenge

Liquid cargo handling introduces spillage risk, low precision with manual operation, and low flexibility with fixed automation. The system needed to support semi-automated operation with stable behavior while handling heavy loads.

What we built

  • Modular ROS 2 controller stack (separable pipelines)
  • Reusable Cartesian controller base for rapid iteration
  • Vision-based feedback loop with TF2 transforms
  • Compliance layer to keep motion stable under load

Architecture overview

Two controller pipelines feed a shared compliance controller, keeping the system modular and testable.

A) Cartesian controller pipeline

1
Keyboard / Hand guidance input
2
Error computation (Pose / Orientation)
3
Cartesian motion controller
4
Compliance controller
5
UR10e execution

B) Vision feedback pipeline

1
RealSense RGB-D stream
2
YOLO detection + filtering
3
TF2 frame transform → base
4
3D target error computation
5
Compliance controller → UR10e

Core components

ROS 2 HumbleC++ros2_controlTF2PoseStamped error modelYOLORealSense D435

Before vs after

BEFORE
  • Manual operation: fatigue + inconsistent precision
  • Fixed automation: hard to adapt to changing conditions
  • Weak feedback loops for correction and recovery
AFTER
  • Precise Cartesian targets with real-time error correction
  • Modular pipelines: swap input source (keyboard/vision)
  • Improved stability via compliance control strategy

Results & validation

Precision

High-precision pose tracking using PoseStamped targets and Cartesian error correction.

Responsiveness

Strong response to dynamic input changes; users preferred velocity-based input behavior.

Vision feedback

Accurate for nearby detections; performance impacted by distance/lighting—promising for adaptive manipulation.

Testing approach

  • Target pose tracking accuracy
  • Responsiveness to sudden target changes
  • Stability under motion execution
  • User evaluation survey + preferences

Challenges & mitigations

Vision limitations

Performance reduces with distance and adverse lighting; multi-detection ambiguity appears in cluttered scenes.

  • Confidence filtering
  • Depth-based 3D localization
  • Frame transform validation checks

Stability under load

Heavy-load manipulation requires stable compliance behavior and predictable correction dynamics.

  • Compliance controller boundary
  • Error smoothing strategies
  • Reusable Cartesian base for consistent behavior

Business impact

Why this matters operationally

  • Reduced spillage risk through more precise, stable motion
  • Less operator fatigue with semi-automated control
  • Higher adaptability vs fixed automation systems
  • Reusable control framework that scales to future logistics systems

Want this level of stability in your robotics stack?

If your system behaves well in demos but degrades in real conditions, we can help diagnose failure modes and ship fixes.

Contact us