PhD Researcher

Zongyuan Zhang

Project based at

Lead Partner Organisation

Contact task execution by robot with non-rigid fixation

Start Date: 1st March 2024

Expected End Date: 1st September 2027

Supervisory Team

• Principal Supervisors: Professor Jonathan Roberts,
• Associate Supervisors: Professor Will Browne and Dr Chris Lehnert

The Challenge

Many industrial processes such as grinding, polishing, sanding and welding require robots to perform high-precision contact tasks while maintaining consistent force against a surface. Traditionally, these applications rely on rigidly fixed robotic systems and carefully positioned workpieces to achieve the required levels of accuracy and stability.

However, manufacturing environments increasingly require flexibility. Mobile robots can significantly expand the workspace available to robotic systems and allow them to interact with large workpieces positioned in arbitrary locations. The trade-off is that mobile platforms introduce movement, compliance and positioning uncertainty, reducing system rigidity and making precise contact tasks significantly more difficult.

This project addressed the challenge of enabling robots with non-rigid fixation to perform high-precision contact tasks reliably and cost-effectively in real-world manufacturing environments.

Research Questions

What We Did

Research Activities

The project investigated robotic grinding and polishing applications in partnership with Vaulta, focusing on the removal of oxide layers from electrical contact points used in battery manufacturing.

Key research activities included:

• Developing robotic grinding and polishing solutions for large, variably positioned components.
• Investigating hybrid force/position control strategies for contact tasks performed on non-rigid systems.
• Exploring microphone-based acoustic sensing as an alternative to traditional force-torque sensors.
• Collecting and analysing high-frequency vibration and sound signals generated during grinding operations.
• Developing machine learning approaches to interpret acoustic feedback and estimate contact conditions.
• Extending contact-task research toward humanoid robot platforms.

Methods

The project combined:

• Hybrid force/position control techniques.
• Acoustic sensing and signal processing.
• Machine learning models for contact-state estimation.
• Mobile manipulation research.
• Robotic grinding and polishing experiments.
• Hardware prototyping and software development.

A novel sensing approach was developed that uses low-cost microphones to capture vibration signals generated during grinding. These signals can be used to infer contact conditions and provide feedback for closed-loop control, reducing reliance on expensive force sensors.

Industry Engagement

The project included a successful industry placement with Vaulta, where research was translated into manufacturing applications.

Industry engagement activities included:

• Collaborative development of robotic grinding solutions for Vaulta’s manufacturing requirements.
• Testing and validation of robotic surface preparation methods.
• Knowledge transfer between researchers and industry personnel.

An earlier collaboration was explored with Urban Art Project (UAP) for sculpture sanding applications. While this partnership did not proceed because of timeline constraints, the work helped identify broader industrial use cases for mobile robotic contact tasks.

Site Visits, Observations, Prototypes and Demonstrations

Project activities included:

• Industry placement activities at Vaulta.
• Robotic grinding and polishing prototype development.
• Acoustic sensing system development and testing.
• Contact-task demonstrations using the Unitree G1 humanoid robot.
• Early-stage humanoid manipulation demonstrations using a dexterous robotic hand to perform surface-cleaning tasks.

Results, Outcomes and Impact

Research Findings

The project generated important new knowledge in several areas:

Sensor Substitution Viability: The research demonstrated that microphone-based vibration feedback can effectively substitute for, or augment, traditional force-torque sensors during high-precision robotic contact tasks.

Control Loop Optimisation: The project produced new insights into stabilising hybrid force/position control systems when operating from mobile or non-rigid robotic platforms.

Mobile Manipulation Under Compliance: The work advanced understanding of how robots can interact with flexible workpieces and environments where rigidity cannot be guaranteed.

Acoustic Force Estimation: The research established a foundation for future work investigating how acoustic signals can be translated into accurate force estimates for robotic control applications.

Tools and Frameworks Developed

Robotic Grinding and Polishing Framework: A robotic grinding and polishing application framework tailored to Vaulta’s manufacturing requirements. The framework automates the removal of oxide layers from electrical contact points while maintaining consistent cleaning quality without requiring rigid positioning fixtures.

Low-Cost Acoustic Sensing System: A hardware and software solution that captures high-frequency vibration feedback during grinding and polishing tasks. The system provides a practical, low-cost alternative to conventional force sensing technologies.

Acoustic Learning Framework: An algorithmic framework that maps acoustic data to robotic contact states, enabling intelligent monitoring and control during contact-based manufacturing processes.

Hybrid Force/Position Control Scheme: A control architecture designed to improve the stability of robotic contact tasks undertaken on mobile or non-rigid robotic platforms.

Industry Outcomes

The project delivered significant benefits to industry, including:

• Demonstrating cost-effective alternatives to expensive force-torque sensing systems.
• Enabling robotic contact-task automation without the need for highly rigid positioning infrastructure.
• Providing Vaulta with practical solutions for battery manufacturing processes.
• Increasing capability in advanced robotics and automation through researcher industry placement activities.
• Demonstrating how flexible robotic systems can replace traditional fixed manufacturing layouts.
• Creating opportunities to transfer the technology to other manufacturing applications, including sanding, polishing and surface finishing processes.

The project also initiated discussions regarding future deployment opportunities in automated sculpture sanding and other advanced manufacturing applications.

Publications

• “Acoustic Feedback for Closed-Loop Force Control in Robotic Grinding” by Zongyuan Zhang, Christopher Lehnert, Will N. Browne, Jonathan M. Roberts at 2026 IEEE International Conference on Robotics and Automation (ICRA), Vienna, Austria.