Over the last few decades, robots have become highly skilled at manipulating rigid objects, which has led to their widespread use in manufacturing, healthcare, and everyday services. However, their ability to handle soft, flexible materials – such as cables, clothing, or food – remains limited. Among these, a particularly challenging category is Deformable Linear Objects (DLOs), like cables or ropes. These objects have one dimension significantly longer than the other two. They can be found everywhere – from household devices to complex machinery in the automotive, aerospace and electronics industries – and they often prevent full automation of assembly tasks.
In his doctoral research, Pablo Malvido Fresnillo developed new methods to address this problem. These methods enable robots to "see" and plan how to handle these flexible objects.
“Unlike solid parts, cables can bend, twist, overlap, or get entangled, which makes it incredibly difficult for robots to understand and manipulate them accurately,” Malvido Fresnillo explains.
His work focuses on wire harnesses – bundles of cables used in all modern machines. These are particularly challenging to automate due to their high density of cables with often unpredictable shapes and configurations. By using advanced computer vision and deep learning techniques, the system can identify and estimate the shape of multiple cables even when they overlap or are partially hidden. It then uses this information to plan how to manipulate them with two robotic arms moving in a coordinated way.
To make the system practical for real-world use, Malvido Fresnillo also developed a user-friendly graphical interface and a programming-by-demonstration system, allowing non-experts to teach the robot new tasks by simply showing them, rather than coding them.
“This makes it easier for small and medium-sized companies to use robotics, as these companies often do not have their own programming experts," he says.
The technologies developed in his dissertation were integrated into a complete robotic system and tested on an actual wire harness assembly task. The results show that dual-arm robots can successfully perform tasks like cable separation, routing, and insertion, which have traditionally been done by human workers.
Malvido Fresnillo’s work brings robotics a step closer to tackling the manipulation Deformable Linear Objects, which is a growing concern not only in manufacturing but also in areas like household robotics and healthcare.
Public defence on Friday 5 September
The doctoral dissertation of M.Sc. Pablo Malvido Fresnillo in the field of Automation Engineering titled Perception and Planning for Dual-arm Robotic Manipulation of Multi-Deformable Linear Objects in Wire Harness Assembly will be publicly examined at the Faculty of Engineering and Natural Sciences at Tampere University at 13:00 on Friday 5 Sep 2025 at Hervanta campus, Rakennustalo building and auditorium room RG202 (Korkeakoulunkatu 5, Tampere).
The Opponent will be Associate Professor Yiannis Karayiannidis from Lund University. The Custos will be Professor Jose L. Martinez Lastra from the Faculty of Engineering and Natural Sciences at Tampere University.