Tuesday, November 4
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
15:00 - 16:00
Perception and Planning for Dual-arm Robotic Manipulation of Multi-Deformable Linear Objects in Wire Harness Assembly
Pablo Malvido Fresnillo
Postdoctoral Researcher at Tampere University (Finland)
Abstract
Deformable Linear Objects (DLOs), such as cables and ropes, are very common in both industry and everyday life, yet their robotic manipulation remains a major challenge. Unlike rigid objects, DLOs can bend, twist, overlap, and become entangled, which makes their perception and control significantly more complex. These challenges are amplified in scenarios involving multiple DLOs (MDLOs), where close adjacency and intricate arrangements further complicate manipulation.
This talk presents advances in robotic perception and planning aimed at addressing these difficulties. On the perception side, a vision-based system is introduced to estimate the shape of MDLOs under challenging conditions, including occlusions, constant overlaps, and cluttered backgrounds. On the planning side, a set of dual-arm motion coordination functions is developed, enabling robots to manipulate DLOs by synchronizing their movements according to different policies. This allows precise shape control without relying on external supports.
Additionally, to bridge the gap between advanced methods and practical usability, two complementary technologies are presented: a task-level programming by demonstration (PbD) framework that captures and digitizes human manipulation knowledge, and an intuitive graphical user interface (GUI) for controlling and monitoring ROS-based robotic systems. The integration of these components into a unified robotic platform is demonstrated through wire harness assembly, showcasing both the effectiveness of the proposed solutions and their potential for broader applications.
Biography
Pablo Malvido Fresnillo is a postdoctoral researcher in the FAST-Lab research group at Tampere University, Finland. He holds a Ph.D. (2025) in Automation Science and Engineering from Tampere University. He received his B.Sc. degree (2017) in Industrial Engineering and his M.Sc. degree (2019) in Electronics and Automation Engineering, both from the University of Vigo, Spain. His research interest includes robotic manipulation of deformable linear objects (DLOs), computer vision, dual-arm manipulation, and programming by demonstration. Since 2020, he has been involved in three European research projects—REMODEL, AGILEHAND, and AGRIMATE—focused on developing and implementing hardware and software technologies for robotic manipulation of deformable objects.
16:00 - 17:00
Data-driven manipulation and control of deformable objects
Ignacio Cuiral Zueco
Researcher, University of Zaragoza (Spain)
Abstract
Reliably handling deformable objects is a critical challenge for robots in fields like agriculture, food handling, manufacturing, healthcare, logistics, etc. My research focuses on developing scalable, data-efficient methods for deformable object manipulation with a focus on controlling the shape of deformable objects. I work on perception systems that capture deformations, methods to represent and compare object shapes, and control strategies that guide objects toward desired configurations. By combining data-driven strategies and models with analytical tools, I seek both practical performance and stability guarantees. My current research interests are now focused on enhancing the robustness of deformable object manipulation in dynamic, unstructured environments. This involves addressing the inherent limitations of vision-based strategies by exploring the integration of multi-modal feedback from force control, tactile sensing, and proprioception.
Biography
Ignacio Cuiral-Zueco is a robotics researcher specializing in the manipulation and shape control of deformable objects, developing data-driven methods for modeling, perception, and control. He earned his PhD in Systems Engineering and Computer Science (Cum Laude) from the University of Zaragoza, with a thesis recognized as a finalist for the Georges Giralt PhD Award and, as part of this distinction, presented at the European Robotics Forum 2025 in Stuttgart.
Thursday, November 6
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
10:00 - 11:00
Hierarchical Bio-Inspired Learning for Legged Locomotion
Guillaume Bellegarda
Postdoctoral Researcher at École Polytechnique Fédérale de Lausanne (Switzerland)
Abstract
The ability to efficiently move in complex environments is a fundamental property both for animals and for robots, and the problem of locomotion and movement control is an area in which neuroscience, biomechanics, and robotics can fruitfully interact. Bio-inspired robots and numerical models can be used to explore the interplay of the four main components underlying animal locomotion, namely central pattern generators (CPGs), reflexes, descending modulation, and the musculoskeletal system. After briefly reviewing different models for animals, I will present our recent work on integrating deep reinforcement learning with CPGs to study this interplay for quadruped locomotion.
Biography
Guillaume Bellegarda is a postdoctoral researcher in the Institute of Mechanical Engineering at École Polytechnique Fédérale de Lausanne (EPFL). He was previously a postdoctoral researcher at University of Southern California, and received his Ph.D. and M.S. degrees in Electrical and Computer Engineering from University of California, Santa Barbara, and his B.S. degree in Electrical Engineering and Computer Science from University of California, Berkeley. His research draws inspiration from machine learning, model-based control, and neuroscience to maximize explainable performance for dynamic robotic systems, as well as to deepen understanding of their biological system counterparts to adapt to real world situations.
11:00 - 12:00
Scaling Robot Skill Acquisition through Prior Knowledge
Quantao Yang
Postdoctoral Researcher, KTH Royal Institute of Technology (Sweden)
Abstract
Robots are increasingly expected to learn diverse skills efficiently to perform complex tasks in real-world environments. However, scaling robot skill learning often faces challenges of data inefficiency, limited generalization, and safety. In this talk, I will show how prior knowledge—ranging from structured priors in reinforcement learning to semantic cues from large-scale video and vision-language-action models—can be used to accelerate learning and improve generalization. I will present recent works in physics-prior regularized reinforcement learning, category-level skill transfer, and semantic action flow for video-driven policy learning. Together, these methods highlight how embedding prior knowledge into learning frameworks enables adaptable, and data-efficient robot skill acquisition for manipulation
Biography
Quantao Yang is a postdoctoral researcher at the Robot Perception and Learning (RPL) group, KTH Royal Institute of Technology. Previously, he was a research scientist at ABB Corporate Research in Sweden. He earned his PhD in computer science from Örebro University and was a visiting PhD scholar at UT Austin. His long-term research goal is to enable robots to learn efficiently and generalize as humans, with a focus on robot manipulation through reinforcement learning and imitation learning. He is passionate about developing intelligent robotic agents capable of performing tasks across diverse real-world environments.
Tuesday, November 11
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
15:00 - 16:00
From Ego to Eco-System: Perception & Fusion for Autonomous and Connected Mobility
Simone Mentasti
Researcher at Politecnico di Milano (Italy)
Abstract
Autonomy succeeds when perception is both timely and shareable. In this talk I trace our journey from an ego-vehicle stack to a infrastructure-based shared perception systems.
We begin with the context, our autonomous platform, sensing suite, and constraints, and show how vision/LiDAR/radars fusion underpins three pillars:
- lane and road-marking detection that remains stable under wear, rain, and glare
- state estimation lavaraging sensor fusion in GNSS-denied scenarios
- obstacle detection & multi-object tracking with robust association from lidars cameras and radars
We then widen the lens to mapping as the backbone: high-fidelity maps for localization, but also dataset generation and digital twins. I’ll detail how occupancy-level fusion and static-vs-dynamic segmentation feed repeatable evaluation loops, and how the digital twin (CARLA) lets us stress-test synchronization, calibration drift, occlusions, and bandwidth limits in safety-critical edge cases.
Building on that, we step into V2X cooperative perception: exchanging tracklets or occupancy summaries rather than raw data, aligning agents in space and time, and fusing confidence to see beyond line-of-sight. Finally, we land in the real world with infrastructure-side perception, roadside cameras/LiDAR for early hazard discovery, near-miss analysis, and increased road safety.
Biography
Simone Mentasti is a Researcher at the Artificial Intelligence and Robotics Laboratory (AIRLab), DEIB, Politecnico di Milano. He earned his M.S. in Computer Science from Università Statale di Milano in 2017 and his Ph.D. in Information Technology from Politecnico di Milano in 2022. His work centers on sensor fusion and obstacle detection and tracking, with applications ranging from autonomous vehicles and indoor robotics to agriculture.
In addition to his research at AIRLab, Simone collaborates with Politecnico’s Mechanical Engineering Department on urban mobility and self-driving car projects. Since 2022, he has also been working in the Smart Eyewear Lab, a Joint Research Center between EssilorLuxottica and Politecnico di Milano where he contributes to the design and development of next-generation smart eyewear technologies.
16:00 - 17:00
Online robot learning in hardware: Challenges and opportunities
Thomas Berrueta
Postdoctoral Researcher at CALTECH (USA)
Abstract
Online learning is essential for robot autonomy, but hardware deployment presents major hurdles in data efficiency, model fidelity, and operational safety. This talk provides a direct overview of my work in this domain, focusing on safety-critical systems. First, I tackle the data acquisition problem, showing that effective learning from scratch demands direct and robust state exploration. Next, I address the sim-to-real gap with a framework that leverages bandit learning techniques to directly optimize true closed-loop system performance. To manage the risks of online adaptation in high-speed hardware, I then introduce provably stable and smooth learning methods for system components, such as sensor noise, that ensure downstream controllers remain stable. Finally, I unify these concepts by exploring the duality of “planning to learn” and “learning to plan,” presenting active learning frameworks that generate safe, informative actions and hybrid architectures that guarantee recoverability. This talk synthesizes these results to offer a clear perspective on building intelligent robots that learn safely and effectively on hardware in real-time.
Biography
Thomas Berrueta is an interdisciplinary roboticist and Postdoctoral Scholar at the California Institute of Technology’s Computing and Mathematical Sciences Department. Since joining Prof. Soon-Jo Chung’s Autonomous Robotics and Control Laboratory, Thomas has led algorithm design and development for high-performance robotic hardware platforms, such as spacecraft and autonomous race cars traveling at over 250kph. He received a Ph.D. in Mechanical Engineering from Northwestern University under the supervision of Prof. Todd Murphey, where he was awarded the Presidential Fellowship—the highest honor conferred to a graduate student by the university—and named a Microsoft Future Leader in Robotics and AI. Moreover, his work has been featured in coverage from news outlets like Ars Technica, Popular Mechanics, Scientific American, Gizmodo, and Science. Thomas Berrueta’s research explores the role of embodiment in robot learning and control, seeking to make autonomous systems more adaptable, robust, and life-like by integrating physical self-awareness with techniques from reinforcement learning, stochastic processes, and optimal control.
Thursday, November 13
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
10:00 - 11:00
Dynamic manipulation of deformable objects, past work and challenges
Adriá Colomé
Research Scientist at the Institut de Robòtica i Informàtica Industrial (Spain)
Abstract
Robotic manipulation of deformable objects, in particular clothes, presents several challenges for the scientific community. This kind of manipulation has been mostly done in an offline manner: the robot observes, plans, and executes such a plan with a controller. However, humans tend to do things differently: we observe, plan and then run a control loop that combines observation, control, and our expectation of the cloth’s dynamics. This is the reason why robots still struggle to perform fast actions in cloth manipulation: they often do not consider dynamics, in any of these meanings: dynamic in terms of fast motion, dynamic in terms of adaptive, or dynamic in terms of physics-awareness. In this talk, I will discuss our previous work trying to bridge the gap between humans and robots in terms of dynamic cloth manipulation, and point towards current and future challenges.
Biography
Adrià Colomé is a research scientist in Barcelona, specializing in robotics. He earned two bachelor degrees in Mathematics and Industrial Engineering in 2009, followed by a M.Sc. (2011) and a Ph.D. (2017) in Robotics and Automatic Control from the Technical University of Catalonia (UPC). His research interests span a range of robotics topics, including robot kinematics and dynamics, robot motion learning, variable impedance control, and the manipulation of deformable objects. Dr. Colomé has published several papers and contributed to various projects in these areas.
Monday, November 24
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
09:00 - 10:00
Feedback Models for Robotic Cloth Manipulation
Oriol Barbany
Monday, December 1
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
10:00 - 11:00
In-Hand Slip-Aware Object Manipulation
Gabriel Arslan Waltersson
PhD Student at Chalmers University of Technology (Sweden)
Abstract
Traditional robotic manipulation often assumes a rigid grasp, where an object remains fixed within the gripper during motion. But what if this assumption were relaxed, allowing controlled slippage within the grasp? Such slip-aware manipulation unlocks new possibilities—from safely handling delicate items, to expanding the workspace of a manipulator, to enabling more efficient and dexterous motions. Despite its promise, this area remains largely unexplored, though recent advancements in tactile sensing and actuation have opened up new opportunities to investigate slip-aware manipulation.
In this seminar, we present our approach to slip-aware manipulation using simple parallel grippers. Although these grippers offer only a single degree of freedom (open/close), their capabilities can be significantly enhanced through tactile perception, fast actuation, and accurate contact modeling. We will discuss how hardware design, sensing, control algorithms, learning, and friction modeling can be integrated into an architecture that enables in-hand manipulation without relying on external sensing.
Biography
Gabriel Arslan Waltersson is a PhD student at Chalmers University, researching in-hand slip-aware object manipulation with robotic grippers. His work spans modeling and estimation of friction and contact dynamics, hardware and sensor design, and perception and control. The goal of his research is to extend the capabilities of parallel robotic grippers by relaxing the rigid-grasp assumption and enabling controlled slippage between the hand and the object.
11:00 - 12:00
Scalable and Generalizable Robot Learning: from Simulated to Real-World Applications
Gabriele Tiboni
Postdoctoral Researcher at JMU Würzburg and Pearl Lab at TU Darmstadt (Germany)
Abstract
Directly applying reinforcement learning algorithms to real-world robots poses safety concerns and challenges in terms of data inefficiency. Physics simulators, on the other hand, allow for cheap data collection and eliminate the need for human supervision during training. We present the field of simulation-to-real transfer for robot learning, aiming to safely train reinforcement learning policies in simulation that can be directly transferred to real-world robots. To this end, techniques such as domain randomization will be described in detail as a means for robust and adaptive transfer of the reality gap. Particularly, we will demonstrate how policies can be trained to generalize across physical parameters via curriculum learning approaches, and how real-world data can be further integrated to facilitate the sim-to-real transfer.
Biography
Gabriele Tiboni received his M.Sc. degree in Data Science and Engineering from Politecnico di Torino, Italy, in collaboration with Aalto University, Finland, in 2021, where he completed his Master thesis on safe and efficient simulation-to-real transfer of robot behavior. He subsequently pursued the National Ph.D. AI Programme at Politecnico di Torino focusing on generalizable and scalable Robot Learning algorithms. His research includes 3D learning for complex motion generation via imitation, and generalization across real-world environment dynamics for robot manipulation via Reinforcement Learning. During his doctoral studies, he enrolled in the ELLIS PhD&Post-Doc Programme and the ELIZA School of Excellence. He is currently a Postdoctoral Researcher jointly in the LiteRL Group at JMU Würzburg, and Pearl Lab at TU Darmstadt. His research vision targets the development of intelligent robotic systems capable of robust and adaptive generalization to new conditions.
Tuesday, December 2
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
15:00 - 16:00
Achieving Coordination with Multiple Heterogeneous Agents
Martina Lippi
Assistant Professor, Roma Tre University (Italy)
Abstract
Deploying heterogeneous human-multi-robot teams in complex environments poses several challenges related to coordination and adaptability under uncertain and dynamic conditions. This talk explores frameworks that combine optimization-based formulations with adaptive planning strategies to address diverse agent capabilities, uncertainties, and variable human factors. The approaches provide coordination of the multi-agent system, robustness to dynamic conditions, and mechanisms for incorporating human feedback into the decision-making process. Validations in realistic simulated and laboratory settings demonstrate the applicability of these methods to heterogeneous multi-agent collaboration.
Biography
Martina Lippi received the M.Sc. (cum laude) and Ph.D. degrees in Information Engineering from the University of Salerno, Italy, in 2017 and 2020, respectively. In 2019, she was a Visiting Scholar with the KTH Royal Institute of Technology, Sweden. From November 2020 to June 2022, she was a Postdoctoral Researcher with Roma Tre University, Italy, where she has been an Assistant Professor since June 2022. Her research interests include multi-robot systems, human–robot interaction, distributed control and agricultural robotics.
16:00 - 17:00
Adaptive Sensor Fusion for Autonomous Driving in Harsh Weather Conditions
Thu Ngo
Postdoctoral Researcher, Halmstad University (Sweden)
Abstract
In autonomous driving, the perception system plays a crucial role in enabling vehicles to perceive and interpret their surroundings. This comprises a suite of sensors, including a camera, lidar, radar, and an advanced algorithm to identify and track objects such as pedestrians, vehicles, road signs, and traffic lights. However, adverse weather conditions such as rain, snow, fog, and low-light environments can significantly degrade the sensor performance and challenge the reliability of the perception model. This seminar presents the recent advances in adaptive sensor fusion techniques to enhance the robustness and safety in such challenging conditions. We discuss dynamic fusion strategies across data, feature, and decision levels, including sensor weighting, multimodal learning, and domain adaptive fusion. In addition, we introduce the proposed datasets tailored for adverse weather conditions. Case studies from recent research illustrate how adaptive fusion techniques improve object detection, semantic segmentation, and multi-task learning under weather-induced disturbances. Finally, the seminar highlights emerging research directions and practical considerations for real-world deployment of adaptive fusion systems in autonomous vehicles.
Biography
Ngo Thien Thu is a Postdoctoral Researcher at the School of Information Technology, Halmstad University, Sweden. She works on the ROADVIEW project, focusing on adaptive sensor fusion algorithms for autonomous driving in harsh weather condition. Her research interests include deep learning, computer visionand their applications in real-time environment.
Thursday, December 4
Seminar Room M:3170-73 Dept. of Automatic Control, M-building, LTH
13:00 - 14:00
Safe and Efficient Structured Learning for Robotics
Raffaello Camoriano
Assistant Professor at Politecnico di Torino, Affiliated Researcher with the Istituto Italiano di Tecnologia (Italy)
Abstract
Recent progresses in machine learning targeting longstanding perception and control problems are propelling the field forward by demonstrating what is achieavable, yet with virtually unlimited data, energy, computations, memory, and labelling resources. Nonetheless, robotics imposes a variety of resource budgets on learning systems, thus requiring more flexible and efficient learning methods in terms of computational costs, often with limited expert supervision. Furthermore, robots are required to operate safely in dynamic environments, which requires the ability to learn from structured data (e.g., sequences, manifolds, etc.) and comply with constraints.
In this talk, we explore recent works for efficient and structured learning in robotics, covering how structured prediction enables imitation and model learning for robotic manipulation and humanoid robot locomotion. To conclude, I will also present recent works on incremental learning for robotic prosthetics and avenues for future research in this domain.
Biography
Raffaello Camoriano received the B.Sc. and M.Sc. degrees in computer and robotics engineering and the Ph.D. degree in bioengineering and robotics from the University of Genoa, Italy, in 2011, 2013, and 2017, respectively. From 2014 to 2022, he has been a Research Fellow and a Postdoctoral Researcher with the Istituto Italiano di Tecnologia, Genoa. He is currently an Assistant Professor with the Department of Control and Computer Engineering, Politecnico di Torino, Turin, Italy, and an Affiliated Researcher with the Istituto Italiano di Tecnologia. He is author of 25 international peer-reviewed articles in the areas of efficient and structured learning, continual learning, reinforcement learning, and robotics. Prof. Camoriano was the recipient of the IEEE Computational Intelligence Society Italy Section Chapter’s 2017 Best Ph.D. Thesis Award. He is a Member of the European Laboratory for Learning and Intelligent Systems (ELLIS). He is currently an Associate Editor for IEEE Robotics and Automation Letters, Springer Machine Learning, and IEEE/RSJ IROS, and a member of the IEEE Robotics and Automation Society’s Member Activities Board (MAB
14:00 - 15:00
Machine learning for scaled autonomous racing and beyond
Piotr Kicki
Postdoctoral Researcher at IDEAS Research Institute, Assistant professor at the Institute of Robotics and Machine Intelligence at Poznan University of Technology (Poland)
Abstract
In this talk I would like to present our recent advancements in applying machine learning to model and control dynamical systems at the edge of their physical capabilities, with a particular emphasis on scaled autonomous racing. I’ll discuss topics like identifying and predicting dynamics models for efficient state estimation, motion prediction and control, as well as the use of deep reinforcement learning and model predictive control for beating the humans in RC racing.
Biography
Piotr Kicki is a post-doc in the robotics team at IDEAS Research Institute and an assistant professor at the Institute of Robotics and Machine Intelligence at Poznan University of Technology. He received his B.Eng. and M.Sc. degrees in automatic control and robotics from Poznan University of Technology, Poland in 2018 and 2019, respectively. He completed his Ph.D. from the same university in 2024. His main research interests focus on robot learning with a special emphasis on motion planning and control.
