In this talk I will review work on the analysis of motion capturing data and similar applications using techniques of shape analysis and deep learning. I will then consider a method for learning the Lagrangian and forces for mechanical systems using the discrete Lagrange d'Alembert principle. The case of manifold valued data and data on Lie groups will also be discussed if time permits. Applications to mechanical system will be considered.
Elena Celledoni is a professor at the Department of Mathematical Sciences at the Norwegian University of Science and Technology. She received her Master from the University of Trieste, and her Ph.D from the University of Padua, Italy. She held post doc positions at the University of Cambridge, UK, at the Mathematical Sciences Research Institute, Berkeley, California and at NTNU. Her research field is in numerical analysis and in particular structure preserving algorithms for differential equations and geometric numerical integration. She is a member of the editorial board of SIAM Review, SISC and Math. Comp.
The optimization of multibody systems atracted some interest from the early starts of the discipline but the possibilities and the number of works published shot up in the last decade. Nowadays, the dynamics of multibody systems is satisfactorily solved by many codes in a general fashion, using different coordinates and formulations of the equations of motion. For this reason, the focus was moved to more complex problems like the desgin optimization and optimal control of multibody systems. This talk focus on the optimization of the dynamics of multibody systems, considering both rigid and flexible bodies and its application to different types of problems, some of them classical or academic and some industrial applications.
For optimization of simple systems based on the dynamic response with few parameters, gradient-free optimization methods can be acceptable, or even convenient, but for large systems and/or large number of parameters the so-called global optimization methods render prohibitive in terms of computational time and gradient based methods arise as the best solution to the problem.
Sensitivity analysis of MBS is the discipline needed for calculating gradients of objective functions dependent on the dynamic response of the system and it adds great complexity on top to the regular dynamics simulations, specially for flexible multibody systems. Different techniques available for sensitivity analysis will be visited in this talk, as well as techniques to calculate the derivatives of the equations of motion terms involved in the sensitivity equations
Daniel Dopico became Associate Professor at University of La Coruña (Spain) in 2010 focused on real-time formulations for the dynamics of multibody systems and he was research scientist at Virginia Tech (USA) in 2012 and 2013 where his interest on sensitivity analysis and optimization of multibody systems started. His current research interests are formulations and sensitivity-based optimization methods for rigid and flexible multibody dynamics, topic on which he has been working during the last 14 years. He is the main developer of MBSLIM: general code for the dynamics, sensitivity analysis and gradient based optimization of multibody systems.
Multibody dynamics (MBD) simulation tools have become indispensable to robotics researchers and practitioners, forming a core part of modern robot design, analysis, and control. Today, a wide range of commercial and open-source MBD simulators are available, many of which are widely adopted across diverse robotic applications. In this talk, I will begin by outlining key use-cases for multibody dynamics modeling in robotics and provide a comparative overview of leading simulation platforms. I will then illustrate how these tools have supported our research through two contrasting case studies: one focused on the dynamics of space debris capture and on-orbit servicing, and the other on the development of autonomous log-loading machinery for forestry operations. Drawing from these experiences, I will discuss several challenges encountered when using commercial simulation software, with particular attention to the modeling of contact interactions.
Dr. Inna Sharf is a professor in the Department of Mechanical Engineering at McGill University, Montreal, Canada. Sharf conducts research in robotics and autonomous systems with applications to space robotics and sustainable timber harvesting. Sharf has worked on problems related to space debris remediation, on-orbit servicing and more recently, autonomy for timber-harvesting machines. Sharf has an established publication record in robotics, dynamics and control journals and conferences and has given numerous presentations on her research. She has supervised over 60 PhD and Master’s level engineering students over her career. Sharf recently completed a three-year term on Robotics and Automation Society Administrative Committee. She is an associate fellow of American Institute of Aeronautics and Astronautics, a member
Multibody system dynamics (MBD) provides a fundamental framework for advancing robotic systems in many fields and many applications, yet significant challenges remain in translating theoretical models into reliable real-world performance. This keynote address explores these barriers and identifies key bridges to overcome them through recent advances in modeling, control, and cooperative strategies.
In space robotics, the complex dynamics of orbital manipulation require robust solutions to handle uncertain contact forces and disturbances. It is demonstrated that model-based reinforcement learning can effectively generalize control policies across varying contexts through the integration of projected inverse dynamics and probabilistic forward models. Efficient adaptation for cooperative manipulation tasks is addressed, while stability is maintained in unstructured environments.
For swarm robotics, the integration of particle swarm optimization with multibody dynamics offers a promising path to coordinate large-scale robotic teams. By incorporating mechanical properties such as mass and inertia into swarm guidance algorithms, robots can navigate in complex environments while maintaining cohesion. Additional refinements using extremum-seeking techniques help mitigate localization errors, allowing decentralized control without precise state feedback.
Underwater robotics faces unique challenges due to hydrodynamic forces and limited sensing. Advanced dynamic modeling based on recursive Newton-Euler formulations provides a foundation for vehicle-manipulator coordination, while disturbance observers enhance stability during manipulation tasks. Further improvements in trajectory tracking have been achieved through nonlinear sliding mode control, which suppresses chattering while compensating for time-varying disturbances.
Across these domains, a common theme emerges: the need to balance model-based precision with adaptive, real-time responsiveness. Future progress will depend on tighter integration between MBD, distributed control, and learning-based strategies—bridging the gap between theoretical frameworks and practical robotic deployment in challenging environments.
Qirong Tang received his doctoral degree (Dr.-Ing.) in 2012 from University of Stuttgart, Germany. From 2012 to 2014, he was a senior research associate at the Institute of Engineering and Computational Mechanics, University of Stuttgart. He is currently a full professor (with distinguish) and the founding director of Laboratory of Robotics and Multibody System, as well as the leader of the Intelligent Unmanned Systems Group. Dr. Tang serves as the deputy director of Sino-German School and the director of Sino-German Doctoral School. He is also the director of Discipline of Mechanical Engineering, Chairman of Academic Committee of School of Mechanical Engineering, at Tongji University, Shanghai, China. Dr. Tang and his group's research interests include swarm robotics, multibody system dynamics, robotic manipulator and so on.
Contact by email: imsd2026@us.es
© Copyright IMSD, ES