Scuola di dottorato SIDRA 2016
|Coordinator:||Alessandro Macchelli (University of Bologna)|
|Cristian Secchi (University of Modena and Reggio)|
|Confirmed Speakers:||Alessandro Macchelli (University of Bologna)
|Bernhard Maschke (University Claude Bernard of Lyon)|
|Cristian Secchi (University of Modena and Reggio)|
|Stefano Stramigioli (University of Twente)|
An emerging trend in control and robotic research seems to be the integration with different areas such as neuro-science, artificial intelligence or cognition. Nevertheless, it is still important to present basic and general methods, techniques and tools for the modeling, control and simulation of physical systems (e.g., mechanical, electro-mechanical and thermodynamical systems). This is even more fundamental for young researchers and scientists, which should have a solid knowledge and background of general approaches for dealing with multidisciplinar and complex systems. Examples in robotics of such systems are devices for advanced manipulation (dextrous hands), telemanipulation and haptic systems, cooperating robots and legged robots. These are open systems, i.e. devices that may be generically described as a system with a direct physical interface with its environment. It is important to note that the notion of open system naturally arises also in all the other energy domains, where an interaction among physical systems is present, domains which may seem very different from the electro-mechanical one.
It turns out that elegant and general mathematical tools for modelling and controlling interacting physical systems (also coming from different domains) exist, and that their use can be very helpful in solving complex problems, where other approaches may lead to more heuristic or confuse solutions.
These methodological tools are framed in the Hamiltonian formalism. In particular, in order to describe and to manipulate these dynamical models in a systematic way, it is convenient to use a coordinatefree, geometric framework for their mathematical formulation, especially because of the intrinsic and strong nonlinearities in their system behavior. The framework of port-Hamiltonian systems, where the physical components are formulated as generalised Hamiltonian systems either in the lumped and in the distributed parameter case, and coupled to each other through power ports, will be presented in this Summer School. In this context, the resulting complex physical system can be geometrically described as a Hamiltonian system with respect to the geometric object of a Dirac structure.
Main goal of this Summer School is to present methods, techniques and tools for modelling and controlling complex interacting dynamical systems, using an integrated system approach allowing to deal with physical components stemming from robotics and also from different domains (e.g. electrical, mechanical, thermodynamic), both in the lumped-parameter and in the distributed-parameter case. To conclude, a general aim of this initiative is to explain in a clear and comprehensive manner the foundations of port-Hamiltonian system theory and of passivity based control. Concepts at the moment spread in a number of journal and conference papers are then illustrated in a compact and integrated form. This will allow to the audience to get the basic concepts and, if interested, to rapidly being able to access the state-of-the-art in this field.
Teaching aids and instructional materials
According to the spirit of the latest editions, the teaching style of the lectures will be a traditional one, making use as much as possible of the blackboard for derivations, complemented as appropriate by videos, simulations, and diagrams. The speakers will provide a collection of handouts as material supporting the lectures. This material will be available at least one week before the start of the School.
The PhD School will be taught in English.
The course is organised into five main topics, briefly presented below. Each argument requires an half-day class.
Passivity and passivity-based control. This initial part aims at illustrating the basic ideas and results of passivity-based control. Once the intuition behind the concept of passivity/dissipativity is discussed together with the main definitions, the relation between passivity-based control and Lyapunov stability is discussed. Then, stabilisation by output feedback and via interconnection (e.g., small-gain) is presented.
Port-Hamiltonian modelling. The class of port-Hamiltonian systems, that are a particular type of passive systems, is presented starting from the basic concepts on physical modelling, i.e. on bond-graphs. The key geometric features (e.g., invariants and Dirac structures) are discussed in the lumped paramenter case, and the extension of the classical port-Hamiltonian system theory towards the description of irreversible processes is discussed. This latter point leads to the definition and characterisation of the class of irreversible port-Hamiltonian systems.
Control of port-Hamiltonian systems. This part of the course is devoted to control synthesis
within the port-Hamiltonian framework. The starting point is the general theory of passivity based control, that leads to simple control schemes based on damping injection and on energy balancing. More powerful control schemes that are able to exploit the port-Hamiltonian structure to obtain stability in closed-loop are also discussed. Among them, main emphasis is given the control techniques that rely on generalised canonical transformations, and on interconnection and damping assignment passivity-based control (IDA-PBC). Finally, basic results on the control of irreversible port-Hamiltonian systems are illustrated.
Infinite dimensional port-Hamiltonian systems. This topic deals with the extension of the
port-Hamiltonian formalism presented so far to the distributed parameter case, i.e. to systems described by PDEs. The classical formulation of an infinite dimensional port-Hamiltonian system derived from conservation laws is discussed and, for a class of system with one-dimensional domain, the key results on boundary control are illustrated. Finally, a brief overview on spatial discretisation techniques within the port-Hamiltonian framework are also presented.
Port-Hamiltonian systems in robotics. In this conclusive part of the course, the applications of
port-Hamiltonian modelling and control to the field of robotics is presented. In particular, the use of port-Hamiltonian modeling and control for the control of interaction of robotic systems (both fixed and aerial robots) will be discussed. The concept of energy tank is introduced, and its use in teleoperation of swarms of aerial vehicles is described. Finally, an overview of the future directions of port-Hamiltonian systems in robotics will is provided.
|Day 1 - Thursday, July 6, 2017|
|9:00 – 10:30||Welcome - Introduction to Physical Modelling P1||Macchelli|
|11:00 – 12:30||Introduction to Physical Modelling P2||Macchelli|
|15:00 – 16:30||Energy Aware Robotics||Stramigioli|
|17:00 – 18:30||Energy-based Control of Port-Hamiltonian Systems||Maschke|
|Day 2 – Friday, July 7, 2017|
|9:00 – 10:30||Irreversible Port-Hamiltonian Systems. Modelling and Control P1||Maschke|
|11:00 – 12:30||Irreversible Port-Hamiltonian Systems. Modelling and Control P2||Maschke|
|15:00 – 16:30||Passivity and Port-Hamiltonian Based Control for Advanced Robotics P1||Secchi|
|17:00 – 18:30||Passivity and Port-Hamiltonian Based Control for Advanced Robotics P2||Secchi|
|Day 3 – Saturday, July 8, 2017|
|9:00 – 10:30||Modelling and Control of Distribute Port-Hamiltonian Systems P1||Macchelli|
|11:00 – 12:30||Modelling and Control of Distribute Port-Hamiltonian Systems P2||Macchelli|
Here it is possible to download the program (pdf version).