|The interaction of a robot with its surrounding environment or with a human operator takes place by means of power exchanges. In both cases, the primary requirement for a control algorithm is to guarantee the stability of the interaction regardless of the uncertainties due to a partial knowledge of the environment.Robust stability can be obtained if controllers and the controlled systems behave as physical
passive systems , as stated by Hogan with
the principle of physical equivalence:
"A unified approach to both the design of the controller and the physical hardware can be developed by postulating that, taken together, the hardware (e.g. sensors, actuators, and structures) and software (e.g. neural networks, brains, or computers) is still a physical system in the same sense that the hardware alone is."
This principle implies that physical modeling techniques can be used in order to describe the behavior of interaction controllers. Moreover, since physical systems are passive in the sense that only a finite amount can be extracted from them, the passivity formalism provides a set of powerful tools to analyze the interconnection of passive systems. Therefore the analysis of
complex systems for remote and virtual interaction can be carried out by following an approach based on the interconnection of simple physical elements, according to the philosophy of bond graphs and port Hamiltonian systems.
This ideas inspired the research activity that is presented in the remainder of this thesis. In particular several aspects of teleoperation and haptic systems have been analyzed by means of the
fundamental concepts of energy and passivity:
- After a brief overview of teleoperation and haptic systems in chapter 1,
- chapter 2 presents a theoretical and experimental analysis of classical control algorithms for time-delayed teleoperation showing that stability is not an intrinsic feature. Therefore small perturbations on the operating conditions can cause unstable behaviors. On the other hand, physically consistent control strategies allow to obtain robust stability.
- The same concepts are applied in chapter
3 to a non-conventional teleoperation system where a haptic device is used to command the motion of a mobile platform. In particular, this application shows how the navigation among unknown obstacles can be improved by means of force feedback. The nonholonomic constraint affecting the kinematics of the mobile robot is handled by means of an intrinsically passive control strategy.
- On the other hand, the physically based control strategies tend to introduce spurious dynamics that affect user perception of the remote or virtual environment. The knowledge of contact dynamics could be then used to improve the transparency of the overall system. In particular, chapter 4 shows how the usual linear model does not provide an energetically consistent description of contact dynamics. A nonlinear model is taken into account and a parameters estimation procedure is designed in order to evaluate its suitability to describe the behavior of stiff and compliant materials.
- Finally, it should be noted that the analogy between control algorithms and their physically equivalent elements does not hold if unmodeled dynamics cannot be neglected because of the high gains requested. This is the typical situation of haptic rendering, where stiff virtual walls are rendered to the user by means of high gain controllers. A detailed study relating the maximum controller stiffness to the physical and technological parameters of a haptic device is presented in chapter 5.
- Chapter 6 reviews the main results
obtained in previous chapters and draws future research lines.