The support vector machine (SVM) is a method for classification and for function approximation. This method commonly makes use of an /spl epsi/-insensitive cost function, meaning that errors smaller than /spl epsi/ remain unpunished. As an alternative, a least squares support vector machine (LSSVM) uses a quadratic cost function. When the LSSVM method is used for function approximation, a nonsparse solution is obtained. The sparseness is imposed by pruning, i.e., recursively solving the approximation problem and subsequently omitting data that has a small error in the previous pass. However, omitting data with a small approximation error in the previous pass does not reliably predict what the error will be after the sample has been omitted. In this paper, a procedure is introduced that selects from a data set the training sample that will introduce the smallest approximation error when it will be omitted. It is shown that this pruning scheme outperforms the standard one.

The design and realization of an on-line learning motion controller for a linear motor is presented, and its usefulness is evaluated. The controller consists of two components: 1) a modelbased feedback component and 2) a learning feedforward component. The feedback component is designed on basis of a simple second order linear model, which is known to have structural errors. In the design, emphasis is placed on robustness. The learning feedforward component is a neural-network-based controller, comprised of a onehidden-layer structure with second-order B-spline basis functions. Simulations and experimental evaluations show that, with little effort, a high-performance motion system can be obtained with this approach.

Abstract-Different branches of technology are striving to come up with new advancements that will enhance civilization and ultimately improve the way of life. In the robotics community, a stride has been made to bring the use of personal robots in office and home environments on the horizon. Safety is one of the critical issues that must be guaranteed for successful acceptance, deployment and utilization of domestic robots. Unlike the barrier based operational safety guarantee that is widely used in industrial robotics, safety in domestic robotics deals with a number of issues such as intrinsic safety, collision avoidance, human detection and advanced control techniques. In the last decade, a number of researchers have presented their works that highlighted the issue of safety in a specific part of the complete domestic robotics system. This paper presents a general survey of various safety related publications that focus on safety criteria & metrics, mechanical design & actuation and controller design.

This paper presents a general passivity based interaction controller design approach that utilizes a combined energy and power based safety norms to assert safety of domestic robots. Since these robots are expected to cohabit the same environment with a human user, analysing and ensuring their safety is an important requirement. Safety analysis of domestic robots determine whether a robot achieves a desired safety level according to some quantitative safety metrics. When it comes to controller design for human friendly robots, it often involves introducing compliance and ensuring asymptotic stability using impedance control technique and passivity theories. The controller proposed in this work also uses a passive design that extends the standard impedance control scheme with energy and power based safety metrics to ensure that safety requirements defined in these norms are achieved by domestic robots. The effectiveness of the proposed guideline is illustrated with simulation and experimental results.

Abstract-Robust stability of controlled mechanical systems is often obtained using collocated actuator-sensor-pairs. Collocation enables the implementation of a passive control law, which is robustly stable, irrespective of structural modeling errors. Within the context of vibration control, this knowledge is used to obtain robust active damping. However, collocated control is inherently in terms of "local" coordinates, whereas vibration analysis is usually in terms of "modal" coordinates. Therefore, modal decoupling of the collocated control loops is required. It is shown that, under mild conditions, transformation of the control problem from local into modal coordinates yields control loops that again enable the implementation of passive and thus robustly stable control laws. The presented theory is illustrated by means of experiments on the six-degrees-of-freedom (DOF) actively controlled lens suspension within a micro-lithography machine.