Fig. 11. Two successive outdoor range scans, smoothed and segmented by the line mask. The line intersection points ( ) in the first scan, and inverted triangles ( ) in the second, can be used as occlusion-invariant, point features. VII. CONCLUSIONUnlike RANSAC, the presented algorithm smoothes range data, to produce a local description of features, which, in some circumstances, can be more beneficial than global descriptions for robot navigation. Also, its computational complexity is independent of the number of model outliers, and is less affected by the use of higher order geometric models.The smoothing and segmentation of range data is fundamentally different from that of image data. Structure preserving, and noise reduction algorithms in vision, use the local intensity gradient as a measure of noise. Range values are completely environment dependent, and not constant between features. Therefore, in this algorithm, the Mahalanobis distance between observed range values and their geometric-model-based predictions is used as the "measure of noise." A mask weighting function of the Mahalanobis distance was derived, which behaves as the diffusion coefficient in the anisotropic diffusion equation, often applied in vision, which guarantees that no new features are introduced with increase of scale. This mask can be applied iteratively, providing smoothing at different scales. The results demonstrated that the number of extracted features (lines or circles) converged to the true number with increase of scale, and the error between the extracted and true feature coordinates converged to a minimum. It has been shown that with increase of scale, the algorithm automatically reduces noise, only within the model-compliant regions of the range scans, yielding superior, postsmoothing, segmentation possibilities. REFERENCES[1] P. Perona and J. Malik, "Scale-space and edge detection using anisotropic diffusion," IEEE Trans.

No abstract

This paper addresses the problem of steady-state position and force tracking in bilateral teleoperation. Passivity-based control schemes for bilateral teleoperation provide robust stability against network delays in the feedback loop and velocity tracking, but do not guarantee steadystate position and force tracking in general. Position drift due to data loss and offset of initial conditions is a well-known problem in such systems. In this paper, we introduce a new architecture, which builds upon the traditional passivity-based configuration by using additional position control on both the master and slave robots, to solve the steady-state position and force-tracking problem. Lyapunov stability methods are used to establish the range of the position control gains on the master and slave sides. Experimental results using a single-degree-of-freedom master/slave system are presented, showing the performance of the resulting system.

Abstract-The problem of existence and stability of equilibria of linear systems with constant power loads is addressed in this paper. First, we correct an unfortunate mistake in our recent paper [10] pertaining to the sufficiency of the condition for existence of equilibria in multiport systems given there. Second, we give two necessary conditions for existence of equilibria. The first one is a simple linear matrix inequality hence it can be easily verified with existing software. Third, we prove that the latter condition is also sufficient if a set defined by the problem data is convex, which is the case for single and two-port systems. Finally, sufficient conditions for stability and instability for a given equilibrium point are given. The results are illustrated with two benchmark examples.

We address the problem of global Lyapunov stability of discrete-time recurrent neural networks (RNNs) in the unforced (unperturbed) setting. It is assumed that network weights are fixed to some values, for example, those attained after training. Based on classical results of the theory of absolute stability, we propose a new approach for the stability analysis of RNNs with sector-type monotone nonlinearities and nonzero biases. We devise a simple state-space transformation to convert the original RNN equations to a form suitable for our stability analysis. We then present appropriate linear matrix inequalities (LMIs) to be solved to determine whether the system under study is globally exponentially stable. Unlike previous treatments, our approach readily permits one to account for non-zero biases usually present in RNNs for improved approximation capabilities. We show how recent results of others on the stability analysis of RNNs can be interpreted as special cases within our approach. We illustrate how to use our approach with examples. Though illustrated on the stability analysis of recurrent multilayer perceptrons, the approach proposed can also be applied to other forms of time-lagged RNNs.