Line search algorithms with guaranteed sufficient decrease

1994 ◽  
Vol 20 (3) ◽  
pp. 286-307 ◽  
Author(s):  
Jorge J. Moré ◽  
David J. Thuente
2009 ◽  
Author(s):  
M. Fernanda P. Costa ◽  
Edite M. G. P. Fernandes ◽  
Theodore E. Simos ◽  
George Psihoyios ◽  
Ch. Tsitouras

Author(s):  
Hiroyuki Sugiyama ◽  
Yoshihiro Suda

In this investigation, contact search algorithms for the analysis of wheel/rail contact problems are discussed, and the on-line and off-line hybrid contact search method is developed for multibody railroad vehicle dynamics simulations using the elastic contact formulation. In the hybrid algorithm developed in this investigation, the off-line search that can be effectively used for the tread contact is switched to the on-line search when the contact point is jumped to the flange region. In the two-point contact scenarios encountered in curve negotiations, the on-line search is used for both tread and flange contacts to determine the two-point contact configuration. By so doing, contact points on the flange region given by the off-line tabular search are never used, but rather used as an initial estimate for the online iterative procedure for improving the numerical convergence. Furthermore, the continual on-line detection of the second point of contact is replaced with a simple table look-up. It is demonstrated by several numerical examples that include flange climb and curve negotiation scenarios that the proposed hybrid contact search algorithm can be effectively used for modeling wheel/rail contacts in the analysis of general multibody railroad vehicle dynamics.


1997 ◽  
Vol 6 (2) ◽  
pp. 205-229 ◽  
Author(s):  
L. PRONZATO ◽  
H. P. WYNN ◽  
A. A. ZHIGLJAVSKY

Certain convergent search algorithms can be turned into chaotic dynamic systems by renormalisation back to a standard region at each iteration. This allows the machinery of ergodic theory to be used for a new probabilistic analysis of their behaviour. Rates of convergence can be redefined in terms of various entropies and ergodic characteristics (Kolmogorov and Rényi entropies and Lyapunov exponent). A special class of line-search algorithms, which contains the Golden-Section algorithm, is studied in detail. Their associated dynamic systems exhibit a Markov partition property, from which invariant measures and ergodic characteristics can be computed. A case is made that the Rényi entropy is the most appropriate convergence criterion in this environment.


VLSI Design ◽  
1999 ◽  
Vol 9 (1) ◽  
pp. 91-104 ◽  
Author(s):  
Joon Shik Lim ◽  
S. Sitharama Iyengar ◽  
Si-Qing Zheng

This paper presents new heuristic search algorithms for searching combined rectilinear (L1) and link metric shortest paths in the presence of orthogonal obstacles. The Guided Minimum Detour (GMD) algorithm for L1 metric combines the best features of mazerunning algorithms and line-search algorithms. The Line-by-Line Guided Minimum Detour (LGMD) algorithm for L1 metric is a modification of the GMD algorithm that improves on efficiency using line-by-line extensions. Our GMD and LGMD algorithms always find a rectilinear shortest path using the guided A* search method without constructing a connection graph that contains shortest paths. The GMD and the LGMD algorithms can be implemented in O(m+eloge+NlogN) and O(eloge+NlogN) time, respectively, and O(e+N) space, where m is the total number of searched nodes, e is the number of boundary sides of obstacles, and N is the total number of searched line segments. Based on the LGMD algorithm, we consider not only the problems of finding a link metric shortest path in terms of the number of bends, but also the combined L1 metric and link metric shortest path in terms of the length and the number of bends.


Sign in / Sign up

Export Citation Format

Share Document