Closing the gap: Improved bounds on optimal POMDP solutions

POMDP algorithms have made significant progress in recent years by allowing practitioners to find good solutions to increasingly large problems. Most approaches (including point-based and policy iteration techniques) operate by refining a lower bound of the optimal value function. Several approaches (e.g., HSVI2, SARSOP, grid-based approaches and online forward search) also refine an upper bound. However, approximating the optimal value function by an upper bound is computationally expensive and therefore tightness is often sacrificed to improve efficiency (e.g., sawtooth approximation). In this paper, we describe a new approach to efficiently compute tighter bounds by i) conducting a prioritized breadth first search over the reachable beliefs, ii) propagating upper bound improvements with an augmented POMDP and iii) using exact linear programming (instead of the sawtooth approximation) for upper bound interpolation. As a result, we can represent the bounds more compactly and significantly reduce the gap between upper and lower bounds on several benchmark problems. Copyright © 2011, Association for the Advancement of Artificial Intelligence. All rights reserved.

Closely correlating lower and upper bound plastic analysis of real slabs

An engineer assessing the load-carrying capacity of an existing reinforced concrete slab is likely to use elastic analysis to check the load at which the structure might be expected to fail in flexure or in shear. In practice, many reinforced concrete slabs are highly ductile in flexure, so an elastic analysis greatly underestimates the loads at which they fail in this mode. The use of conservative elastic analysis has led engineers to incorrectly condemn many slabs and therefore to specify unnecessary and wasteful flexural strengthening or replacement. The lower bound theorem is based on the same principles as the upper bound theorem used in yield line analysis, but any solution that rigorously satisfies the lower bound theorem is guaranteed to be a safe underestimate of the collapse load. Jackson presented a rigorous lower bound method that obtains very accurate results for complex real slabs.

Evolution of localized blobs of swirling or buoyant fluid with and without an ambient magnetic field.

We investigate the evolution of localized blobs of swirling or buoyant fluid in an infinite, inviscid, electrically conducting fluid. We consider the three cases of a strong imposed magnetic field, a weak imposed magnetic field, and no magnetic field. For a swirling blob in the absence of a magnetic field, we find, in line with others, that the blob bursts radially outward under the action of the centrifugal force, forming a thin annular vortex sheet. A simple model of this process predicts that the vortex sheet thins exponentially fast and that it moves radially outward with constant velocity. These predictions are verified by high-resolution numerical simulations. When an intense magnetic field is applied, this phenomenon is suppressed, with the energy and angular momentum of the blob now diffusing axially along the magnetic field lines, converting the blob into a columnar structure. For modest or weak magnetic fields, there are elements of both types of behavior, with the radial bursting dominating over axial diffusion for weak fields. However, even when the magnetic field is very weak, the flow structure is quite distinct to that of the nonmagnetic case. In particular, a small but finite magnetic field places a lower bound on the thickness of the annular vortex sheet and produces an annulus of counter-rotating fluid that surrounds the vortex core. The behavior of the buoyant blob is similar. In the absence of a magnetic field, it rapidly develops the mushroomlike shape of a thermal, with a thin vortex sheet at the top and sides of the mushroom. Again, a simple model of this process predicts that the vortex sheet at the top of the thermal thins exponentially fast and rises with constant velocity. These predictions are consistent with earlier numerical simulations. Curiously, however, it is shown that the net vertical momentum associated with the blob increases linearly in time, despite the fact that the vertical velocity at the front of the thermal is constant. As with the swirling blob, an imposed magnetic field inhibits the formation of a vortex sheet. A strong magnetic field completely suppresses the phenomenon, replacing it with an axial diffusion of momentum, while a weak magnetic field allows the sheet to form, but places a lower bound on its thickness. The magnetic field does not, however, change the net vertical momentum of the blob, which always increases linearly with time.

Appraisal of PD6493:1991 fracture assessment procedures

PD6493:1991 fracture assessment have been performed for a range of large-scale fracture mechanics tests conducted at TWI in the past. These tests cover several material groups, including pressure vessel steels, pipeline steels, stainless steels and aluminium alloys, including parent material and weldments. Ninety-two wide plate and pressure vessel tests have been assessed, following Levels 1, 2 and 3 PD6493:1991 procedures. In total, over 400 assessments have been performed, examining many features of the fracture assessment procedure including toughness input, proof testing, residual stress assumptions and stress state (tension, bending and biaxial). In all cases the large scale tests have been assessed as one would actual structures: i.e., based on lower bound toughness values obtained from small scale fracture toughness specimens.

Nearest neighbour decoding and pilot-aided channel estimation in stationary Gaussian flat-fading channels

We study the information rates of non-coherent, stationary, Gaussian, multiple-input multiple-output (MIMO) flat-fading channels that are achievable with nearest neighbour decoding and pilot-aided channel estimation. In particular, we analyse the behaviour of these achievable rates in the limit as the signal-to-noise ratio (SNR) tends to infinity. We demonstrate that nearest neighbour decoding and pilot-aided channel estimation achieves the capacity pre-logwhich is defined as the limiting ratio of the capacity to the logarithm of SNR as the SNR tends to infinityof non-coherent multiple-input single-output (MISO) flat-fading channels, and it achieves the best so far known lower bound on the capacity pre-log of non-coherent MIMO flat-fading channels. © 2011 IEEE.