Scalable hybrid computation with spikes

We outline a hybrid analog-digital scheme for computing with three important features that enable it to scale to systems of large complexity: First, like digital computation, which uses several one-bit precise logical units to collectively compute a precise answer to a computation, the hybrid scheme uses several moderate-precision analog units to collectively compute a precise answer to a computation. Second, frequent discrete signal restoration of the analog information prevents analog noise and offset from degrading the computation. And, third, a state machine enables complex computations to be created using a sequence of elementary computations. A natural choice for implementing this hybrid scheme is one based on spikes because spike-count codes are digital, while spike-time codes are analog. We illustrate how spikes afford easy ways to implement all three components of scalable hybrid computation. First, as an important example of distributed analog computation, we show how spikes can create a distributed modular representation of an analog number by implementing digital carry interactions between spiking analog neurons. Second, we show how signal restoration may be performed by recursive spike-count quantization of spike-time codes. And, third, we use spikes from an analog dynamical system to trigger state transitions in a digital dynamical system, which reconfigures the analog dynamical system using a binary control vector; such feedback interactions between analog and digital dynamical systems create a hybrid state machine (HSM). The HSM extends and expands the concept of a digital finite-state-machine to the hybrid domain. We present experimental data from a two-neuron HSM on a chip that implements error-correcting analog-to-digital conversion with the concurrent use of spike-time and spike-count codes. We also present experimental data from silicon circuits that implement HSM-based pattern recognition using spike-time synchrony. We outline how HSMs may be used to perform learning, vector quantization, spike pattern recognition and generation, and how they may be reconfigured.     

Load Balancing Parallel Explicit State Model Checking

This paper first identifies some of the key concerns about the techniques and algorithms developed for parallel model checking; specifically, the inherent problem with load balancing and large queue sizes resultant in a static partition algorithm. This paper then presents a load balancing algorithm to improve the run time performance in distributed model checking, reduce maximum queue size, and reduce the number of states expanded before error discovery. The load balancing algorithm is based on generalized dimension exchange (GDE). This paper presents an empirical analysis of the GDE based load balancing algorithm on three different supercomputing architectures—distributed memory clusters, Networks of Workstations (NOW) and shared memory machines. The analysis shows increased speedup, lower maximum queue sizes and fewer total states explored before error discovery on each of the architectures. Finally, this paper presents a study of the communication overhead incurred by using the load balancing algorithm, which although significant, does not offset performance gains.     

The physics of e-commerce supply chains

The Internet is rapidly improving our ability to support information flows across global supply chains. However, the business impact of these IT investments depends not just on the information flows supported by such systems, but also on other non-information related characteristics of the supply chain. It is the complex interplay among these characteristics that results in business performance, or lack thereof. In order to better understand the business impact of IT investments in the supply chain, we are motivated to examine the ‘physics’ of supply chain structures. By physics, we mean the performance patterns inherent in its informational and physical characteristics. Using the systems dynamics methodology, we model basic information and physical characteristics of supply chains and examine their impact on some common measures of performance. Experiments with the models suggest that, in addition to information delays, physical delays also have a major impact on the stability of supply chains, as well as on operating cost. Moreover, the tradeoff between chain stability and responsiveness appears to be nonlinear, suggesting that a small compromise in responsiveness may yield larger gains in stability. Multi-tier chains appear to be less ‘stiff’ in responding to demand fluctuations, implying that their information systems must be specifically designed to overcome this structural tendency. These results have managerial implications in terms of designing the information and physical structure of a supply chain as well as for its operation.     

Fabrication of Integrated Vertical Mirror Surfaces and Transparent Window for Packaging MEMS Devices

A scheme for creating metal-coated vertical mirrors in silicon, along with an integrated transparent package lid for assembling, packaging, and testing microelectromechanical systems (MEMS) devices is presented. Deep reaction ion etching (DRIE) method described here reduces the loading effect and maintains a uniform etch rate resulting in highly vertical structures. A novel self-masking lithography and liftoff process was developed to ensure that the vertical mirrors undergo uniform metallization while leaving a transparent window for optical probing. Front side of a Si wafer was shallow-etched using DRIE to define an eventual optical window. This surface was then anodically bonded to a Pyrex wafer. Backside Si was then patterned to define thin channels around the optical window. These channels were vertically etched using DRIE, after which the unattached portions of the window region were removed. Negative photoresist was spun on the remaining vertical structures and the stack was exposed from the Pyrex side using Si structures as a self-mask. Subsequent metal sputtering and liftoff results in the metallized top and mirror sidewalls while leaving a clear window. These integrated mirrors and lids are then bonded to the active MEMS mirrors. Various processes and results are illustrated with an example of packaged corner cube retroreflectors (CCRs)     

Rapidly exploring random graphs: motion planning of multiple mobile robots

Rapidly exploring random trees (RRT) and probabilistic roadmaps (PRM) are sampling-based techniques being extensively used for robot path planning. In this paper, the tree structure of the RRT is generalized to a graph structure which enables a greater exploration. Exploration takes place simultaneously from multiple points in the map, all explorations fusing at multiple points producing well-connected graph architecture. Initially, in a typical RRT manner, the search algorithm attempts to reach the goal by expansions, and thereafter furtherer areas are explored. With some additional computation cost, as compared to RRT with a single robot, the results can be significantly improved. The so-formed graph is similar to roadmap produced by PRM. However as compared to PRM, the proposed algorithm has a more judicious search strategy and is adaptable to the number of nodes as a parameter. Experimental results are shown with multiple robots planned using prioritization scheme. Results show the betterment of the proposed algorithm as compared to RRT and PRM techniques.     

MULTIROBOT MOTION PLANNING USING HYBRID MNHS AND GENETIC ALGORITHMS

Planning the motion of multiple robots deals with computing the motion of all robots avoiding any collision. This article focuses on the use of hybrid Multi Neuron Heuristic Search (MNHS) and Genetic Algorithm (GA). The MNHS is an advancement over the conventional A* algorithm and is better suited for maze-like conditions where there is a high degree of uncertainty. The MNHS contributes toward optimality of the solution, and the GA gives it an iterative nature and enables the approach to be used on high-resolution maps. MNHS works over the set of points returned by the GA in its fitness function evaluation. A priority-based approach is used, in which the priorities are decided by the GA. Path feasibility is speeded up by using the concept of coarser-to-finer lookup called momentum. Experimental results show that the combined approach is able to easily solve the problem for a variety of scenarios.

[Supplementary materials are available for this article. Go to the publisher's online edition of Applied Artificial Intelligence for the following free supplemental resource(s): Videos 1-4]     

Majorisations for the eigenvectors of graph-adjacency matrices

We develop majorisation results that characterise changes in eigenvector components of a graph's adjacency matrix when its topology is changed. Specifically, for general (weighted, directed) graphs, we characterise changes in dominant eigenvector components for single- and multi-row incrementations. We also show that topology changes can be tailored to set ratios between the components of the dominant eigenvector. For more limited graph classes (specifically, undirected, and reversibly-structured ones), majorisations for components of the subdominant and other eigenvectors upon graph modifications are also obtained.     

Comparing stationary standing with an intermittent walking posture during assembly operations

The purpose of this investigation was to evaluate two different standing postures, intermittent walking (dynamic) and stationary, on 11 volunteers. While performing the activity in a stationary posture, volunteers worked barefoot for 25 minutes at a particular workstation, taking up to five passive breaks. The dynamic standing postural activity was similar in every respect with the exception of the nature of the breaks that were spent walking from one workstation to the other. Foot pressure was analyzed using a static evaluation variable (contact area) and a dynamic evaluation variable (area pressure change root mean square [aPcrms]). Foot center of pressure (COP) was analyzed using two of its features: excursion length (LEN) and root mean square velocity of excursions (VEL). In addition, a psychophysical study was used to distinguish the comfort levels of various muscles between the two postures. The quantifications of the four physiological variables revealed a lower comfort index (p < 0.05) in stationary standing compared to dynamic standing during assembling duties. This significant distinction was further corroborated by the results of the psychophysical test. Such a positive influence of the dynamic standing posture on standing comfort can be used to improve industrial productivity. © 2008 Wiley Periodicals, Inc.