Palette images are widely used on the World Wide Web (WWW) and in game-cartridge applications. Many images used on the WWW are stored and transmitted after they are compressed losslessly with the standard graphics interchange format (GIF), or portable network graphics (PNG). Well-known 2-D compression schemes, such as JPEG-LS and JPEG-2000, fail to yield better compression than GIF or PNG due to the fact that the pixel values represent indices that point to color values in a look-up table. To improve the compression performance of JPEG-LS and JPEG-2000 techniques, several researchers have proposed various reindexing algorithms. We investigate various compression techniques for color palette images. We propose a new technique comprised of a traveling salesman problem (TSP)-based reindexing scheme, Burrows-Wheeler transformation, and inversion ranks. We show that the proposed technique yields better compression gain on average than all the other 1-D compressors and the reindexing schemes that utilize JPEG-LS or JPEG-2000.
In this research a generalized software framework that enables accurate computer aided design of MEMS devices is developed. The proposed simulation engine utilizes a novel material property estimation technique that generates effective material properties at the microscopic level. The material property models were developed based on empirical analysis and the behavior extraction of standard test structures. A literature review is provided on the physical phenomena that govern the mechanical behavior of thin films materials. This survey indicates that the present day models operate under a wide range of assumptions that may not be applicable to the micro-world. Thus, this methodology is foreseen to be an essential tool for MEMS designers as it would develop empirical models that relate the loading parameters, material properties, and the geometry of the microstructures with its performance characteristics. This process involves learning the relationship between the above parameters using non-parametric learning algorithms such as radial basis function networks and genetic algorithms. The proposed simulation engine has a graphical user interface (GUI) which is very adaptable, flexible, and transparent. The GUI is able to encompass all parameters associated with the determination of the desired material property so as to create models that provide an accurate estimation of the desired property. This technique was verified by fabricating and simulating bilayer cantilevers consisting of aluminum and glass (TEOS oxide) in our previous work. The results obtained were found to be very encouraging.
A decentralized version of particle swarm optimization called the distributed particle swarm optimization (DPSO) approach is formulated and applied to the generation of sensor network configurations or topologies so that the deleterious effects of hidden nodes and asymmetric links on the performance of wireless sensor networks are minimized. Three different topology generation schemes, COMPOW, Cone-Based and the DPSO--based schemes are examined using ns-2. Simulations are executed by varying the node density and traffic rates. Results contrasting heterogeneous vs. homogeneous power reveal that an important metric for a sensor network topology may involve consideration of hidden nodes and asymmetric links, and demonstrate the effect of spatial reuse on the potency of topology generators.
Since untethered sensor nodes operate on battery, and because they must communicate through a multi-hop network, it is vital to optimally configure the transmit power of the nodes both to conserve power and optimize spatial reuse of a shared channel. Current topology control algorithms try to minimize radio power while ensuring connectivity of the network. We propose that another important metric for a sensor network topology will involve consideration of hidden nodes and asymmetric links. Minimizing the number of hidden nodes and asymmetric links at the expense of increasing the transmit power of a subset of the nodes may in fact increase the longevity of the sensor network. In this paper we explore a distributed evolutionary approach to optimizing this new metric. Inspiration from the Particle Swarm Optimization technique motivates a distributed version of the algorithm. We generate topologies with fewer hidden nodes and asymmetric links than a comparable algorithm and present some results that indicate that our topologies deliver more data and last longer.
Diffractive optical elements (DOE) utilize diffraction to manipulate light in optical systems. These elements have a wide range of applications including optical interconnects, coherent beam addition, laser beam shaping and refractive optics aberration correction. Due to the wide range of applications optimal design of DOE has become an important research problem. In the design of the DOEs, existing techniques utilize the Fresnel diffraction theory to compute the phase at the desired location at the output plane. Since this process involves solving nonlinear integral equations, various numerical methods along with robust optimization algorithms have been proposed. However all the algorithms proposed so far assume that the size and the spacing of the elements as independent variables in the design of optimal diffractive gratings. Therefore search algorithms need to be called every time the required geometry of the elements changes, resulting in a computationally expensive design procedure for systems utilizing a large number of DOEs.
In this work, we have developed a novel algorithm that uses neural networks with multiple hidden layers to overcome this limitation and arrives at a general solution for the design of the DOEs for a given application. Inputs to this network are the spacing between the elements and the input/output planes. The network outputs the phase gratings that are required to obtain the desired intensity at the specified location in the output plane. The network was trained using the back-propagation technique. The training set was generated by using genetic algorithm approach as described in literature. The mean square error obtained is comparable to conventional techniques but with much lower computational costs.
Untethered, underwater sensors, deployed for event detection and tracking and operating in an autonomous mode will be required to self-assemble into a configuration, which optimizes their coverage, effectively minimizing the probability that an event in the target area goes undetected. This organized, cooperative, and autonomous, spreading-out of the sensors is complicated due to sensors localized communication. A given sensor will not in general have position and velocity information for all sensors, but only for those in its communication area. A possible approach to this problem, motivated by an evolutionary optimization technique, Particle Swarm Optimization (PSO) is proposed and extended in a novel way. A distributed version of PSO is developed. A distributed version of PSO is explored using experimental fitness to address the coverage problem in a two dimensional area.
When wireless sensors are capable of variable transmit power and are battery powered, it is important to select the appropriate transmit power level for the node. Lowering the transmit power of the sensor nodes imposes a natural clustering on the network and has been shown to improve throughput of the network. However, a common transmit power level is not appropriate for inhomogeneous networks. A possible fitness-based approach, motivated by an evolutionary optimization technique, Particle Swarm Optimization (PSO) is proposed and extended in a novel way to determine the appropriate transmit power of each sensor node. A distributed version of PSO is developed and explored using experimental fitness to achieve an approximation of least-cost connectivity.
Genetic Algorithms are powerful tools, which when set upon a solution space will search for the optimal answer. These algorithms though have some associated problems, which are inherent to the method such as pre-mature convergence and lack of population diversity. These problems can be controlled with changes to certain parameters such as crossover, selection, and mutation. This paper attempts to tackle these problems in GA by having another GA controlling these parameters. The values for crossover parameter are: one point, two point, and uniform. The values for selection parameters are: best, worst, roulette wheel, inside 50%, outside 50%. The values for the mutation parameter are: random and swap. The system will include a control GA whose population will consist of different parameters settings. While this GA is attempting to find the best parameters it will be advancing into the search space of the problem and refining the population. As the population changes due to the search so will the optimal parameters. For every control GA generation each of the individuals in the population will be tested for fitness by being run through the problem GA with the assigned parameters. During these runs the population used in the next control generation is compiled. Thus, both the issue of finding the best parameters and the solution to the problem are attacked at the same time. The goal is to optimize the sensor coverage in a square field. The test case used was a 30 by 30 unit field with 100 sensor nodes. Each sensor node had a coverage area of 3 by 3 units. The algorithm attempts to optimize the sensor coverage in the field by moving the nodes. The results show that the control GA will provide better results when compared to a system with no parameter changes.
Discovering relationships between variables is crucial for interpreting data from large databases. Relationships between variables can be modeled using a Bayesian network. The challenge of learning a Bayesian network from a complete dataset grows exponentially with the number of variables in the database and the number of states in each variable. It therefore becomes important to identify promising heuristics for exploring the space of possible networks. This paper utilizes an evolutionary algorithmic approach, Particle Swarm Optimization (PSO) to perform this search. A fundamental problem with a search for a Bayesian network is that of handling cyclic networks, which are not allowed. This paper explores the PSO approach, handling cyclic networks in two different ways. Results of network extraction for the well-studied ALARM network are presented for PSO simulations where cycles are broken heuristically at each step of the optimization and where networks with cycles are allowed to exist as candidate solutions, but are assigned a poor fitness. The results of the two approaches are compared and it is found that allowing cyclic networks to exist in the particle swarm of candidate solutions can dramatically reduce the number of objective function evaluations required to converge to a target fitness value.
Clustering is necessary for data aggregation, hierarchical routing, optimizing sleep patterns, election of extremal sensors, optimizing coverage and resource allocation, reuse of frequency bands and codes, and conserving energy. Optimal clustering is typically an NP-hard problem. Solutions to NP-hard problems involve searches through vast spaces of possible solutions. Evolutionary algorithms have been applied successfully to a variety of NP-hard problems. We explore one such approach, Particle Swarm Optimization (PSO), an evolutionary programming technique where a 'swarm' of test solutions, analogous to a natural swarm of bees, ants or termites, is allowed to interact and cooperate to find the best solution to the given problem. We use the PSO approach to cluster sensors in a sensor network. The energy efficiency of our clustering in a data-aggregation type sensor network deployment is tested using a modified LEACH-C code. The PSO technique with a recursive bisection algorithm is tested against random search and simulated annealing; the PSO technique is shown to be robust. We further investigate developing a distributed version of the PSO algorithm for clustering optimally a wireless sensor network.