Search

This paper presents a behavior-based adaptive mission planner (AMP) to trace a chemical plume to its source and reliably declare the source location. The proposed AMP is implemented on a REMUS autonomous underwater vehicle (AUV) equipped with multiple types of sensors that measure chemical concentration, the flow velocity vector, and AUV position, depth, altitude, attitude, and speed. This paper describes the methods and results from experiments conducted in November 2002 on San Clemente Island, CA, using a plume of Rhodamine dye developed in a turbulent fluid flow (i.e., near-shore ocean conditions). These experiments demonstrated chemical plume tracing over 100 m and source declaration accuracy relative to the nominal source location on the order of tens of meters. The designed maneuvers are divided into four behavior types: finding a plume, tracing the plume, reacquiring the plume, and declaring the source location. The tracing and reacquiring behaviors are inspired by male moths flying upwind along a pheromone plume to locate a sexually receptive female. All behaviors are formulated by perception and action modules and translated into chemical plume-tracing algorithms suitable for implementation on a REMUS AUV. To coordinate the different behaviors, the subsumption architecture is adopted to define and arbitrate the behavior priorities. AUVs capable of such feats would have applicability in searching for environmentally interesting phenomena, unexploded ordnance, undersea wreckage, and sources of hazardous chemicals or pollutants. [ABSTRACT FROM AUTHOR]

Thin layers of phytoplankton have an important impact on coastal ocean ecology. The high spatial and temporal variability of such layers makes autonomous underwater vehicles (AUVs) ideal for their study. At the Monterey Bay Aquarium Research Institute (MBARI, Moss Landing, CA), the authors have used an AUV for obtaining repeated high-resolution surveys of thin layers in Monterey Bay, CA. The AUV is equipped with ten “gulpers” that can capture water samples when some feature is detected. In this paper, the authors present an adaptive triggering method for an AUV to capture water samples at chlorophyll fluorescence peaks in a thin layer. The algorithm keeps track of the fluorescence background level and the peaks' baseline in real time to ensure that detection is tuned to the ambient conditions. The algorithm crosschecks for concurrent high values of optical backscattering to ensure that sampling targets true particle peaks and not simply physiologically controlled fluorescence peaks. To let the AUV capture the thin layer's peak without delay, the algorithm takes advantage of the vehicle's sawtooth (i.e., yo-yo) trajectory: in one yo-yo cycle, the vehicle makes two crossings of the thin layer. On the first crossing, the vehicle detects the layer's fluorescence peak and saves the peak height; on the second crossing, as the fluorescence measurement reaches the saved peak height (plus meeting additional timing and depth conditions), a sampling is triggered. Based on the thin layer's vertical position in the vehicle's yo-yo profiles, the algorithm selects the pair of detection and triggering crossings so as to minimize the spacing between them. We use the algorithm to postprocess a data set of 20 AUV missions in the 2005 Layered Organization in the Coastal Ocean (LOCO) Experiment in Monterey Bay, CA, and compare its performance with that of a threshold triggering method. In October 2009, the presented method was field tested in an AUV mission in northern Monterey Bay, CA. [ABSTRACT FROM AUTHOR]

This paper proposes a multiclassification algorithm using multilayer perceptron neural network models. It tries to boost two conflicting main objectives of multiclassifiers: a high correct classification rate level and a high classification rate for each class. This last objective is not usually optimized in classification, but is considered here given the need to obtain high precision in each class in real problems. To solve this machine learning problem, we use a Pareto-based multiobjective optimization methodology based on a memetic evolutionary algorithm. We consider a memetic Pareto evolutionary approach based on the NSGA2 evolutionary algorithm (MPENSGA2). Once the Pareto front is built, two strategies or automatic individual selection are used: the best model in accuracy and the best model in sensitivity (extremes in the Pareto front). These methodologies are applied to solve 17 classification benchmark problems obtained from the University of California at Irvine (UCI) repository and one complex real classification problem. The models obtained show high accuracy and a high classification rate for each class. [ABSTRACT FROM AUTHOR]

This paper presents a new learning algorithm for multitask pattern recognition (MTPR) problems. We consider learning multiple multiclass classification tasks online where no information is ever provided about the task category of a training example. The algorithm thus needs an automated task recognition capability to properly learn the different classification tasks. The learning mode is "online" where training examples for different tasks are mixed in a random fashion and given sequentially one after another. We assume that the classification tasks are related to each other and that both the tasks and their training examples appear in random during "online training." Thus, the learning algorithm has to continually switch from learning one task to another whenever the training examples change to a different task. This also implies that the learning algorithm has to detect task changes automatically and utilize knowledge of previous tasks for learning new tasks fast. The performance of the algorithm is evaluated for ten MTPR problems using five University of California at Irvine (UCI) data sets. The experiments verify that the proposed algorithm can indeed acquire and accumulate task knowledge and that the transfer of knowledge from tasks already learned enhances the speed of knowledge acquisition on new tasks and the final classification accuracy. In addition, the task categorization accuracy is greatly improved for all MTPR problems by introducing the reorganization process even if the presentation order of class training examples is fairly biased. [ABSTRACT FROM AUTHOR]

The architectural landscape of high-performance computing stretches from superscalar uniprocessor to explicitly parallel systems to dedicated hardware implementations of algorithms. Single-purpose hardware can achieve the highest performance and uniprocessors can be the most programmable. Between these extremes, programmable and reconfigurable architectures provide a wide range of choice in flexibility, programmability, computational density, and performance. The UCSC Kestrel parallel processor strives to attain single-purpose performance while maintaining user programmability. Kestrel is a single-instruction stream, multiple- data stream (SIMD) parallel processor with a 512-element linear array of 8-bit processing elements. The system design focuses on efficient high-throughput DNA and protein sequence analysis, but its programmability enables high performance on computational chemistry, image processing, machine learning, and other applications. The Kestrel system has had unexpected longevity in its utility due to a careful design and analysis process. Experience with the system leads to the conclusion that programmable SIMD architectures can excel in both programmability and performance. This paper presents the architecture, implementation, applications, and observations of the Kestrel project at the University of California at Santa Cruz. [ABSTRACT FROM AUTHOR]