Virtual Spring-Based 3D Multi-Agent Group Coordination
A virtual spring-based model is proposed for the group safe-distance coordination. Within a specified neighborhood radius, each vehicle forms a virtual connection with each neighbor vehicle by a virtual spring. As the vehicle changes its position, speed and altitude, the total resultant forces on each virtual spring try to maintain zero by moving to the mechanical equilibrium point. The agents then add the simple total virtual spring constraints to their movements to determine their next positions individually.
Together, the multi-agent vehicles reach a group behavior, where each of them keeps a minimal safe-distance with others. A new safe behavior thus arises in the group level. With the proposed virtual spring coordination model, the vehicles need no direct communication with each other, require only minimum local processing resources, and the control is completely distributed. The effects of various neighborhood radius, agent density, and individual vehicle speed on the group behavior are under investigation. New behaviors can now be formulated and studied based on the proposed model, i.e., how a fast driving vehicle can find its way though the crowd by avoiding the other vehicles effortlessly.
Top of Page
Optimal Upwind Sail Boat Control Strategy via Fuzzy Logic
In this project, we examine the kinematics of a simulated sailboat progressing upwind in a sinusoidally oscillating wind. The sailboat is modeled simply as a point moving at a steady state speed in direct proportion to the wind speed, and losing a fraction of that speed in a tack. Two methods of progressing to an upwind destination are considered and compared: The first optimizes the velocity made good (VMG) by tacking when the velocity made good on the other tack exceeds the current VMG by a given percentage. The second method makes use of fuzzy logic to determine the best points to tack and proceed toward the destination.
Top of Page
Parallel Fault Tolerant Floating-Point Multi-Core FPGA Accelerator
To meet the ever increasing needs in floating-point arithmetic speedup of the ever higher precision and finer scale of modern computation-intensive applications, a hardware co-processor accelerator design is proposed. This paper explores the use of a double-precision floating-point multi-core scheme on a Field Programmable Gate Array (FPGA) to work in parallel with Central Processing Unit (CPU).
The shared-memory system architecture chosen will allow high-speed queuing of operations to be performed on the FPGA in a parallel manner. To avoid memory contention, this system requires a high-speed bus controller as well as an on-chip internal floating-point core parallelization bus controller. In the current design, this multi-core FPGA chip replaces one of the on-board dual AMD Opteron PEs (Processing Elements) in DRC system. A fault-tolerant safe guard is also added with duplicated computation plus voting.
Top of Page
Adaptive Parallel Computation-to-Processor Topology Matching
This research effort is aimed at the core problem of the current parallel computing study to revolutionize the current fixed computation-to-machine matching. This paper describes an adaptive technique to optimize the matching between the computation data- flow task graph and the processor network topology. The matching process consists of multithread analysis for task partition, and processor/communication-channel assignment/scheduling.
To tackle the NP-Hard complexity of the matching, a level-based heuristic is designed to prioritize the tasks. A critical path analysis and counting scheme is also added to help determine the task importance order. An A* algorithm is then applied sequentially to the prioritized task list to estimate the cost function to assign the best performing processing elements to the most important tasks. To aid the cost estimate, a shortest-path inter-processor routing matrix is maintained. The resulting algorithm-machine matching adapts based on power/speed/cost/communication/reliability constraints/tradeoffs. The research result will benefit all areas of computation from Instruction Level Parallelism (ILP), Very Long Instruction Word (VLIW) compiler, to cluster/grid resource allocation, fault- tolerance/recovery, algorithm/machine selection, algorithm-specific computer architecture design and adaptive/reconfigurable computing.
Top of Page
Wavelet Texture Analysis for Defect / Abnormality Detection
In any manufacture process, quality control is very essential to ensure that the product is free from defect. Defect Identification, Isolation and Resolution are utmost important to ensure the accuracy of the results achieved by any manufacture process. By applying multi-level wavelet decomposition to the input image and then choosing a group of sub- bands to be restored for best defect detection, the technique used detects a defect in both structural and statistical texture. This makes the proposed technique computationally efficient because that it does not rely on the texture feature extraction on pixel by pixel basis.
The experimental results showed that a few factors are considered to be very important for best defect detection: The type of wavelet bases which various wavelet bases can be used with superiority of compact support orthogonal wavelet bases. Also, 3 or 4 level can enhance the defect properly for a detection task. The image restoration strategy is important in which the different image restoration levels for different texture classes were used for defect detection in textures. The results on statistical texture showed that defect enhancement by restoring the coarse approximation of image is rotation invariant. Also, restoring some details of sub bands to enhance the defect in structural texture may cause unpredictable changes in image rotation. Beside manufacture inspection, this texture analysis can be tuned to inspect space station/shuttle damage or for cancer screening with mouth/colon/cervix ex-vivo scopes.
Top of Page
Image Texture Clustering for Prostate Ultrasound Diagnosis
Motivation/Background - To double the accuracy of ultrasound diagnosis and biopsy guidance, an efficient, integrated platform for image textural analysis and clustering of transrectal prostate ultrasound images into clusters potentially representing cancerous or normal tissue areas is presented. Preliminary image texture analysis has shown the potential for doubling diagnosis accuracy from 38-42% for prostate cancer with current clinical methods, to 88-92%.
In addition, image texture analysis makes prostate cancer tumor locating possible for more precise, less invasive biopsy/treatment, instead of 6-way random biopsy. Due to the tedious derivation of co-occurrence matrix and scanning a sub-image tile sequentially, pixel by pixel across and down the entire image, the initial image texture analysis on a NASA JSC's miniVAX could take 8 days CPU time per image, i.e., more than 5 months analysis for 20 cross-sections per patient, making clinical point- of-care diagnosis impossible.
Statement of the Contribution/Methods - An efficient Image Texture Analysis tool platform on Window PC is constructed via innovative sparse co-occurrence matrix techniques with linked lists to speedup the processing from 8 days to about 5 seconds per image on a PC.
The approach is based on Haralick�s textural features and the Minimum Squared Error (MSE) clustering algorithm. This significant reduction in run time potentially allows more accurate, objective diagnoses to be performed within clinical settings, as well as enables the general investigation of image textural and clustering parameters.
Results/Discussion - Ultrasound diagnosis is proven to be less invasive, portable, at lower screening cost than most other medical images. However, being less visual than most, ultrasound image diagnosis is difficult even for trained professionals, and thus can benefit greatly from computer enhancement. Image texture analysis and clustering are improved from 8 days per image down to 5 seconds on a PC to enable point-of-care computer enhanced diagnosis as well as more accurate tumor locating capabilities for treatment and biopsy guidance. Using this integrated approach, specific results for several sample patient image cases are tested and general conclusions are drawn, yet more images need to be tested.
For future study, the color-coded cluster tissue sample images can potentially represent from normal to various degrees of seriousness of abnormality in sample images similar to Gleason score for prostate cancer. Beside the current 5 Haralick's textural features, wavelet and other faster image texture features are also under consideration for inclusion. The parameter settings of this efficient image texture analysis tool can also be tuned to cluster pixel groups in wide varieties of challenging micro/macro image applications.