Cristiana Bentes

A parallel heuristic for the Vehicle Routing Problem with Simultaneous Pickup and Delivery

This paper presents a parallel approach for solving the Vehicle Routing Problem with Simultaneous Pickup and Delivery (VRPSPD). The parallel algorithm is embedded with a multi-start heuristic which consists of a variable neighborhood descent procedure, with a random neighborhood ordering (RVND), integrated in an iterated local search (ILS) framework. The experiments were performed in a cluster with a multi-core architecture using up to 256 cores. The results obtained on the benchmark problems, available in the literature, show that the proposed algorithm not only improved several of the known solutions, but also presented a very satisfying scalability.

Sensor data analysis for equipment monitoring

Sensors play a key role in modern industrial plant operations. Nevertheless, the information they provide is still underused. Extracting information from the raw data generated by the sensors is a complicated task, and it is usually used to help the operator react to undesired events, other than preventing them. This paper presents SDAEM (Sensor Data Analysis for Equipment Monitoring), an oil process plant monitoring model that covers three main goals: mining the sensor time series data to understand plant operation status and predict failures, interpreting correlated data from different sensors to verify sensors interdependence, and adjusting equipments working set points that leads to a more stable plant operation and avoids an excessive number of alarms. In addition, as time series data generated by sensors grow at an extremely fast rate, SDAEM uses parallel processing to provide real-time feedback. We have applied our model to monitor a process plant of a Brazilian offshore platform. Initial results were promising since some undesired events were recognized and operators adopted the tool to assist them finding good set points for the oil processing equipments.

An integrated CPU–GPU heuristic inspired on variable neighbourhood search for the single vehicle routing problem with deliveries and selective pickups

Environmental issues have become increasingly important to industry and business in recent days. This trend forces the companies to take responsibility for product recovery, and proper recycling and disposal, moving towards the design of sustainable green supply chains. This paper addresses the backward stream in transportation of products, by means of reverse logistics applied to vehicle routing. This problem, called single vehicle routing problem with deliveries and selective pickups, consists in finding a route that starts from the depot and visits all delivery customers. Some pickup customers may also be visited, since the capacity of the truck is not exceeded, and there is also a revenue associated with each pickup. We develop an algorithm inspired on the variable neighbourhood search metaheuristic that explores the power of modern graphics processing unit (GPU) to provide routes in reasonable computational time. The proposed algorithm called four-neighbourhood variable neighbourhood search (FN-VNS) includes a novel high-quality initial solution generator, a CPU–GPU integrated perturbation strategy and four different neighbourhood searches implemented purely in GPU for the local search phase. Our experimental results show that FN-VNS is able to improve the quality of the solution for 51 instances out of 68 instances taken from the literature. Finally, we obtained speedups up to 14.49 times, varying from 17.42 up to 76.84 for each local search, measured over a set of new large-size instances.

Evolving CUDA PTX programs by quantum inspired linear genetic programming

The tremendous computing power of Graphics Processing Units (GPUs) can be used to accelerate the evolution process in Genetic Programming (GP). The automatic generation of code using the GPU usually follows two different approaches: compiling each evolved or interpreting multiple programs. Both approaches, however, have performance drawbacks. In this work, we propose a novel approach where the GPU pseudo-assembly language, PTX (Parallel Thread Execution), is evolved. Evolving PTX programs is faster, since the compilation of a PTX program takes orders of magnitude less time than a CUDA program compilation on the CPU, and no interpreter is necessary. Another important aspect of our approach is that the evolution of PTX programs follows the Quantum Inspired Linear Genetic Programming (QILGP). Our approach, called QILGP3U (QILGP + GPGPU), enables the evolution on a single machine in a reasonable time, enhances the quality of the model with the use of PTX, and for big databases can be much faster than the CPU implementation.

A Region-Growing Segmentation Algorithm for GPUs

This letter proposes a parallel version for graphics processing units (GPU) of a region-growing image segmentation algorithm widely used by the geographic object-based image analysis (GEOBIA) community. Initially, all image pixels are considered as seeds or primitive segments. Fine-grained parallel threads assigned to individual pixels merge adjacent segments iteratively always ensuring to minimize the overall heterogeneity increase. Besides spectral features the merging criterion considers morphological features that can be efficiently computed in the underlying GPU architecture. Two alternatives using different merging criteria are proposed and tested. An experimental analysis upon five different test images has shown that the parallel algorithm may run up to 19 times faster than its sequential counterpart.

Exploring Data Streaming to Improve 3D FFT Implementation on Multiple GPUs

FFT is a well known and widely used algorithm in many scientific and engineering applications. However, FFT is a memory-bound problem that still presents performance challenges to new generations of computer architectures due to its relatively low ratio of computation per memory access. For GPU architectures, where the data transfers between the host CPU memory and the device memory is very expensive, the memory overhead can become a huge bottleneck for large size problems. In this work, we propose an efficient parallel implementation of FFT on multiple GPUs that tackles the overhead of host memory access, by implementing a streaming scheme that hides the data transfer latency. The idea is to divide the problem into smaller ones, generating several lighter and asynchronous memory transfers from host to device enabling the computation for those data simultaneously. We obtained an acceleration of approximately 60% over the non streamed GPU implementation.

Irregular Grid Raycasting Implementation on the Cell Broadband Engine

Direct volume rendering has become a popular technique for visualizing volumetric data from sources such as scientific simulations, analytic functions, medical scanners, among others. Volume rendering algorithms, such as raycasting, can produce high-quality images, however, the use of raycasting has been limited due to its high demands on computational power and memory bandwidth. In this paper, we propose a new implementation of the raycasting algorithm that takes advantage of the highly parallel architecture of the Cell Broadband Engine processor, with 9 heterogeneous cores, in order to allow efficient raycasting of irregular datasets. All the computational power of the Cell~BE processor, though, comes at the cost of a different programming model. Applications need to be rewritten, which requires using multithreading and vectorized code. In our approach, we tackle this problem by distributing ray computations using the visible faces, and vectorizing the lighting integral operations inside each core. Our experimental results show that we can obtain good speedups reducing the overall rendering time significantly.

Towards distributed region growing image segmentation based on MapReduce

Image segmentation is a critical step in image analysis, and usually involves a high computational cost, especially when dealing with large volumes of data. Given the significant increase in the spatial, spectral and temporal resolutions of remote sensing imagery in the last years, current sequential and parallel solutions fail to deliver the expected performance and scalability. This work proposes a scalable and efficient segmentation method, capable of handling efficiently very large high resolution images. The proposed solution is based on the MapReduce model, which offers a highly scalable and reliable framework for storing and processing massive data in cloud computing environments. The solution was implemented and validated using the Hadoop platform. Experimental results attest the viability of performing region growing segmentation in the MapReduce framework.

The Single Vehicle Routing Problem with Deliveries and Selective Pickups in a CPU-GPU Heterogeneous Environment

In this work, we propose a new algorithm to solve a variant of the Vehicle Routing Problem that is the Single Vehicle Routing Problem with Deliveries and Selective Pickups (SVRPDSP). Our algorithm produces good quality solutions that are better than the best known solutions in the literature. In order to reduce the time spent to solve large-sized instances, we also propose here a parallel implementation of our algorithm that explores a heterogeneous environment composed of a CPU and a GPU. Therefore, our algorithm harnesses the tremendous computing power of the GPU to improve the performance of the local searches computation. We obtained average speedups from 2.73 to 16.23 times with our parallel approach.

Sharing design perspectives through storytelling

Design consists of analyzing scenarios and proposing artifacts, obeying the initial set of requirements that lead from initial to goal state. Finding or creating alternative solutions, analyzing them, and selecting the best one are expected steps in the designer’s decision making process. Very often, not a sole designer, but a team of them is engaged in the design process, sharing their expertise and responsibility to achieve optimum projects. In a design team, most conflicts occur due to misunderstanding of one’s assessment of specifications and contexts. Decision explanations play a key role in teamwork success. Designers are rational agents trained to follow rational methods. Acceptable justifications include value function, requirements, constraints, and criteria. Generally, explanations are delivered in a multimedia fashion, composed of text, graphics and gestures, to provide the audience the ability to perceive what was contextually imagined. The more spatial the reasoning is, the richer the explanation channel should be. This paper presents CineADD, a design explanation generation model based on cinema techniques such as animation, scripting, editing, and camera movements. The idea is to provide designers with a tool for describing the way their projects should be visually explained, as in a movie. Designers develop their projects in an active design document environment. Rationale is captured as a design model, so explanations can be generated instead of retrieved. The captured design model serves as a base to visually reconstruct design, giving emphasis and guidance by using movie storytelling techniques. CineADD was implemented for the domain of oil pipeline layout showing the feasibility of this approach. We expect CineADD to become a commodity attachable to any intelligent CAD system.