Juan Manuel García-Chamizo

2012 Special Issue: Autonomous Growing Neural Gas for applications with time constraint: Optimal parameter estimation

This paper aims to address the ability of self-organizing neural network models to manage real-time applications. Specifically, we introduce fAGNG (fast Autonomous Growing Neural Gas), a modified learning algorithm for the incremental model Growing Neural Gas (GNG) network. The Growing Neural Gas network with its attributes of growth, flexibility, rapid adaptation, and excellent quality of representation of the input space makes it a suitable model for real time applications. However, under time constraints GNG fails to produce the optimal topological map for any input data set. In contrast to existing algorithms, the proposed fAGNG algorithm introduces multiple neurons per iteration. The number of neurons inserted and input data generated is controlled autonomous and dynamically based on a priory or online learnt model. A detailed study of the topological preservation and quality of representation depending on the neural network parameter selection has been developed to find the best alternatives to represent different linear and non-linear input spaces under time restrictions or specific quality of representation requirements.

FPGA-based tool path computation: an application for shoe last machining on CNC lathes

Tool path generation is one of the most complex problems in computer aided manufacturing. Although some efficient strategies have been developed, most of them are only useful for standard machining. The algorithm called Virtual Digitizing computes the tool path by means of a virtually digitized model of the surface and a geometry specification of the tool and its motion, so it can even be used in non-standard machining. This algorithm is simple and robust and avoids the problem of tool-surface collision by its own definition. A Virtual Digitizing optimisation that makes the most of specific hardware in order to improve the algorithm efficiency is presented in this paper. A comparative study will show the gain achieved in terms of total computing time. We also present a FPGA-based architecture that can be used to produce rotations with more precision and speed than other well-known classic implementations.

Developing Ubiquitous Sensor Network Platform Using Internet of Things: Application in Precision Agriculture

The application of Information Technologies into Precision Agriculture methods has clear benefits. Precision Agriculture optimises production efficiency, increases quality, minimises environmental impact and reduces the use of resources (energy, water); however, there are different barriers that have delayed its wide development. Some of these main barriers are expensive equipment, the difficulty to operate and maintain and the standard for sensor networks are still under development. Nowadays, new technological development in embedded devices (hardware and communication protocols), the evolution of Internet technologies (Internet of Things) and ubiquitous computing (Ubiquitous Sensor Networks) allow developing less expensive systems, easier to control, install and maintain, using standard protocols with low-power consumption. This work develops and test a low-cost sensor/actuator network platform, based in Internet of Things, integrating machine-to-machine and human-machine-interface protocols. Edge computing uses this multi-protocol approach to develop control processes on Precision Agriculture scenarios. A greenhouse with hydroponic crop production was developed and tested using Ubiquitous Sensor Network monitoring and edge control on Internet of Things paradigm. The experimental results showed that the Internet technologies and Smart Object Communication Patterns can be combined to encourage development of Precision Agriculture. They demonstrated added benefits (cost, energy, smart developing, acceptance by agricultural specialists) when a project is launched.

Image Compression Using Growing Neural Gas

In this paper we study the capacities of characterization and synthesis of objects by using a self-organizing neural model, the Growing Neural Gas. These networks, by means of their competitive learning try to preserve the topology of an input space. This feature is being used for the representation of objects and their movement with topology preserving networks. We characterize the object to be represented by means of the obtained maps and kept information solely on the coordinates and the pixel color of the neurons. With this information it is made the synthesis of the original images, applying mathematical morphology and simple filters using the available information.

Real-time arithmetic unit

In this paper we discuss the paradigm of real-time processing on the lower level of computing systems. An arithmetical unit based on this principle containing addition, multiplication, division and square root operations is described. The development of the computation operators model is based on the imprecise computation paradigm and defines the concept of the adjustable calculation of a function that manages delay and the precision of the results as an inherent and parameterized characteristic. The arithmetic function design is based on well-known algorithms and offers progressive improvement in the results. Advantages in the predictability of calculations are obtained by means of processing groups of k-bits atomically and by using look-up tables. We report an evaluation of the operations in path time, delay and computation error. Finally, we present an example of our real-time architecture working in a realistic context.

Mobile agent system framework suitable for scalable networks

Purpose – To provide a formal framework based on the action and reaction model that allows us to cover the dynamics of multi‐agent systems (MAS) made up of mobile software agents suitable for scalable networks.Design/methodology/approach – This model is based on the operation of the human nervous centers. In the case of systems based on mobile agents, the main problem is the different vision the agents have of the world and the impossibility of being aware of and synchronizing all the influences brought by the different agents acting on it.Findings – This proposal has been compared with the conventional MAS by solving an extension of the predator‐prey problem. The results show the advantages of mobility, as the size of the problem grows in a distributed system.Practical implications – At the present time, the model is being applied in works related to the control of biological systems and also in those related to the network management.Originality/value – From this formulation, a set of refinements on the...In this article we present a formal framework based on the action and reaction model that allows us to cover the dynamics of multi-agent systems (MAS) made up of mobile software agents suitable for scalable networks. This model is based on the operation of the human nervous centres. At the present time, we are applying it in works related with the control of biological systems and also in those related to the network management. In the case of systems based on mobile agents, the main problem is the different vision the agents have of the world and the impossibility of being aware of and synchronizing all the influences brought by the different agents acting on it. We have compared our proposal with the conventional MAS by solving an extension of the predator-prey problem. The results show the advantages of mobility as the size of the problem grows in a distributed system.

Real time motion estimation using a neural architecture implemented on GPUs

Abstract This work describes a neural network based architecture that represents and estimates object motion in videos. This architecture addresses multiple computer vision tasks such as image segmentation, object representation or characterization, motion analysis and tracking. The use of a neural network architecture allows for the simultaneous estimation of global and local motion and the representation of deformable objects. This architecture also avoids the problem of finding corresponding features while tracking moving objects. Due to the parallel nature of neural networks, the architecture has been implemented on GPUs that allows the system to meet a set of requirements such as: time constraints management, robustness, high processing speed and re-configurability. Experiments are presented that demonstrate the validity of our architecture to solve problems of mobile agents tracking and motion analysis.

3D colour object reconstruction based on Growing Neural Gas

With the advent of low-cost 3D sensors and 3D printers, surface reconstruction has become an important research topic in the last years. In this work, we propose an automatic method for 3D surface reconstruction from raw unorganized point clouds acquired using low-cost sensors. We have modified the Growing Neural Gas (GNG) network, which is a suitable model because of its flexibility, rapid adaptation and excellent quality of representation, to perform 3D surface reconstruction of different real-world objects. Some improvements have been made on the original algorithm considering colour information during the learning stage and creating complete triangular meshes instead of basic wire-frame representations. The proposed method is able to create 3D faces online, whereas existing 3D reconstruction methods based on Self-Organizing Maps (SOMs) required post-processing steps to close gaps and holes produced during the 3D reconstruction process. Performed experiments validated how the proposed method improves existing techniques removing post-processing steps and including colour information in the final triangular mesh.

Fast Autonomous Growing Neural Gas

This paper aims to address the ability of self-organizing neural network models to manage real-time applications. Specifically, we introduce fAGNG (fast Autonomous Growing Neural Gas), a modified learning algorithm for the incremental model Growing Neural Gas (GNG) network. The Growing Neural Gas network with its attributes of growth, flexibility, rapid adaptation, and excellent quality of representation of the input space makes it a suitable model for real time applications. However, under time constraints GNG fails to produce the optimal topological map for any input data set. In contrast to existing algorithms the proposed fAGNG algorithm introduces multiple neurons per iteration. The number of neurons inserted and input data generated is controlled autonomous and dynamically based on a priory learnt model. Comparative experiments using topological preservation measures are carried out to demonstrate the effectiveness of the new algorithm to represent linear and non-linear input spaces under time restrictions.

Fast image representation with GPU-based growing neural gas

This paper aims to address the ability of self-organizing neural network models to manage real-time applications. Specifically, we introduce a Graphics Processing Unit (GPU) implementation with Compute Unified Device Architecture (CUDA) of the Growing Neural Gas (GNG) network. The Growing Neural Gas network with its attributes of growth, flexibility, rapid adaptation, and excellent quality representation of the input space makes it a suitable model for real time applications. In contrast to existing algorithms the proposed GPU implementation allow the acceleration keeping good quality of representation. Comparative experiments using iterative, parallel and hybrid implementation are carried out to demonstrate the effectiveness of CUDA implementation in representing linear and non-linear input spaces under time restrictions.