Multi-stage classifier design

In many classification systems, sensing modalities have different acquisition costs. It is often unnecessary to use every modality to classify a majority of examples. We study a multi-stage system in a prediction time cost reduction setting, where the full data is available for training, but for a test example, measurements in a new modality can be acquired at each stage for an additional cost. We seek decision rules to reduce the average measurement acquisition cost. We formulate an empirical risk minimization problem (ERM) for a multi-stage reject classifier, wherein the stage k classifier either classifies a sample using only the measurements acquired so far or rejects it to the next stage where more attributes can be acquired for a cost. If we restrict ourselves to binary classification setting then, to solve the ERM problem, we show that the optimal reject classifier at each stage is a combination of two binary classifiers, one biased towards positive examples and the other biased towards negative examples. We use this parameterization to construct stage-by-stage global surrogate risk, develop an iterative algorithm in the boosting framework and present convergence and generalization results. We test our work on synthetic, medical and explosives detection datasets. Our results demonstrate that substantial cost reduction without a significant sacrifice in accuracy is achievable.     

Compiler-Assisted Checkpointing of Parallel Codes: The Cetus and LLVM Experience

With the evolution of high-performance computing, parallel applications have developed an increasing necessity for fault tolerance, most commonly provided by checkpoint and restart techniques. Checkpointing tools are typically implemented at one of two different abstraction levels: at the system level or at the application level. The latter has become an interesting alternative due to its flexibility and the possibility of operating in different environments. However, application-level checkpointing tools often require the user to manually insert checkpoints in order to ensure that certain requirements are met (e.g. forcing checkpoints to be taken at the user code and not inside kernel routines). This paper examines the transformations required to enable automatic checkpointing of parallel applications in the CPPC application-level checkpointing framework. These transformations have been implemented on two very different compiler infrastructures: Cetus and LLVM. Cetus is a Java-based compiler infrastructure aiming to provide an easy to use and clean IR and API for program transformation. LLVM is a low-level, SSA-based toolchain. The fundamental differences of both approaches are analyzed from the structural, behavioral and performance perspectives.     

How mobile phones perform in collaborative augmented reality (CAR) applications

This paper presents the experimental analysis of mobile phones for Augmented Reality marker tracking, a core task that any CAR application must include. The results show that the most time consuming stage is the marker detection stage, followed by the image acquisition stage. Moreover, the rendering stage is decoupled on some devices, depending on the operative system used. This decoupling process allows avoiding low refresh rates, facilitating the collaborative work. However, the use of multicore devices does not significantly improve the performance provided by CAR applications. Finally, the results show that unless a poor network bandwidth makes the network to become the system bottleneck, the performance of CAR applications based on mobile phones will be limited by the detection stage. These results can be used as the basis for an efficient design of CAR systems and applications based on mobile phones.     

Cooperation Between UAS and Wireless Sensor Networks for Efficient Data Collection in Large Environments

This paper describes a method for collection of data from Wireless Sensor Network (WSN) deployed in large environments using Unmanned Aerial Systems (UAS). Unlike existing approaches, in which the WSN and the UAS act as independent units, the main novelty of the proposed method is that UAS and WSN cooperate to increase the performance of the mission. The proposed method presents two main cooperative behaviors: (1) the results of the WSN operation are used to update the UAS flight plan and; (2) the UAS trajectory is considered in the operation of the WSN in order to improve the data collection performance. The proposed method outperforms non-cooperative UAS-based collection approaches and traditional ground multi-hop collection schemes. The method has been experimented in the airfield of Bellavista in Seville (Spain) in March 2011.     

Trajectory Control of a Quadrotor Subject to 2D Wind Disturbances

The paper addresses the flight control of a quad-rotor subject to two dimensional unknown static/varying wind disturbances. A model separation is proposed to simplify the control of the six-degrees-of-freedom (6DOF) nonlinear dynamics of the flying robot. Such approach allows to deal with quad-rotor’s 3D-motion via two subsystems: dynamic (altitude and MAV-relative forward velocity) and kinematic (nonholonomic-like navigation) subsystems. In terms of control, a hierarchical control is used as the overall control structure to stabilize the kinematic underactuaded subsystem. A control strategy based on sliding-mode and adaptive control techniques is proposed to deal with slow and fast time-varying wind conditions, respectively. This choice not only provides well tracking control but also improves the estimation of unknown disturbance. The backstepping technique is used to stabilize the inner-loop heading dynamics, such recursive design takes into account a constrained heading rate. Promising simulations results show the validity of the proposed control strategy while tracking a time-parameterized straight-line and sinusoidal trajectory.     

Tracking in scale quad-rotors through adaptive augmentation

This paper illustrates the application of an adaptive flight control architecture to a scale quad-rotor. For autonomous vertical takeoff and landing flight, it is common to separate the control problem into an inner fast loop that controls attitude and an outer slow loop that controls the trajectory tracking. In this paper, we augment a conventional proportional and derivative controller conceived mainly for hovering, with an adaptive element using a real-time tuning single hidden layer neural network in a inner–outer loop combined architecture to account for model inversion error cancelation, issued in the feedback linearization process. The results shown in simulations reveal the superior performance of the augmented controller in tracking maneuvers.     

EDHMoR: Evolutionary designer of heterogeneous modular robots

This paper is devoted to the problem of automatically designing feasible and manufacturable robots made up of heterogeneous modules. Specifically, the coevolution of morphology and control in robots is analyzed and a particular strategy to address this problem is contemplated. To this end, the main issues of this approach such as encoding, evaluation or transfer to reality are studied through the use of heterogeneous modular structures with distributed control. We also propose a constructive evolutionary algorithm based on tree-like representations of the morphology that can intrinsically provide for a type of generative evolutionary approach. The algorithm introduces some new elements to smooth the search space and make finding solutions much easier. The evaluation of the individuals is carried out in simulations and then transferred to real robots assembled from the modules considered. To this end, the extension of the principles proposed by classical authors in traditional evolutionary robotics to brain–body evolution regarding how simulations should be set up so that robust behaviors that can be transferred to reality are obtained is considered here. All these issues are analyzed by means of an evolutionary design system called EDHMoR (Evolutionary Designer of Heterogeneous Modular Robots) that contains all the elements involved in this process. To show practical evidences of the conclusions that have been extracted with this work, two benchmark problems in modular robotics are considered and EDHMoR is tested over them. The first one is focused on solving a linear robot motion mission and the second one on a static task of the robot that does not require displacements.     

Statistical inference in a redesigned Radial Basis Function neural network

A Hybrid Learning Process method was fitted into a RBF. The resulting redesigned RBF intends to show how to test if the statistical assumptions are fulfilled and to apply statistical inference to the redesigned RBFNN bearing in mind that it allows to determine the relationship between a response (to a process) and one or more independent variables, testing how much each factor contributes to the total variation of the response is also feasible. The results show that statistical methods such as inference, Residual Analysis, and statistical metrics are all good alternatives and excellent methods for validation of the effectiveness of the Neural Network models. The foremost conclusion is that the resulting redesigned Radial Basis Function improved the accuracy of the model after using a Hybrid Learning Process; moreover, the new model also validates the statistical assumptions for using statistical inference and statistical analysis, satisfying the assumptions required for ANOVA to determine the statistical significance and the relationship between variables.     

A novel automatic algorithm for the segmentation of the lumen of the carotid artery in ultrasound B-mode images

A novel algorithm is proposed for the segmentation of the lumen and bifurcation boundaries of the carotid artery in B-mode ultrasound images. It uses the image contrast characteristics of the lumen and bifurcation of the carotid artery in relation to other tissues and structures for their identification. The relevant ultrasound data regarding the artery presented in the input image is identified using morphologic operators and processed by an anisotropic diffusion filter for speckle noise removal. The information obtained is then used to define two initial contours, one corresponding to the lumen and the other one regarding the bifurcation boundaries, for the application of the Chan-Vese level set segmentation model. A set of longitudinal ultrasound B-mode grayscale images of the common carotid artery was acquired using a GE Healthcare Vivid-e ultrasound system. The results reveal that the new algorithm is effective and robust, and that its main advantage relies on the automatic identification of the carotid lumen, which overcomes the known limitations of the traditional algorithms.