Search

The synchronization requirement is a bottleneck of many distributed optimization algorithms, particularly for solving problems with computationally heterogeneous subproblems and during the occurrence of communication failure/delay. This paper presents a double-loop learning-aided asynchronous alternating direction method of multipliers (LA-ADMM) that has information prediction capability and handles a considerable level of asynchrony between subproblems. A momentum-extrapolation prediction-correction technique is developed to enable subproblems to predict their neighbors missing shared variable information instead of using the latest received values. An online streaming-based anomaly classification is designed to observe the performance of predicted data and control Lagrange multipliers update over the course of iterations. The proposed LA-ADMM reduces under-utilization of computation resources, especially if subproblems are computationally heterogeneous. This algorithm also enhances distributed optimization robustness against communication failure/delay that may result in a considerable level of asynchrony between subproblems. LA-ADMM is applied to solve the optimal power flow problem for several test systems. Promising results are obtained as compared to the classical synchronous ADMM and asynchronous ADMM without the anomaly switch control. [ABSTRACT FROM AUTHOR]

This paper proposes a temporal decomposition strategy to reduce the computation time of security-constrained unit commitment (SCUC). The novelty of this paper is twofold. The scheduling horizon is decomposed into multiple subhorizons. The concept of coupling intervals is introduced, and a set of auxiliary counting variables and logical expressions along with their equivalent linear models are formulated to handle intertemporal ramp constraints and minimum on/off times between consecutive subhorizons. An accelerated analytical target cascading (A-ATC) algorithm is developed to coordinate SCUC subproblems and find the optimal solution for the whole operation horizon in a distributed manner. An initialization strategy is presented to enhance the convergence performance of A-ATC. The proposed algorithm is tested on four test systems. [ABSTRACT FROM AUTHOR]

Simultaneous localization and planning (SLAP) is a crucial ability for an autonomous robot operating under uncertainty. In its most general form, SLAP induces a continuous partially observable Markov decision process (POMDP), which needs to be repeatedly solved online. This paper addresses this problem and proposes a dynamic replanning scheme in belief space. The underlying POMDP, which is continuous in state, action, and observation space, is approximated offline via sampling-based methods, but operates in a replanning loop online to admit local improvements to the coarse offline policy. This construct enables the proposed method to combat changing environments and large localization errors, even when the change alters the homotopy class of the optimal trajectory. It further outperforms the state-of-the-art Feedback-based Information RoadMap (FIRM) method by eliminating unnecessary stabilization steps. Applying belief space planning to physical systems brings with it a plethora of challenges. A key focus of this paper is to implement the proposed planner on a physical robot and show the SLAP solution performance under uncertainty, in changing environments and in the presence of large disturbances, such as a kidnapped robot situation. [ABSTRACT FROM AUTHOR]

We report the design of a two-degree-of-freedom microelectromechanical systems nanopositioner for on-chip atomic force microscopy (AFM). The device is fabricated using a silicon-on-insulator-based process to function as the scanning stage of a miniaturized AFM. It is a highly resonant system with its lateral resonance frequency at ~850 Hz. The incorporated electrostatic actuators achieve a travel range of 16 μm in each direction. Lateral displacements of the scan table are measured using a pair of electrothermal position sensors. These sensors are used, together with a positive position feedback controller, in a feedback loop, to damp the highly resonant dynamics of the stage. The feedback controlled nanopositioner is used, successfully, to generate high-quality AFM images at scan rates as fast as 100 Hz. [ABSTRACT FROM AUTHOR]

The sensitivity of an electrothermal displacement sensor increases with its temperature, whereas a higher temperature range leads to higher thermal noise level, which imposes a tradeoff on the sensor's achievable resolution. We have developed a multiple sensor displacement measurement technique on a 1-degree-of-freedom silicon-on-insulator microelectromechanical systems nanopositioner that mitigates the mentioned tradeoff. To obtain maximum improvement, it is necessary to supply equal power to all of the sensors to ensure equal sensitivity. By combining three identical sensors, we have successfully achieved a 4-dB improvement in signal-to-noise ratio, which is in a good agreement with the averaging theory. Experiments show that the displacement resolution is improved from 0.3 to 0.15 nm/ √(Hz) in the prototype nanopositioner. Furthermore, improvement is possible by increasing the number of sensors around the stage. [ABSTRACT FROM AUTHOR]

Accurate gear defect detection in induction machine-based systems is a fundamental issue in several industrial applications. At this aim, shallow neural networks, i.e., architectures with only one hidden layer, have been used after a feature extraction step from vibration, torque, acoustic pressure and electrical signals. Their additional complexity is justified by their ability in extracting its own features and in the very high-test classification rates. These signals are here analyzed, both geometrically and topologically, in order to estimate the class manifolds and their reciprocal positioning. At this aim, the different states of the gears are studied by using linear (Pareto charts, biplots, principal angles) and nonlinear (curvilinear component analysis) techniques, while the class clusters are visualized by using the parallel coordinates. It is deduced that the class manifolds are compact and well separated. This result justifies the use of a shallow neural network, instead of a deep one, as already remarked in the literature, but with no theoretical justification. The experimental section confirms this assertion, and also compares the shallow neural network results with the other machine learning techniques used in the literature. [ABSTRACT FROM AUTHOR]

The operation of mobile robots in unknown environments typically requires building maps during exploration. As the exploration time and environment size increase, the amount of data collected and the number of variables required to represent these maps both grow, which is problematic since all real robots have finite resources. The solution proposed in this paper is to only retain the variables and measurements that are most important to achieve the robot's task. The variable and measurement selection approach is demonstrated on the task of navigation with a low risk of collision. Our approach has two stages: first, a subset of the variables is selected that is most useful for minimizing the uncertainty of navigation (termed the “focused variables”). And second, a task-agnostic method is used to select a subset of the measurements that maximizes the information over these focused variables (“focused inference”). Detailed simulations and hardware experiments show that the two-stage approach constrains the number of variables and measurements. It can generate much sparser maps than existing approaches in the literature, while still achieving a better task performance—in this case (fewer collisions). An incremental and iterative approach is further presented, in which the two-stage procedure is performed on subsets of the data, and thus, avoids the necessity of performing a resource-intensive batch selection on large datasets. [ABSTRACT FROM AUTHOR]

We present a comprehensive study of the design, modeling, and characterization of an on-chip piezoresistive displacement sensor. The design is based on the bulk piezoresistivity of tilted clamped-guided beams without the need for additional steps to generate doped regions. The sensor is implemented in a one-degree-of-freedom microelectromechanical system (MEMS) nanopositioner, where the beams also function as the suspension system. A standard MEMS fabrication process is used to realize the device on single-crystalline silicon as the structural material. The beams are oppositely tilted to develop tensile and compressive axial forces during stage movement, creating a differential sensing feature. An analytical approach is proposed for modeling and design of the tilted clamped-guided beams. The linearity of the sensor in the differential configuration is investigated analytically. The static, dynamic, and noise characteristics of the sensor are presented, followed by a model-based investigation of the measured dynamic feedthrough. [2015-0030] [ABSTRACT FROM PUBLISHER]

Learning from demonstration provides an attractive solution to the problem of teaching autonomous systems how to perform complex tasks. Reward learning from demonstration is a promising method of inferring a rich and transferable representation of the demonstrator's intents, but current algorithms suffer from intractability and inefficiency in large domains due to the assumption that the demonstrator is maximizing a single reward function throughout the whole task. This paper takes a different perspective by assuming that the reward function behind an unsegmented demonstration is actually composed of several distinct subtasks chained together. Leveraging this assumption, a Bayesian nonparametric reward-learning framework is presented that infers multiple subgoals and reward functions within a single unsegmented demonstration. The new framework is developed for discrete state spaces and also general continuous demonstration domains using Gaussian process reward representations. The algorithm is shown to have both performance and computational advantages over existing inverse reinforcement learning methods. Experimental results are given in both cases, demonstrating the ability to learn challenging maneuvers from demonstration on a quadrotor and a remote-controlled car. [ABSTRACT FROM PUBLISHER]

We report a novel piezoresistive microelectromechanical system (MEMS) differential displacement sensing technique with a minimal footprint realized through a standard MEMS fabrication process, whereby no additional doping is required to build the piezoresistors. The design is based on configuring a pair of suspension beams attached to a movable stage so that they experience opposite axial forces when the stage moves. The resulting difference between the beam resistances is transduced into a sensor output voltage using a halfbridge readout circuit and differential amplifier. Compared with a single piezoresistive flexure sensor, the design approximately achieves 2, 22, and 200 times improvement in sensitivity, linearity, and resolution, respectively, with 1.5-nm resolution over a large travel range exceeding 12 μm. [ABSTRACT FROM AUTHOR]