Search

This paper introduces a multi-constraint predictive control algorithm along with a safety layer to guarantee path tracking and object avoidance in emergency situations. A controller switching mechanism is designed which switches the controller between a main and an emergency controller. As the main controller, a nonlinear multi-constraint model predictive controller (MPC) is designed. The MPC algorithm is compared with Stanley and PID methods in terms of their efficiency to validate the MPC as the main controller. However, in unexpected situations, the high computational time of the planner and MPC modules threatens the safety of the vehicle. In order to respond as quickly as possible, emergency braking and maneuver systems are added which could be triggered separately per different situations. Two different emergency scenarios have been implemented in CARLA simulator and Python environment to evaluate the proposed method.

In this study, a passive fault-tolerant controller is designed for the dynamic system based on nonlinear fractional-order theory. The closed-loop control structure contains a linear quadratic regulator for calculating the optimum input signal based on physical criteria as well as nonlinear fractional-order PID. In fact, the summation of the controlling inputs, which are computed by the optimized robust nonlinear fractional-order PID controllers (NFOC), is modified by the LQR unit to achieve higher reliability and robustness. Using optimized robust NFOC along with LQR, the desired controller robustness and performance are evaluated by comparing its quantified performance indicator and optimized simple and fractional-order PID controllers' efficiency. This controller can improve the system’s tracking performance by less amount of control effort and reject the effects of the faulty actuator on the system’s dynamics.

In recent years, brain-computer interfaces (BCI) have gained lots of popularity as an assistive method for individuals with disabilities to regain their communication or mobility and interact with the world more naturally. The utility of such systems rely on their accuracy in decoding user intentions. In this research, we build on previous BCI research that explored CNN models for decoding motor imagery tasks. Inspired by denoising autoencoders, we introduce a deeper CNN model with very high drop out rates that performs better at motor imagery tasks by reducing the effects of EEG data artefacts on the decoder. We tested the model on a hand squeeze data-set where our model has outperformed the previous best classifiers. Additionally, given the deeper nature of this model, we have explored its limitations by comparing its performance as both the individual sample size and the number of training data change.

The classification of motor imagery tasks using scalp EEG signals is a complicated procedure in BCI especially when the task comprises multiple gestures of the same hand. In this paper, we present a classification method to distinguish three grasp motion classes (cylindrical, spherical, and lumbrical) of one hand over two-day training sessions in 15 subjects in a public dataset. We have developed Two ensemble methods consisting of (anomaly detection + fully connected neural network) and (anomaly detection + convolutional neural network) to classify grasp motion and have achieved more than 80% classification accuracy in 3 subjects and an average accuracy of 57% among the full cohort. Our results confirm the possibility of utilizing neural networks to decode motor movement intentions from scalp EEG in a complicated task.

Brain computer interfaces can help individuals with movement disabilities such as locked-in syndrome, tetraplegia and cerebral palsy by providing a direct interface between the brain and various external mobility and assistive devices, such as spellers, wheelchairs, and prosthetics. Inspired by the recent advancements in computer vision, in this paper, we investigate the use of convolutional neural networks in decoding brain signals and evaluate the performance of our models on an existing dataset of EEG signals acquired through a hand squeeze task. Our model outperforms other neural network models previously applied on this dataset both in terms of accuracy, and training speed. Moreover, our model is not limited by factors such as optimal order for the EEG channels and it is patient-invariant. We further investigate the effects of network architecture on decoder performance and training time.

The conventional Wonham-Ramadge supervisory control framework of discrete event systems enforces a closed discrete event system to generate correct behaviors under certain environments, which can be captured by an appropriate plant model. Nevertheless, such control methods cannot be directly applied for many practical engineering systems nowadays since they are open systems and their operation heavily depends on nontrivial interactions between the systems and the external environments. These open systems should be controlled in such a way that accomplishment of the control objective can be guaranteed for any possible environment, which may be dynamic, uncertain and sometimes unpredictable. In this paper, we aim at extending the conventional supervisory control theory to open discrete event systems in a reactive manner. Starting from a novel input-output automaton model of an open system, we consider control objectives that characterize the desired input-output behaviors of the system, based on which a game-theoretic approach is carried out to compute a reactive supervisor that steers the system to fulfill the specifications regardless of the environment behaviors. We present a necessary and sufficient conditions for the existence of such a reactive supervisor. Furthermore, illustrative examples are given throughout this paper to demonstrate the key definitions and the effectiveness of the proposed reactive supervisor synthesis framework.

Erbium-doped fiber amplifier (EDFA) is an optical amplifier/repeater device used to boost the intensity of optical signals being carried through fiber optic communication networks. A highly accurate EDFA model – to predict the signal gain for each channel – is required because of its crucial role in optical network management and optimization. EDFA channel inputs (i.e. features) either carry signal or are idle, therefore they can be treated as binary features. However, channel outputs (and the corresponding signal gains) are continuous values. Labeled training data is very expensive to collect for EDFA devices, therefore we devise an active learning strategy suitable for binary features to overcome this issue. We propose to take advantage of sparse linear models to simplify the predictive model. This approach improves signal gain prediction and accelerates active learning query generation. We show the performance of our proposed active learning strategies on simulated data and real EDFA data.

Formal design of integrated task and motion planning of mobile manipulators in a common workspace, where other agents can also manipulate objects in the workspace, is a challenging research question due to the tight coupling between agents, dynamic environments, and the complexity of mission specifications. We are therefore motivated to propose a multi-layered reactive integrated mission and motion planning for mobile-robotic manipulator systems that are operating in a partially known environment. At the high-level, it receives high-level mission specifications that are expressed in linear temporal logic, and a mission planner constructs a symbolic two-player game between the robots and their environment to synthesize a strategy that adapts to changes in the workspace imposed by other robots. A bilateral synergistic layer is developed to map the designed mission plan to an integrated task and motion planner, which constructs a set of robot tasks to move the objects according to the mission strategy. In the low-level planning stage, verifiable motion controllers are designed that can be incrementally composed to guarantee a safe motion planning for each high-level induced task.

Language identification is an automatic process of detecting the language of a speech utterance. As an application example, in automatic translation technology, before any recognition or translation, the spoken language must be recognized. In this paper we propose two methods for training and testing the language classifiers. One uses Kullback Leibler Divergence (KLD) for improved training of GMMs and the other is the use of Frame Selection Decoding (FSD) for classification. The resulting system leads to significant improvement over the baseline system. Here, acoustic features are extracted directly from speech, and in order to add temporal variations, delta and shifted delta cepstral parameters are added to the features. Our approach has led to a language identification performance of 78.6% among 11 languages using the OGI database and relative reduction error rate of 27.95% when compared with a baseline system employing GMM-UBM for classification.

Virtual metrology in quality control deals with drifts in product quality that occur during non-sampling periods. This approach enables a hundred percent control and improves the precision of statistical control, specially while there is no sampling activity in manufacturing process. The main challenge in virtual metrology is inaccurate predictions. As such, the choice of an appropriate algorithm for prediction is crucial. We compare several algorithms that can be used for prediction in virtual metrology. The comparison over different prediction algorithms is made on a simulated data inspired from virtual metrology application.

This paper aims to investigate the task decomposition problem of multi-agent systems. Task decomposition among agents refers to a process to decompose a given global task into subtasks for individual agents. The decomposition is not arbitrary and should be done in such a way that the satisfaction of the sub-tasks by all agents individually would imply the accomplishment of the global task collectively. In this paper, it is assumed that agents are modeled by labeled transition systems, and the global specification is given as a subclass of Computation Tree Logic (CTL) formulas. It is also assumed that the global CTL specification is broadcasted to and known by all agents. Agents could be heterogeneous and have different capabilities. In order to obtain subtasks for each agent with a maximum potential for fault tolerance, our basic idea is to let each agent contribute to their maximum capabilities in the sense of satisfying a maximum number of sub-formulas of the global specification. The maximum satisfaction set is achieved through a modified CTL model checking algorithm. These maximum satisfiable sub-formulas can be used as the subtask for the corresponding agent. Furthermore, based on assume-guarantee reasoning, sufficient conditions are derived to guarantee the satisfaction of the global CTL specification provided that each agent fullfill its own subtasks. A two-robot cooperative motion planning example is given to illustrate the results.

Potenza is a first generation 64b ARM v8 processor and memory sub-system of the X-GeneTM server platform [1]. The Potenza processor module (PMD) is an integrated design unit, comprising two identical cores sharing a 256KB L2 cache, and is designed to be scalable for different server configurations. Each PMD contains 84 million transistors, occupying over 14.8mm2, and averages 4.5W under representative workloads. The initial platform is configured with 4 PMDs, a shared 8MB L3 cache, and 4 DRAM channels arranged around a central switch. Potenza can operate up to 3GHz at 0.9V supply and is fabricated in a 40nm bulk CMOS technology using 10 metal layers (Fig. 5.8.7).

This paper presents an energy-efficient 3-stage Pseudo Single-phase family of Flip-flops (PSPFF) targeted for use in a 3GHz microprocessor in a 40nm, 0.9V CMOS technology. With latencies in line with the fast pulsed-latch and an average switching energy comparable to the master-slave flip-flop, PSPFF achieves an energy-delay product (EDP) which is 42% and 24% lower than the pulsed-latch and the master-slave flip-flop respectively. Measurement results confirm an improvement of at least 300MHz in operating frequency when using the PSPFF in place of the master-slave flip-flop.

Flying insects possess a surprisingly competent vision and navigation system, which enables them to control various manoeuvres in their flight. In this paper, we have developed a hemispherical omni-directional optical sensor inspired by the compound eye of flying insects. Here, we show the feasibility of a system based on a low complexity optoelectronic system that estimates the optic flow, essentially based on biological findings on the fly Elementary Motion Detectors (EMDs). The valuable properties of our hemispherical compound eye include being lightweight, low power consumption, panoramic field of view and multi resolution. The developed modular system with 128 photoreceptors (up-gradable to 256), with a payload of some hundred grams and power consumption less than 300 mW is light enough to make it suitable to be mounted on UAVs, MAVs and mobile robot applications.

The possibility of employing the demand response in the electricity market environment can lead to an improvement on the power system reliability and unit's production scheduling. To inquire these concepts, two scenarios are defined to schedule day-ahead energy and spinning reserve market simultaneously based on deterministic criteria. In the first scenario, reserve capacity can only be provided by committed generators while in the second scenario, the contribution of demand response in reserve market is permitted. This approach has been implemented over the IEEE 24-bus test system using mixed integer linear programming (MILP) and solved by GAMS software. The achieved results are compared from both economic and technical point of views which indicate the advantages and disadvantages of applying demand response in the system scheduling. Moreover, two different bidding scenarios for demand response are examined.

In this paper a new algorithm for determination of optimal active power reserve capacity requirements and energy and spinning reserve allocation in power system during the 24 hours is presented. In the proposed method, the optimum amount of reserve capacity per hour is determined proportional to network security level required by the system operator. In this way, it is essential to define a proper index for measuring system security. Thus, we have used Expected Load Not Served (ELNS) to evaluate system security in each hour. The proposed method has been implemented over the IEEE 24-bus test system and the results are compared with a deterministic security approach, which considers certain and fixed amount of reserve capacity in each hour. This comparison is done from economic and reliability points of view. The promising results show the effectiveness of the proposed model which is solved using mixed integer linear programming (MILP) and GAMS software is used.

Recently, the controllability problem of multi-agent systems is significantly explored; however, the majority of studies have been focused on the classical controllability approaches. This paper investigates the necessary and sufficient conditions of structural controllability for high order dynamic multi-agent systems. We consider a group of agents in a leader-follower framework under a fixed topology structure. It is assumed that, the agents interconnection is a weighted graph with freely chosen weights and each agent has a high order controllable canonical dynamic. Under this setup we show that the structural controllability of such a network is directly determined by agent interconnection. It is shown that a set of weights can be found which make the entire network controllable if and only if the graph is connected. Finally, we present a numerical example and simulation to illustrate the results.

Nowadays, hyper-spectral remote sensing imaging systems are able to acquire several hundreds of spectral bands. Increasing spectral bands provide the more information for land cover and separate similarity classes and classification accuracy potentially could increase. Nevertheless classification of hyper-spectral imagery by conventional classifiers suffers from Hughes phenomenon. Namely, by increasing spectral bands, for a fixed number of training samples, classification accuracy is reduced. One of the solutions for overcoming the mentioned problem is reducing the dimension of input space based on feature selection techniques. Traditional feature selection techniques have several limitations in performance and finding the global optimum subset selection of feature in hyper-spectral images. In this paper a novel feature selection algorithms based on an Ant Colony Optimization (ACO) presents. ACO techniques are based on the behavior of real ant colonies. Evaluating of obtained results from classification accuracy of AVIRIS image data set shows effectiveness of this algorithm as it achieves fewer features and higher classification accuracy rather than other non-parametric optimization methods such as Genetic Algorithm.

Graphics processing is the driving force behind the demand for high-bandwidth DRAMs. Accelerating the pace of bandwidth improvement, fifth-generation graphics DDRs will operate at data rates up to 5.33Gb/s, and support single-ended signaling for low pin-count. A significant design challenge is to ensure proper signal transmission over single-ended wires at rates previously attainable only with differential pairs. We present single-ended transceiver design techniques for 5.33Gb/s operation. In addition to the receiver and transmitter, a CML-to-CMOS converter and an integrated serializer/level-shifter are described (Fig. 7.5.1). The circuits are fabricated in 0.13µm 1.2V CMOS. The chip area is 5.7×7.0mm2 and is housed in a quadratic BGA package with 289 balls.

Evidence based decision making in the context of chronic disease management requires long term tracking and analysis of patient data. This study demonstrates how disease data tracking can help in understanding underlying patterns in chronic disease progression. Latent Growth Modeling (LGM) is used as a tool to analyze the long term chronic data related to the progression of Multiple Sclerosis (MS). The survey data has been collected on a bi-annual basis by the North American Research Committee on Multiple Sclerosis (NARCOMS), a project of the Consortium of Multiple Sclerosis Centers for the purpose of clinical trial recruitment and epidemiological research. This data set allows for study of MS progression, by measuring three base models: Patient Determined Disease Steps (PDDS), Overall Health and Emotional Health. MS patient data are grouped as early, middle and late disease status. This study analyzes three temporal data points spanning three years and identifies patient traits that are both patient and physician controlled. Empirical evidence confirms many practitioner observations.

This paper builds a new collaborative contract that stresses on long-term relationship between a single supplier and a single retailer in a supply chain. This relationship is based on commitment and trust within the supply chain. Both participants of the supply chain take effort to improve quality. The objective function of the proposed model is to maximize the total revenue of the supplier as well as that of the buyer. The new contract takes the quality issue in addition to the unobservable commitment issue into consideration.

Data recovery and retiming of 4.8Gb/s fully buffered DIMM serial links are described. A 2.4GHz retiming FIFO with an integrated insertion MUX is used to minimize the thru-latency. Fabricated in a 0.13mum 1.5V CMOS technology, the chip occupies 9.2 times 4.5mm2. Using wide-band CML techniques, the input sensitivity with a minimum eye-opening of 0.35UI, is better than 50mVp-pat a BER of 10 -12

The AMD-K7 (TM) processor is an out-of-order, three-way superscalar x86 microprocessor with a 15-stage pipeline, organized to allow 500+MHz operation. The processor can fetch, decode, and retire up to three x86 instructions per cycle to independent integer and floating-point schedulers. The schedulers can simultaneously issue up to nine operations to seven integer and three floating point execution resources. The cache subsystem and memory interface minimize effective memory latency and provide high bandwidth data transfers to and from these execution resources. The processor contains separate instruction and data caches, each 64 kB and two-way set-associative. The data cache is banked and supports concurrent access by two loads or stores, each up to 64 b in length. The processor contains logic to directly control an external L2 cache. The L2 data interface is 64 b wide and supports bit rates up to 2/3 the processor clock rate. The system interface consists of a separate 64 b data bus.