Current evaluations of Continual Learning (CL) methods typically assume that there is no constraint on training time and computation. This is an unrealistic assumption for any real-world setting, which motivates us to propose: a practical real-time evaluation of continual learning, in which the stream does not wait for the model to complete training before revealing the next data for predictions. To do this, we evaluate current CL methods with respect to their computational costs. We conduct extensive experiments on CLOC, a large-scale dataset containing 39 million time-stamped images with geolocation labels. We show that a simple baseline outperforms state-of-the-art CL methods under this evaluation, questioning the applicability of existing methods in realistic settings. In addition, we explore various CL components commonly used in the literature, including memory sampling strategies and regularization approaches. We find that all considered methods fail to be competitive against our simple baseline. This surprisingly suggests that the majority of existing CL literature is tailored to a specific class of streams that is not practical. We hope that the evaluation we provide will be the first step towards a paradigm shift to consider the computational cost in the development of online continual learning methods.

The revolution of Internet – enabled instrumentation is emerging as a revolution in Measurement and Automation. New standards are being developed for transmitting data and connecting instruments to the Internet. The main purpose of this thesis is to design and develop a system to Integrate various Real – time experiments and be able to monitor and control them over Internet using LabVIEW. LabVIEW is a graphical programming package capable of data acquisition, data analysis, data presentation and real time remote control. In this thesis a Real – time system is developed which integrates several real time experiments and remote control access over Internet is provided using LabVIEW. The latest remote panel technology is used to provide the remote control access. There are four Real – time instrument experiments developed in this thesis project. They are Automatic Mixing Controller, The Digital Storage Oscilloscope from Gould Instruments, Temperature Controller and an Electronic Recording Rain/Precipitation Gauge. These four instruments are connected and communicated from the main computer by external Data Acquisition Board (DAQ) for the Automatic Mixing Controller, By General Purpose Interface Board (GPIB) for the Oscilloscope, by Data Acquisition board (DAQ) for the Temperature controller and by Serial port for the Electronic Recording Rain/Precipitation Gauge. A system is developed to integrate all the applications listed above into one application and are monitored and controlled remotely over Internet using LabVIEW.

Organic contamination on spaceflight hardware is an ongoing concern for spaceflight safety. In addition, for the goal of analyzing for possible evidence of extra-terrestrial life, it is necessary to consider the presence of terrestrial contamination. This paper will introduce and evaluate a new method using a direct analysis real time (DART) ionization source paired with a high resolution time of flight mass spectrometer (TOFMS) for the determination of organic contamination involved in spaceflight hardware and ground support materials. This novel analytical technique has significant advantages over current methodologies. Materials analyzed in this study were historically considered as probable contaminants in spaceflight related substrates. A user determined library was generated due to the non-traditional mass spectra generated by the DART. Continual improvement of analytical methods for the detection of trace levels of contaminants in potential drinking water sources is of extreme importance to both regulatory communities and concerned citizens. This paper will evaluate a novel analytical method using stir bar sorbtive (SBSE) extraction techniques combined with analysis with a DART TOFMS. Compounds of interest will include several representative pharmaceutical contaminants of emerging concern listed in EPA method 1694. Optimal SBSE and DART experimental parameters will be investigated along with accuracy, precision, limits of detection and calibration linearity.

In order to better understand the changes occurring in the internal environment of the pyrolysis process a method of monitoring the internal environment in real time is the key objective of this study. To accomplish this objective four tasks were laid out in order to develop an effective way of monitoring the changes in gases present as pyrolysis is occurring as well as in material activation processing. For all processing the self-activation process was used which combines pyrolysis and thermal activation into a single step process. In the first task 10 hard wood species were activated and the resulting properties were compared to see the impact of wood species on the resulting carbon structures. In order to understand the impact of gas concentration on the resulting carbons the second task developed a gas sensor array which effectiveness was corroborated using GC-MS and then comparisons of the changes in the resulting were made. For the third task the gas sensor array was used to analyze the production of CO2 gas and a triple Gaussian model was developed to model the changes in gas production throughout processing. H2 gas production was modeled in the fourth task using the same Gaussian model as the third, where the results of both gas productions were compared showing the impact of processing parameters on gas production. With these four tasks completed we can see how our processing effects wood species similarly but at different rates, gas concentration was linked to changes in carbon structure, the effectiveness of our sensor was proven, a triple Gaussian model was developed to around gas production, and the impact of processing parameters on gas production was observed. With this Information a link between resulting carbon structure and gas content of the pyrolysis can be done and the changes in the pyrolysis environment were monitored in real time.

Industry 4.0 is associated with the technological complexity of designing frameworks for factory automation. In that case, since the robots are used in factory operations, this paper proposes a framework that can be used for (near) real-time applications (RTA) in the robotic system. Real-time application is a time constraint that works within a time frame, making it essential to set up a high-speed system in data computation and processing. Monitoring sensors are exposed to different physical variables, such as noise and vibration temperature, from the system, which leads to inefficient data and delay in the data computations. Adaptive Sampling Algorithm A.S.A. was used to reduce the amount of data to be computed. A self-adaptive software (i.e., Rainbow framework) was used to implement the algorithm—Hardware-in-the-loop (HiL) simulation technique used in the simulation of A.S.A. The hardware used in this scenario is the Speedgoat real-time target machine. The proposed framework was tested in an implementation scenario where the robot system had high latency, above 10ms, and had to be reduced to 5ms and below. The results showed fewer samples were collected from the test signal after implementing the algorithm. Hence reducing high latency and increasing real-time application in robot systems. In summary, the proposed framework could be used to develop real-time applications in robotic systems for Industry 4.0.

Many types of devices based on power converters have been proposed and studied for utility applications. In recent years most of the control systems for these converters have been digital. Unfortunately, such digital controllers, which are often based on a digital signal processor (DSP), are difficult to model in simulation. Thus, hardware prototypes are usually required. This thesis presents a tool for fast prototyping that helps overcome these difficulties. Namely, a hardware-in-the-loop simulation is provided for the digital controller in order to evaluate control algorithms without the voltage source converter and power system. The controller in the loop design methodology is described and the division between the real-time power system model and the hardware controller with an interface is shown. Also, the modulation type, integration time step selection and synchronization between the controller and the real-time system simulation are discussed. The hardware configuration for the real-time simulator and the software implementation of the simulator is discussed. In this thesis an example application of a shunt active compensator following this formal procedure is presented. The active compensator prototyping was first developed in MATLAB/Simulink. Then, following a formal design procedure, the power system was modeled in a digital simulator and the controller was implemented in a digital controller board. Finally, a hardware-in-the-loop test was carried out to validate the performance of the hardware controller for the active compensator. Although the tools and methods presented here are aimed at shunt connected current controller application, they may be generalized for use in the development of any digitally controlled power electronic converter.

A cornerstone of next generation intelligent transportation systems (ITS) is a seamless integration of in-vehicle networking with existing wireless telephony infrastructure. Remote access to on-board diagnostics and performance data is a crucial requirement for ITS. This thesis investigated the performance benefits of a data integrity buffering technique for an extensible vehicle position and performance tracking system (VPPTS). In support of this investigation, a VPPTS prototype was developed. The prototype used available technologies and interfaces to industry-standard communications channels and is a demonstration of a next-generation intelligent transportation system (ITS). The data integrity buffering technique under investigation was shown to provide quantitative improvements in successful VPPTS data transmissions. The use of this technique addressed important deficiencies in real-time data transmission for these types of systems over wireless networks.

One of the major objectives at Mississippi State University’s Power and Energy Research Laboratory (PERL) is to develop an adaptive protective controller for Shipboard Power System (SPS) protection. This thesis work focuses on developing an overcurrent relay model in LabVIEW software and validating the developed model by conducting Hardware-in-the-Loop (HIL) tests with Real-Time Digital Simulator (RTDS) and commercial Schweitzer Engineering Laboratories (SEL)-351S directional over-current relay. Simulation results show that the developed relay model is quite flexible, efficient and can be used in real-time. Modeling efforts to establish a HIL platform using National Instruments devices have been presented here. This thesis work also proposes a high-performance and low-cost National Instruments-PXI platform for power system simulations. Two-bus, eight-bus and shipboard power system (SPS) test cases are developed using Matlab/Simulink.Software-in-the-Loop (SIL) tests are conducted for these test cases with Matlab/Simulink overcurrent relay model for several fault conditions. To determine the performance of the NI-PXI system, open loop tests are done between the NI-PXI and the SEL-351S relay and these results are compared with the results of open loop test conducted between the RTDS and SEL-351S relay. HIL tests are done between the NI-PXI system and the dSPACE relay model. HIL tests are also done between the NI-PXI and the commercial SEL-351S relay. These results show that the NI-PXI controller can be used as a power system simulator.

No description available.

No description available.