Anytime Control Algorithms for Embedded Real-Time Systems

In this paper, we consider the problem of designing controllers for linear plants to be implemented in embedded platforms under stringent real-time constraints. These include preemptive scheduling schemes, under which the execution time allowed for control software tasks is uncertain. In a conservative Hard Real-Time (HRT) design approach, only a control algorithm that (in the worst case) is executable within the minimum time slot guaranteed by the scheduler would be employed. In the spirit of modern Soft Real-Time (SRT) approaches, we consider here an "anytime control" design technique, based on a hierarchy of controllers for the same plant. Higher controllers in the hierarchy provide better closed-loop performance, while typically requiring longer execution time. Stochastic models of the scheduler and of algorithm execution times are used to infer probabilities that controllers of different complexity can be executed at different periods. We propose a strategy for choosing among executable controllers, maximizing the usage of higher controllers, which affords better exploitation of the computational platform than the HRT design while guaranteeing stability (in a suitable stochastic sense). Results on the robustness with respect to uncertainties affecting the scheduler model, and on bumpless transfer for tracking problems are also reported. Simulation results on the control of two prototypical mechanical systems show that performance is substantially enhanced by our anytime control technique w.r.t. worst case-based scheduling.

When designing today's highly complicated systems consisting of several hardware and software modules, it is essential to estimate the performance such as worst-case or best-case execution time in early design stages. Such estimation is essential to explore architecture and hardware/software partitioning in system-level design. A maximum execution time estimated topologically without considering false-paths is longer than the real. In this paper, we propose an static estimation method of maximum execution time in system-level designs, considering false-paths. Also, we adopt an approximation approach in order to avoid the path explosion problem. The experimental results show that our method can provide much smaller estimated maximum execution time than the method without considering false-paths. At the same time, the results show us that the maximum execution time can be estimated to a very small range, by applying both simulation-based method and our static method.

Negotiations were introduced in [6] as a model for concurrent systems with multiparty decisions. What is very appealing with negotiations is that it is one of the very few non-trivial concurrent models where several interesting problems, such as soundness, i.e. absence of deadlocks, can be solved in PTIME [3]. In this paper, we introduce the model of timed negotiations and consider the problem of computing the minimum and the maximum execution times of a negotiation. The latter can be solved using the algorithm of [10] computing costs in negotiations, but surprisingly minimum execution time cannot.This paper proposes new algorithms to compute both minimum and maximum execution time, that work in much more general classes of negotiations than [10], that only considered sound and deterministic negotiations. Further, we uncover the precise complexities of these questions, ranging from PTIME to varDelta _2^P-complete. In particular, we show that computing the minimum execution time is more complex than computing the maximum execution time in most classes of negotiations we consider.

The application scheduling is an important requirement in the cloud computing context. It allows to define the required resources to execute applications tasks following predefined criteria, for instance, maximum execution time, number of virtual machines, volume of data, among others. Selecting process to choose the most appropriate execution structure is driven by scheduling algorithms. This paper proposes a scheduling mechanism for data processing in cloud computing environments. Such mechanism analyzes some specific variables in the business context of a software house specialized in software for lawyers and law offices. The main goal of this mechanism is to fulfill the seasonal company's demand using IaaS services and considering two policies: (i) the maximum execution time allowed by the application may not be exceeded and (ii) the data have to be processed considering the lowest possible monetary cost. The proposed solution generates strategies to select the best set of virtual machines to process the current bunch of data considering the amount of data, the estimated execution time for each specific strategy and the monetary cost of the virtual machines sets. In the context of this work, the strategy concept means the schedule of a set of virtual machines to process a specific amount of data, load balancing decisions and the parallelism of application's execution flow. The proposed solution has resulted in great impact for that company since it allowed the vertiginous increase of the amount of clients served.

PREEMPTIVE AND NON-PREEMPTIVE REAL-TIME SCHEDULING BASED ON NEURAL NETWORKS

One of the major goals of the Service Oriented Architecture is to support automatic Web Service (WS) composition and execution, allowing a user query to be satisfied by a Composite WS (CWS). User queries express functional and non-functional (QoS) requirements. QoS requirements indicate, for example, the maximum execution time or price allowed for a CWS execution. In this work, we propose a model to support CWS executions while maintaining the QoS requirements, even in presence of failures. QoS monitoring is performed during the entire execution of a CWS in order to satisfy QoS requirements, influencing the choice of the fault-tolerance strategy selection in case of failures.

Scheduling and resource allocation problems require numerous constraints and objectives. Latency constraints and network parameters have to be considered explicitly to produce realistic scheduling schemes. Finding an optimum scheduling algorithm is classified in the category of NP-Hard problems. Therefore, heuristic approaches are used to produce near optimal solutions with a variety of configurations. The heuristic approach presented in this paper is formulated in the framework of graph theory considering a processor network composed of a bounded number of non identical parallel processors. Every processor is able to execute only one, specific subset of tasks, having enough memory to store the communication data. The starting procedure of the algorithm uses a depth first search to find the set of all paths from source to sink. The principle procedure creates combinations of processors and task paths, seeking to allocate sets of adjacent tasks to the same compatible processor. Task selection is based on the maximum total execution time by simultaneously reducing the processor's idle time. The paper ends with a small detailed numerical example.

A fundamental question in the field of communication-limited control is how low the closed loop data rate can be made before a given dynamical system is impossible to stabilise by any coding and control law. Analogous to the role of entropy in Shannon source coding, this number defines the smallest data rate sufficient to achieve 'reliable' closed loop performance. The objective here is to analyse this quantity for a general, finite-dimensional, discrete-time stochastic linear plant. By inductive arguments employing the entropy power inequality of information theory, an explicit expression for the infimum mean-square-stabilising data rate is derived, under very mild conditions on the initial state and noise probability densities.

Real-time evacuation guidance is a critical safety function for intelligent Cyber-Physical Systems (CPS). The core challenge lies in executing complex, multi-objective path finding algorithms like Iterative Dynamic Programming (IDP) under stringent real-time constraints. This paper presents a scalable, accelerated framework designed around “EnMod-DP,” a novel suite of hybrid algorithms that integrate Dynamic Programming with Reinforcement Learning (RL) heuristics. We evaluate 20 algorithms across varying grid sizes from 5×5 to 100×100. Our results demonstrate that Parallel EnMod-DP solvers achieve significant acceleration on large-scale grids, with the Parallel Hybrid DP-RL solver reducing execution time by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\approx 45 {\%}$</tex> compared to serial baselines. The primary novelty of this work lies in its engineering implementation: bridging the gap between theoretical optimality and real-time viability through a robust, parallelized architecture deployable on edge systems.

SummaryThe process of node identification is referred to as localization, and it is rapidly gaining popularity in the field of WSN. Different node identification processes have different findings, benefits, challenges, costs, effectiveness, and applications. In this work, the position error of the Amorphous algorithm is minimized by optimizing the hop size. For optimization of the hop size of the Amorphous algorithm, two different optimization algorithms, such as ALO and GWO, are considered. Proposed Amorphous‐ALO and Amorphous‐GWO provide higher accuracy rates of 33.89% and 4.22% than traditional Amorphous as well as ensemble approaches. Amorphous‐ALO and Amorphous‐GWO provide position errors 2.9161 and 2.9164 respectively, which are very similar. Therefore, to determine the suitable optimization algorithm for Amorphous, the minimum, average, and maximum execution times of Amorphous‐ALO and Amorphous‐GWO are considered. The approach that has less execution time is considered as most suitable for Amorphous. Amorphous‐ALO takes 67.76, 69.60 and 84.24 s for minimum, average and maximum execution whereas Amorphous‐GWO takes 65.17, 65.46 and 65.66 s for minimum, average and maximum execution respectively. As Amorphous‐GWO takes less execution time than Amorphous‐ALO; therefore, GWO is more suitable for optimization in Amorphous algorithm.

Induction machines have become the standard for highly demanding industrial applications. This has led to the utilization of modern discrete-time control techniques (such as model predictive control) that require the estimation of internal variables that are not subject to measurement (such as the rotational velocity in sensorless applications). From this point of view, it is fundamental to have accurate discrete-time models of induction machines, particularly given their nonlinear nature, so that control techniques perform according to design requirements. In spite of the above, the modeling of induction machines has not received much attention in the literature, even though more powerful machines and faster microcontrollers are currently being used. To better understand induction machine models for control, in this paper, we develop and compare various discrete-time models of induction machines based on Euler, Taylor, and Runge–Kutta methods. In addition, we compare the Extended Kalman Filter and Unscented Kalman Filter for state estimation in terms of accuracy and computational burden. The models are derived and compared through extensive Monte Carlo simulations and the state estimation techniques are compared in terms of root mean squared error, execution time, and maximum absolute error. Our simulations show that, in general, the Taylor method yields more accurate models than both the Runge–Kutta and Euler methods. In particular, the Taylor method results in a root mean square error that is one order of magnitude smaller than the Euler method for stator current and rotor flux linkages. For rotor angular speed, the Runge–Kutta methods are more accurate than both the Taylor and Euler methods, resulting in a root mean square error that is two orders of magnitude smaller than the Euler method. On the other hand, the Extended Kalman Filter results in smaller execution time than the Unscented Kalman Filter, up to two orders of magnitude. In terms of root mean squared error and maximum absolute error, both filtering algorithms perform similarly.

Most previous research done in probabilistic schedulability analysis assumes a known distribution of execution times for each task of a real-time application. This is however not trivial to determine it with a high level of confidence. Methods based on measurements are often biased since not in general exhaustive on all the possible execution paths, whereas methods based on static analysis are mostly worst-case execution time - WCET-oriented. Using static analysis, this work proposes a method to obtain probabilistic distributions of execution times. It assumes that the given real time application is divided into multiple tasks, whose source code is known. Ignoring in this paper hardware considerations and based only on the source code of the tasks, the proposed technique allows designers to associate to any execution path an execution time and a probability to go through this path. A source code example is presented to illustrate the method.

Evaluation of Level of Confidence and Optimization of Roll-back Recovery with Checkpointing for Real-Time Systems

As there is a rise in number of academic institutes all over the world, shortlisting the universities to apply has become a tedious task for aspiring graduate student. There are many crucial factors involved while submitting the application in any university. The cost involved while applying for any university is one of them. If student profile gets rejected, this may lead to wastage of time and money both. In order to handle this problem, we developed a machine learning-based admission prediction system. This proposed prediction system considers various parameters such as English language score, university rank, statement of purpose, letter of recommendation, cumulative grade point average, and research experience for predicting the chances of admission in a university using ensemble model. Total of 13 machine learning-based prediction models were trained and tested. These 13 models were divided into three categories: baseline models, ensemble models, and deep models. Different performance parameters such as root mean squared error, mean squared logarithmic error, and R2 were used. Execution time for each model was also recorded. In category wise comparison, ensemble models outperformed baseline models and deep models. While in total of 13 models, multiple linear regression and ridge regression outperformed all other models (in terms of high-R2 score, less root mean square error, and less execution time) followed by gradient boosting regression and extra tree regression. KeywordsBaseline modelsDeep modelsEnsemble modelsGraduate admissionPredictionRegression and ridge regression

Speech enhancement is a vital area of research, the performance of speech based human machine applications such as automatic speech recognition system, in car communication depends on the quality of speech communicated. Different methodologies have been used by various researchers to improve the quality of speech signal. In this paper an attempt is made to analyze the performance of Least Mean Square (LMS) and Recursive Least Squared (RLS) adaptive filter algorithm for speech enhancement application. The performance indices used for the evaluations is Mean Square Error (MSE), Signal to Noise Ration (SNR) and execution time. The detail analysis is done and experimentally the results are validated and certain modifications are suggested in the algorithm. The experimentation revels that LMS have fast convergence than RLS. The computational complexity of RLS is very high as compared to LMS.