Default Controller Research Articles

Uncertainty-Aware Online Learning of Dynamic Thermal Control in Data Center with Imperfect Pretrained Models

The growing demand for cloud computing and storage necessitates the expansion of Data Centers, thereby increasing energy consumption and environmental footprint. The data-driven dynamic thermal control presents a promising solution to mitigate this issue, as it can offer improved thermal control strategies and address the intricate and stochastic nature of the thermal environment. Nonetheless, existing solutions typically exhibit deficiencies in terms of reliability and practicability. To this end, this paper proposes a feasible uncertainty-aware online learning method with reliable dynamics models and controllers. Specifically, Deep Learning models with the capacity for uncertainty quantification are employed to account for the imperfections in pre-trained models, hence providing reliable predictions. The Cross-Entropy Method and Monte Carlo trajectory sampling solve optimization problems while considering uncertainties. It is mathematically analyzed that the uncertainty quantified by the system dynamics model can be captured with sufficient trajectory samples, rendering a reliable controller. A practical framework is proposed and tested on a widely used Data Center model to mitigate the insufficient capacity of Deep Learning models to quantify uncertainties. The results demonstrate its superior performance during the initial training stage, achieving a 66% reduction in violations and 3.76% power savings compared to the default controller. Continuous active online learning enhances its adaptability, resulting in an 80% reduction in violations and a 6% energy saving. The thermal control strategies are thus improved by the proposed new solutions and insights.

Hybrid supervised hierarchical control of a biological wastewater treatment plant.

In wastewater treatment intensification, hierarchical control structures are developed to improve the plant's performance. This paper proposes two novel hybrid supervised hierarchical control structures for specifying the dissolved oxygen concentration in the last aerobic reactor of the wastewater treatment plant (WWTP) based on the nitrification rate and the ammonia level in this reactor. These structures combine the optimum disturbance rejection PI control (OPI), adaptive neuro-fuzzy inference system (ANFIS), and genetic algorithms (GA) to reduce energy consumption and operational costs, improve effluent quality, and reduce the number and percentage of times the established maximum concentration of pollutants in the effluent of the WWTP is violated. The proposed control strategy is implemented and evaluated using benchmark simulation model no. 1 (BSM1). The OPI-ANFIS-GA configuration significantly enhances effluent quality in dry, rainy, and stormy weather conditions, reducing total nitrogen violations by 50.17%, 63.35%, and 47.35%, respectively. Then, 6.79% and 7.12% of aeration energy and 1.44% and 1.46% of operational costs are reduced in dry and rain weather conditions. The OPI-ANFIS configuration enhanced significant energy savings and a cost reduction in storm weather conditions. Both configurations led to a 49.89% decrease in total suspended sludge (TSS) during stormy weather conditions. The proposed controller significantly improves the performance of the WWTP in all weather scenarios compared to the default controller and similar controllers found in the literature.

Dynamic Posture Stabilization Of Humanoid Robot NAO Using 3D-Multilinked Dual Spring-Loaded Inverted Pendulum Model for Uneven and Inclined Floor

Gait pattern performance, for its crucial significance in humanoid robot stabilization, has attracted the attention of researchers worldwide. Although simplified models highlight major features, bipedal walking has bewildered the researchers. Therefore, for a precise understanding of the bipedal model, a state-of-the-art, simplified model has been proposed in this paper which comprises a 3D-multilinked dual spring-loaded inverted pendulum (3D-MDSLIP) while acknowledging the vertical fluctuations of the center of mass (CoM). In addition, the model considers upper body movement and its effects on the stabilization of the humanoid robot. The mathematical modeling of a humanoid walking over the obstacle and slope is demonstrated to precisely understand the problem. The tuning process of the parameters and postures in a humanoid robot is complex and time-consuming. For proper walking of a robot over uneven terrains and slopes, tuning of the PID controller is achieved using converged teaching-learning based optimization (CTLBO) technique for a central pattern generator (CPG) gait, as introduced in the paper. The optimal gait angles are applied to the experimental and simulated NAO to successfully navigate the provided terrain. Thus, the experimental and simulation results jointly show that the proposed CPG-CTLBO gait learning technique is feasible for finding an optimal gait pattern for the humanoid robot within a deviation of 5%. The energy efficiency of the proposed controller is compared with the default controller of NAO based on the average electronic current in sagittal and lateral movement. Further, it is examined for the energy consumption for several slopes, and the results obtained are acceptable, showing the controller is efficient. Additionally, it has been compared with an existing technique for walking a humanoid robot on uneven terrains. The graph obtained using the proposed technique demonstrates the superiority of the proposed technique.

Mitigating an adoption barrier of reinforcement learning-based control strategies in buildings

Reinforcement learning (RL) algorithms have shown great promise in controlling building systems to minimize energy use, operational cost, and occupant discomfort. RL agents learn a control policy by interacting with the physical or simulated environment that represents building systems, occupants, and the outside world. Yet, a large amount of data is needed to learn a near-optimal control policy in the physical building, which requires months or years to collect. Moreover, an agent’s performance while training can be quite poor, causing occupant discomfort and additional costs. Learning in simulation does not have such real-world impacts, but differences between buildings, and indeed between simulation and physical buildings, potentially lead to poor performance when a policy learned in simulation is deployed in a physical building. This paper addresses part of the sim-to-real problem by training on one set of simulated (source) buildings and then deploying to a novel simulated (target) building. This approach significantly reduces the training cost of RL on the target building by 1) learning a large number of policies on prototype buildings, 2) evaluating these policies on historical data obtained from the target building’s environment and selecting the best ones according to the evaluation result, and 3) using the best policies to control the target building while continuing to learn. The proposed approach involves learning a diverse population of control policies using a novel diversity-induced RL algorithm, and policy clustering, evaluation, and selection techniques. Three case studies show our approach assigns policies to the target building that outperform the default controller by 4.0–30.4%, without sacrificing thermal comfort. Similarly, they outperform policies that are learned only on the target building (i.e., without transfer) by 24.9–74.9% and 16.2–72.2% before and after 500 months of training, respectively.

Deep reinforcement learning towards real-world dynamic thermal management of data centers

Deep Reinforcement Learning has been increasingly researched for Dynamic Thermal Management in Data Centers. However, existing works typically evaluate the performance of algorithms on a specific task, utilizing models or data trajectories without discussing in detail their implementation feasibility and their ability to deal with diverse work scenarios. The lack of these works limits the real-world deployment of Deep Reinforcement Learning. To this end, this paper comprehensively evaluates the strengths and limitations of state-of-the-art algorithms by conducting analytical and numerical studies. The analysis is conducted in four dimensions: algorithms, tasks, system dynamics, and knowledge transfer. As an inherent property, the sensitivity to algorithms settings is first evaluated in a simulated data center model. The ability to deal with various tasks and the sensitivity to reward functions are subsequently studied. The trade-off between constraints and power savings is identified by conducting ablation experiments. Next, the performance under different work scenarios is investigated, including various equipment, workload schedules, locations, and power densities. Finally, the transferability of algorithms across tasks and scenarios is also evaluated. The results show that actor-critic, off-policy, and model-based algorithms outperform others in optimality, robustness, and transferability. They can reduce violations and achieve around 8.84% power savings in some scenarios compared to the default controller. However, deploying these algorithms in real-world systems is challenging since they are sensitive to specific hyperparameters, reward functions, and work scenarios. Constraint violations and sample efficiency are some aspects that are still unsatisfactory. This paper presents our well-structured investigations, new findings, and challenges when deploying deep reinforcement learning in Data Centers.

Seamless Transition and Fault-Ride-Through by Using a Fuzzy EO PID Controller in AVR System

One of the most crucial control aspects in electric power networks is ensuring constant voltage levels throughout different circumstances. To overcome this problem, an automatic voltage regulator (AVR) is installed in the electrical power networks to preserve the voltage at its allowable range. In this paper, a fuzzy-based PID controller was used to enhance the AVR dynamic performance under several operating conditions. Moreover, an Equilibrium Optimizer (EO) algorithm was used to initialize the controller gains. The validation of the proposed controller was proven through three different systems. The first was a simple AVR system under fixed and dynamic references. The proposed fuzzy EO PID controller proved its superiority in this case through the reduction of voltage overshoot by around 3–28% compared with the reported methods in the literature. Then, the fault-ride through capability of the proposed controller was proven through the second system, which was the Kundur two-area system suffering from a 3-phase fault condition, where the overshoot with the proposed controller was reduced by 4–7% compared with the default controller and modern weighted method reported in the literature. Lastly, an IEEE 9-bus system performance was tested with the proposed controller under normal, faulty, and dynamic loading conditions. Again, the proposed controller succeeded in reducing the maximum overshoot by around 5% compared with the default controller in the system. Moreover, the proposed controller achieved a seamless transition between the islanding and grid-connected mode of operation.

Multi-objective optimization technique for trajectory planning of multi-humanoid robots in cluttered terrain

Humanoid robots hold a decent advantage over wheeled robots because of their ability to mimic human exile. The presented paper proposes a novel strategy for trajectory planning in a cluttered terrain using the hybridized controller modeled on the basis of modified MANFIS (multiple adaptive neuro-fuzzy inference system) and MOSFO (multi-objective sunflower optimization) techniques. The controller works in a two-step mechanism. The input parameters, i.e., obstacle distances and target direction, are first fed to the MANFIS controller, which generates a steering angle in both directions of an obstacle to dodge it. The intermediate steering angles are obtained based on the training model. The final steering angle to avoid obstacles is selected based on the direction of the target and additional obstacles in the path. It is further works as input for the MOSFO technique, which provides the ultimate steering angle. Using the proposed technique, various simulations are carried out in the WEBOT simulator, which shows a deviation under 5% when the results are validated in real-time experiments, revealing the technique to be robust. To resolve the complication of providing preference to the robot during deadlock condition in multi-humanoids system, the dining philosopher controller is implemented. The efficiency of the proposed technique is examined through the comparisons with the default controller of NAO based on toques produces at various joints that present an average improvement of 6.12%, 7.05% and 15.04% in ankle, knee and hip, respectively. It is further compared against the existed navigational strategy in multiple robot systems that also displays an acceptable improvement in travel length. In comparison in reference to the existing controller, the proposed technique emerges to be a clear winner by portraying its superiority.

Multi-objective trajectory planning of humanoid robot using hybrid controller for multi-target problem in complex terrain

Humanoid robotics is an emerging area of interest in the current engineering research scenario, owing to its ability to impersonate human deportment and emulate various jobs. The given article emphasizes the development and implementation of a hybrid navigational controller to optimize the path length, energy demand, and time spent for accomplishing assigned tasks. The proposed navigational controller is developed by hybridizing the metaheuristic Improved Spider Monkey Optimization (ISMO) approach and the Regression Analysis (RA) approach. Various input parameters like obstacle and target locations are fed to the RA approach that implements a proper navigational direction selection. And it forwards to the SMO approach that is improved using piecewise B-Spline path smoother, which exercises further refinement of the output turning angle and smoothness of path around obstacles. Simulations and real-time experiments are undertaken using different controllers involving single robot systems, which shows the proposed controller’s superiority. An average improvement of 13.72% and 13.94% in path length against RA in simulation and experiment, respectively, and an average improvement of 7.59% and 7.5% in path length against ISMO in simulation and experiment, respectively, is obtained. It is further evaluated for navigation by implementing in a single robot having a multi-target problem. Multiple robot navigation has to deal with the self-collision situations that are solved by prioritizing the specified robot using the dining philosopher controller. It is implemented in the proposed controller for navigation of multiple robots to solve the conflict. Both scenarios are tested in the simulation environment and ratified in the experimental environment. Average deviation under 5% for path length and time spent for single robot navigation and multiple robot navigation is obtained, which shows a good agreement with each other. Energy efficiency test has been performed in contrast to default controller of NAO for various joints, and an average improvement of 8.16%, 5.9% and 20.57%, has been recorded in torque for ankle, knee and hip, respectively. Comparison is carried with an established navigational controller in a similar environmental setup shows an improvement of 8.6% and 10.365%, respectively, in path length and time spent. The results obtained from these setups prove the proposed hybrid controller to be robust, efficient and superior while performing path planning.

Dynamic walking of humanoid robot on flat surface using amplified LIPM plus flywheel model

PurposeHumanoid robots have complicated dynamics, and they lack dynamic stability. Despite having similarities in kinematic structure, developing a humanoid robot with robust walking is quite difficult. In this paper, an attempt to produce a robust and expected walking gait is made by using an ALO (ant lion optimization) tuned linear inverted pendulum model plus flywheel (LIPM plus flywheel).Design/methodology/approachThe LIPM plus flywheel provides the stabilized dynamic walking, which is further optimized by ALO during interaction with obstacles. It gives an ultimate turning angle, which makes the robot come closer to the obstacle and provide a turning angle that optimizes the travel length. This enhancement releases the constraint on the height of the COM (center of mass) and provides a larger stride. The framework of a sequential locomotion planer has been discussed to get the expected gait. The proposed method has been successfully tested on a simulated model and validated on the real NAO humanoid robot.FindingsThe convergence curve defends the selection of the proposed controller, and the deviation under 5% between simulation and experimental results in regards to travel length and travel time proves its robustness and efficacy. The trajectory of various joints obtained using the proposed controller is compared with the joint trajectory obtained using the default controller. The comparison shows the stable walking behavior generated by the proposed controller.Originality/valueHumanoid robots are preferred over mobile robots because they can easily imitate the behaviors of humans and can result in higher output with higher efficiency for repetitive tasks. A controller has been developed using tuning the parameters of LIPM plus flywheel by the ALO approach and implementing it in a humanoid robot. Simulations and experiments have been performed, and joint angles for various joints are calculated and compared with the default controller. The tuned controller can be implemented in various other humanoid robots

Exergy-based control strategies for the efficient operation of building energy systems

Abstract This article addresses the merits of exergy-based control strategies in comparison to energy-based ones in terms of increasing the energetic efficiency of a building and/or decreasing the operating costs of Heating, Ventilation and Air Conditioning (HVAC) systems. In this study, a dynamic exergy model for building energy systems is demonstrated to calculate exergy destruction which is a true measure of thermodynamic inefficiencies in a system. The exergy-based control strategies minimize exergy destruction obtained from the building exergy model while satisfying thermal comfort constraints. In this way, HVAC systems operate in a low-exergy fashion. The results of this study show that the implementation of exergy-based control strategies reduces the quality mismatch between energy supply and demand, and therefore, increases the sustainability of building energy systems. To prove the practicability of the idea, application of exergy-based control strategies in different types of case studies is investigated. The first case study deals with the design of optimal local controllers for the end-users without considering the entire energy supply chain. It is shown that the total operating cost of the system can be decreased by 13% through the implementation of an exergy-based control strategy. The second case study proposes an exergy-based control algorithm that, in addition to the end-users, also considers the upstream components and sub-systems in the energy supply chain. This control algorithm has been tested in a part of the heating system in the main building of the E.ON Energy Research Center (ERC), and the results shows that the proposed control strategy is sufficiently fast and the mass flow rates inside the consumer circuits are relatively stable. The third case study demonstrates a large-scale application, in which an Exergy-based Linear Model Predictive Controller (ExLiMPC) is implemented in a dynamic model representing the working principles of the HVAC and control systems of the main building of the E.ON ERC. A comparison between the default controller and the ExLiMPC shows that the ExLiMPC is able to decrease the overall energy demand of the building by 23.1%.

Grammaticalization of subject agreement on evidence from Italo–Romance

AbstractIn this paper we consider the grammaticalization of subject agreement adducing first-hand synchronic evidence from presentational VS constructions in Italo-Romance dialects. While the existing literature has placed emphasis on the pragmatic properties of the controller, we explore its semantic properties. We argue that variation in subject agreement can only be fully captured with reference to an independently established semantic scale of subjecthood that is based on the position of arguments in semantic representation, and hence on their lexical entailments.Unaffected actoris the default controller in accusative alignment. The patterns of dialect microvariation arise, in our analysis, from variation in macrorole assignment in presentational focus. Our proposal formalizes at the discourse-semantics-syntax interface the idea that, in presentational VS constructions, the core argument S may be treated as part of the predicate, thus failing to control grammatical subject agreement. This happens in presentational focus because the predication is about an implicit topic. The latter can trigger a type of pronominal agreement that is comparable to Bresnan and Mchombo’s (Bresnan, Joan & Sam A. Mchombo. 1987. Topic, pronoun and agreement in Chichewa.Language63. 741–782) anaphoric agreement. This study provides robust arguments in support of an understanding of subject agreement as the grammaticalization of semantic-relation contrasts, as well as pragmatic-role distinctions.

Intelligent HEV Fuzzy Logic Control Strategy Based on Identification and Prediction of Drive Cycle and Driving Trend

Real-time drive cycles and driving trends have a vital impact on fuel consumption and emissions in a vehicle. To address this issue, an original and alternative approach which incorporates the knowledge about real-time drive cycles and driving trends into fuzzy logic control strategy was proposed. A machine learning framework called MC_FRAME was established, which includes two neural networks for self-learning and making predictions. An intelligent fuzzy logic control strategy based on the MC_FRAME was then developed in a hybrid electric vehicle system, which is called FLCS_MODEL. Simulations were conducted to evaluate the FLCS_MODEL using ADVISOR. The simulation results indicated that comparing with the default controller on the drive cycle NEDC, the FLCS_MODEL saves 12.25% fuel per hundred kilometers, with the HC emissions increasing by 22.7%, the CO emissions reducing by 16.5%, the NOx emissions reducing by 37.5% and with the PM emissions reducing by 12.9%. A conclusion can be drawn that the proposed approach realizes fewer fuel consumption and less emissions.

Fault-tolerant model predictive control of a wind turbine benchmark

This paper aims to solve the fault tolerant control problem of a wind turbine benchmark. A hierarchical controller with model predictive pre-compensators, a global model predictive controller and a supervisory controller is proposed. In the model predictive pre-compensator, an extended Kalman Filter is designed to estimate the system states and various fault parameters. Based on the estimation, a group of model predictive controllers are designed to compensate the fault effects for each component of the wind turbine. The global MPC is used to schedule the operation of the components and exploit potential system-level redundancies. Extensive simulations of various fault conditions show that the proposed controller has small transients when faults occur and uses smoother and smaller generator torque and pitch angle inputs than the default controller. This paper shows that MPC can be a good candidate for fault tolerant controllers, especially the one with an adaptive internal model combined with a parameter estimation and update mechanism, such as an extended Kalman Filter.

Task Refinement for Autonomous Robots Using Complementary Corrective Human Feedback

A robot can perform a given task through a policy that maps its sensed state to appropriate actions. We assume that a hand-coded controller can achieve such a mapping only for the basic cases of the task. Refining the controller becomes harder and gets more tedious and error prone as the complexity of the task increases. In this paper, we present a new learning from demonstration approach to improve the robot's performance through the use of corrective human feedback as a complement to an existing hand-coded algorithm. The human teacher observes the robot as it performs the task using the hand-coded algorithm and takes over the control to correct the behavior when the robot selects a wrong action to be executed. Corrections are captured as new state-action pairs and the default controller output is replaced by the demonstrated corrections during autonomous execution when the current state of the robot is decided to be similar to a previously corrected state in the correction database. The proposed approach is applied to a complex ball dribbling task performed against stationary defender robots in a robot soccer scenario, where physical Aldebaran Nao humanoid robots are used. The results of our experiments show an improvement in the robot's performance when the default hand-coded controller is augmented with corrective human demonstration.

Fuzzy Control Model Optimization for Behavior-Consistent Traffic Routing Under Information Provision

This paper presents an H-infinity filtering approach to optimize a fuzzy control model used to determine behavior-consistent (BC) information-based control strategies to improve the performance of congested dynamic traffic networks. By adjusting the associated membership function parameters to better respond to nonlinearities and modeling errors, the approach is able to enhance the computational performance of the fuzzy control model. Computational efficiency is an important aspect in this problem context, because the information strategies are required in subreal time to be real-time deployable. Experiments are performed to evaluate the effectiveness of the approach. The results indicate that the optimized fuzzy control model contributes in determining the BC information-based control strategies in significantly less computational time than when the default controller is used. Hence, the proposed H-infinity approach contributes to the development of an efficient and robust information-based control approach.

Molecular and cellular correlates of the developmental acquisition of mechanisms modulating ingestive behavior

Postnatal development in most mammals is accompanied by the acquisition of controls of ingestion. In rodents, the initial and default controller appears to be gastric stretch. In the second week of life, rat pups acquire the ability to sense the presence of nutrients within the gut and appropriately modulate ingestion. In the third week of life, rat pups start to become weaned from the dam's milk and begin independent ingestion. There have been strong indications that neuropeptide Y is a stimulator of ingestion in adults, although there was very little information in pups. Dr. Gerard Smith initiated a series of studies that provide strong evidence to indicate that hypothalamic neuropeptide Y (NPY) neurons are strong candidates for providing the ability of preweaning rat pups to modulate ingestion according to caloric intake. Moreover, the studies also suggest that the overactivity of hypothalamic NPY neurons presage the onset of hyperphagia in syndromes associated with defects in leptin signaling.