Length Of Horizon Research Articles

A robust constrained model predictive control scheme for norm‐bounded uncertain systems with partial state measurements

SummaryA robust model predictive control scheme for a class of constrained norm‐bounded uncertain discrete‐time linear systems is developed under the hypothesis that only partial state measurements are available for feedback. The proposed strategy involves a two‐phase procedure. Initialization phase is devoted to determining an admissible, though not optimal, linear memoryless controller capable to formally address the input rate constraint; then, during on‐line phase, predictive capabilities complement the designed controller by means of N steps free control actions in a receding horizon fashion. These additive control actions are obtained by solving semidefinite programming problems subject to linear matrix inequalities constraints. As computational burden grows linearly with the control horizon length, an example is developed to show the effectiveness of the proposed approach for realistic control problems: the design of a flight control law for a flexible unmanned over‐actuated aircraft, where the states of the flexibility dynamics are not measurable, is discussed, and a numerical implementation of the controller within a nonlinear simulation environment testifies the validity of the proposed approach and the possibility to implement the algorithm on an onboard computer.

Multi-step reinforcement learning for model-free predictive energy management of an electrified off-highway vehicle

The energy management system of an electrified vehicle is one of the most important supervisory control systems which manages the use of on-board energy resources. This paper researches a ‘model-free’ predictive energy management system for a connected electrified off-highway vehicle. A new reinforcement learning algorithm with the capability of ‘multi-step’ learning is proposed to enable the all-life-long online optimisation of the energy management control policy. Three multi-step learning strategies (Sum-to-Terminal, Average-to-Neighbour Recurrent-to-Terminal) are researched for the first time. Hardware-in-the-loop tests are carried out to examine the control functionality for real application of the proposed ‘model-free’ method. The results show that the proposed method can continuously improve the vehicle’s energy efficiency during the real-time hardware-in-the-loop test, which increased from the initial level of 34% to 44% after 5 h’ 35-step learning. Compared with a well-designed model-based predictive energy management control policy, the model-free predictive energy management method can increase the prediction horizon length by 71% (from 35 to 65 steps with 1 s interval in real-time computation) and can save energy by at least 7.8% for the same driving conditions.

THE USAGE THE LYAPUNOV INDICATOR FOR DETERMINATION OF DROPS

At this time, there is a widespread use of numerical control lathes, which aims to develop technical solutions for automatic control of the state of the technological system, in particular the tool. In the process of cutting, the cutter fails as a result of a dull (end of durability period), breaks (complete destruction of the cutting edge or crumbles into particles of cutting material), softens (loss of hardness in steel and welding are almost unobservable in hard alloy) and, in the end, the result is increased wear.At the same time, the main requirement is a higher reliability of the results obtained, this is largely determined by the chosen methods of information analysis, which characterizes the state of the technological system. Durability of the tool should provide high performance and the lowest material costs associated with the manufacture of the product. This resistance is called economic. Its approximate values for different cutting tools, adopted in the standards for cutting conditions, are given in the relevant directories.The most promising methods are synergistic. One of the most important results of synergetics was the conclusion that the length of the time horizon of the prediction of circulation was limited even for relatively simple systems. This applies to systems that are sensitive to initial conditions. The main task is to choose a characteristic or group of characteristics by which it is supposed to evaluate the state of the processing system and establish the dependence of the selected characteristic on time. As the initial data it is supposed to use a vibrosignal which is generated by the process of turning, as the initial data – wear of the cutter.Calculations of cross-correlation showed that its value reaches 0.72 Ra/μm, which indicates the presence of a relationship between the values. Three characteristic areas can be identified. In the first region there is an intensive increase in the values of the Lyapunov index. In the second area, there are slight fluctuations in the indicator values, and when a certain wear chamfer threshold is reached - changes in the Lyapunov indicator values are extremely small (about 10-5 ... 10-6) – this area corresponds to a period of catastrophic wear.As established during the study was observed in all the experiments carried out at the finishing turning. Analyzing the obtained results, we can conclude that the use of the Lyapunov indicator is justified for tracking the pathological changes in the mill of the technological processing system – that is, by the value of the Lyapunov indicator, one can judge the occurrence or non-occurrence of catastrophic tool wear.

Prediction of Near-Crashes from Observed Vehicle Kinematics using Machine Learning

This study introduces a machine learning model for near-crash prediction from observed vehicle kinematics data. The main hypothesis is that vehicles tend to experience discernible turbulence in their kinematics shortly before involvement in near-crashes. To test this hypothesis, the SHRP2 NDS vehicle kinematics data (speed, longitudinal acceleration, lateral acceleration, yaw rate, and pedal position) are utilized. Several machine learning algorithms are trained and comparatively analyzed including K nearest neighbor (KNN), random forest, support vector machine (SVM), decision trees, Gaussian naïve Bayes (Gaussian NB), and adaptive boost (AdaBoost). Sensitivity analysis is performed to determine the optimal prediction horizon length (the time period before the occurrence of a near-crash) and the turbulence horizon length (the time period during which near-crash related changes in vehicle kinematics take place). The results indicate that optimal prediction performance can be achieved at a 1 s prediction horizons and a 3 s turbulence horizon. At these values, the AdaBoost model outperforms all other models in relation to its recall (100%), precision (98%), and F1-score (99%). These values imply that the near-crash prediction model is highly efficient in predicting most instances of near-crashes with minimal false near-crash predictions. This promising prediction performance offers a viable tool for supporting crash avoidance systems in the emerging connected/autonomous vehicle technology.

Dynamic Learning of Sequential Choice Bandit Problem under Marketing Fatigue

Motivated by the observation that overexposure to unwanted marketing activities leads to customer dissatisfaction, we consider a setting where a platform offers a sequence of messages to its users and is penalized when users abandon the platform due to marketing fatigue. We propose a novel sequential choice model to capture multiple interactions taking place between the platform and its user: Upon receiving a message, a user decides on one of the three actions: accept the message, skip and receive the next message, or abandon the platform. Based on user feedback, the platform dynamically learns users’ abandonment distribution and their valuations of messages to determine the length of the sequence and the order of the messages, while maximizing the cumulative payoff over a horizon of length T. We refer to this online learning task as the sequential choice bandit problem. For the offline combinatorial optimization problem, we show a polynomialtime algorithm. For the online problem, we propose an algorithm that balances exploration and exploitation, and characterize its regret bound. Lastly, we demonstrate how to extend the model with user contexts to incorporate personalization.

Single-agent Finite Impulse Response Optimizer vs Simulated Kalman Filter Optimizer


 
 
 
 Single-agent Finite Impulse Response Optimizer (SAFIRO) is a new estimation-based optimization algorithm which mimics the work procedure of the ultimate unbiased finite impulse response (UFIR) filter. In a real UFIR filter, the horizon length, N, plays an important role to obtain the optimal estimation. In SAFIRO, N represents the repetition number of estimation part that needs to be done in find-ing an optimal solution. On the other hand, Simulated Kalman Filter (SKF) is also an estimation- based optimization algorithm inspired by the estimation capability of Kalman filtering. In literature, substantial amount of works has been devoted to SKF, both in applied research and fundamental enhancements. Thus, in this paper, a performance comparison of both SAFIRO and SKF is presented. It is found that the SAFIRO outperforms the SKF significantly.
 
 
 

On the design of online production scheduling algorithms

We discuss a systematic way to design online scheduling algorithms to achieve better closed-loop performance. Specifically, we address short-term scheduling, and focus on determining the horizon length and re-optimization time-step, and explore objective-function modifications and added constraints. First, we show that finding open-loop and closed-loop schedules are two different problems and that there is considerable potential to improve the closed-loop performance, but this task is non-trivial. Second, we outline and quantitatively define three key production system characteristics that are pertinent to the design of online scheduling algorithms: (i) production load; (ii) order-size max-mean relative difference; and (iii) time-constants. Third, we discuss how these affect the choice of the re-optimization time-step and the horizon length. Fourth, we evaluate the role of demand uncertainty and how it can be mitigated through better tuning of the above algorithmic parameters. Finally, we present a framework that summarizes the design procedure for online scheduling algorithms.

Advanced Vehicle State Monitoring: Evaluating Moving Horizon Estimators and Unscented Kalman Filter

Active safety systems must be used to manipulate the dynamics of autonomous vehicles to ensure safety. To this end, accurate vehicle information, such as the longitudinal and lateral velocities, is crucial. Measuring these states, however, can be expensive, and the measurements can be polluted by noise. The available solutions often resort to Bayesian filters, such as the Kalman filter, but can be vulnerable and erroneous when the underlying assumptions do not hold. With its clear merits in handling nonlinearities and uncertainties, moving horizon estimation (MHE) can potentially solve the problem and is thus studied for vehicle state estimation. This paper designs an unscented Kalman filter, standard MHE, modified MHE, and recursive least squares MHE to estimate critical vehicle states, respectively. All the estimators are formulated based upon a highly nonlinear vehicle model that is shown to be locally observable. The convergence rate, accuracy, and robustness of the four estimation algorithms are comprehensively characterized and compared under three different driving maneuvres. For MHE-based algorithms, the effects of horizon length and optimization techniques on the computational efficiency and accuracy are also investigated.

A game-theoretic approach for selecting optimal time-dependent thresholds for anomaly detection

Adversaries may cause significant damage to smart infrastructure using malicious attacks. To detect and mitigate these attacks before they can cause physical damage, operators can deploy anomaly detection systems (ADS), which can alarm operators to suspicious activities. However, detection thresholds of ADS need to be configured properly, as an oversensitive detector raises a prohibitively large number of false alarms, while an undersensitive detector may miss actual attacks. This is an especially challenging problem in dynamical environments, where the impact of attacks may significantly vary over time. Using a game-theoretic approach, we formulate the problem of computing optimal detection thresholds which minimize both the number of false alarms and the probability of missing actual attacks as a two-player Stackelberg security game. We provide an efficient dynamic programming-based algorithm for solving the game, thereby finding optimal detection thresholds. We analyze the performance of the proposed algorithm and show that its running time scales polynomially as the length of the time horizon of interest increases. In addition, we study the problem of finding optimal thresholds in the presence of both random faults and attacks. Finally, we evaluate our result using a case study of contamination attacks in water networks, and show that our optimal thresholds significantly outperform fixed thresholds that do not consider that the environment is dynamical.

On the sensitivity and uncertainty of wave energy conversion with an artificial neural-network-based controller

This work addresses with sensitivity and uncertainty of the energy conversion of an oscillation-body wave energy converter with an artificial neural-network-based controller. The smart controller applies the model predictive control strategy to implement real-time latching control to the wave energy converter. Since the control inputs are future wave forces, an artificial neural network is developed and trained by the machine learning algorithm to predict the short-term wave forces based on the real-time measurement of wave elevation. The sensitivity of wave energy conversion with respect to wave frequency and receding horizon length are investigated. Uncertainties of the neural network that lead to the prediction deviation are identified and quantified, and their influences on the energy conversion are examined. The control command is derived inappropriately in the presence of prediction deviation leading to the reduction of energy absorption. Moreover, it is the phase deviation that reduces the energy absorption.

SVR-AMA: An Asynchronous Alternating Minimization Algorithm With Variance Reduction for Model Predictive Control Applications

This paper focuses on the design of an asynchronous dual solver suitable for model predictive control (MPC) applications. The proposed solver relies on a state-of-the-art variance reduction (VR) scheme, previously used in the context of proximal stochastic gradient methods (Prox-SVRG) and on the alternating minimization algorithm (AMA). The resultant algorithm, a stochastic AMA with VR (SVR-AMA), shows geometric convergence (in the expectation) to a suboptimal solution of the MPC problem and, compared to other state-of-the-art dual asynchronous algorithms, allows one to tune the probability of the asynchronous updates to improve the quality of the estimates. Two novel accelerated versions of the Prox-SVRG (and, by duality, of SVR-AMA) are also provided. We apply the proposed algorithm to a specific class of splitting methods, that is, the decomposition along the length of the prediction horizon. Numerical results on the longitudinal control problem of an Airbus passenger aircraft show the benefits that we can gain in terms of computation time when using our proposed solver with an adaptive probability distribution.

PCN100 ECONOMIC EVALUATION OF FOLFIRINOX VERSUS GEMCITABINE IN THE TREATMENT OF METASTATIC PANCREATIC CANCER IN JAPAN

The purpose of this analysis was to evaluate the cost-effectiveness of FOLFIRINOX in patients with metastatic pancreatic cancer, comparing with gemcitabine. Partitioned survival model was developed to predict costs and quality-adjusted life years (QALYs) in FOLFIRINOX and gemcitabine arms. Direct medical costs were considered from the perspective of the Japanese healthcare system. Time horizon and cycle length of the model was set to 5 years and 1 month, respectively. An annual discount rate of 2% for both costs and QALYs was applied. Data on overall survival and progression-free survival was derived from the published randomized controlled trial. Cost parameters were estimated by using the JMDC claims database. All costs were estimated in Japanese yen and converted to US dollars with the currency exchange rate of $ 1 = JPY100. Utilities were derived from published sources other than Japan. The incremental cost-effectiveness ratio (ICER) of FOLFIRINOX compared with gemcitabine was estimated. Deterministic sensitivity analysis was performed to assess uncertainty in parameter settings. Compared with gemcitabine, FOLFIRINOX incurred an additional cost of $ 28,518 and conferred an additional 0.247 QALY. This resulted in the ICER of $ 115,421 per QALY gained. Costs for post-progression had a relatively large impact on the base case result, but the ICERs remained higher than $ 50,000 per QALY over the full range of model parameters. Applying the willingness to pay threshold of $ 50,000 per QALY, FOLFIRINOX might not be cost-effective. Further research is required on utilities of Japanese patients with pancreatic cancer.

Sequential Choice Bandits: Learning with Marketing Fatigue

Motivated by the observation that overexposure to unwanted marketing activities can lead to customer dissatisfaction, we consider a setting where a platform offers a sequence of messages to its users and is penalized when users abandon the platform due to marketing fatigue. We propose a novel sequential choice model to capture multiple interactions taking place between the platform and its users: upon receiving a message, a user decides on whether to accept or reject the message. If she chooses to reject, she would then decide to either receive the next message in the sequence or abandon the platform. Based on user feedback, the platform dynamically learns users' abandonment distribution and the relevance of the recommended content. With a goal to maximize the cumulative payoff over a horizon of length T, the platform dynamically adjusts the sequence of messages and the order in which the messages are shown to a user. We refer to this online learning task as the sequential choice bandit (SC-Bandit) problem. For the offline combinatorial optimization problem, we show a polynomial-time algorithm. For the online problem, we consider two variants, depending on whether contexts are included, and propose algorithms that balance exploration and exploitation. Lastly, we evaluate the performance of our algorithms with both synthetic and real-world datasets.

Cost-effectiveness of cannabinoids for pediatric drug-resistant epilepsy: protocol for a systematic review of economic evaluations

BackgroundDrug-resistant epilepsy negatively impacts the quality of life and is associated with increased morbidity and mortality and high costs to the healthcare system. Cannabis-based treatments may be effective in reducing seizures in this population, but whether they are cost-effective is unclear. In this systematic review, we will search for cost-effectiveness analyses involving the treatment of pediatric drug-resistant epilepsy with cannabis-based products to inform decision-making by public healthcare payers about reimbursement of such products. We will also search for cost-effectiveness analyses of other pharmacologic treatments for pediatric drug-resistant epilepsy, as well as estimates of healthcare resource use, costs, and utilities, for use in a subsequent cost-utility analysis to address this decision problem.MethodsWe will search the published and gray literature for economic evaluations of cannabis-based products and other pharmacologic treatments for pediatric drug-resistant epilepsy, as well as resource utilization and utility studies. Two independent reviewers will screen the title and abstract of each identified record and the full-text version of any study deemed potentially relevant. Study and population characteristics, the incremental cost-effectiveness ratio (ICER), as well as total costs and benefits, will be extracted, and quality will be assessed by use of the Drummond and CHEERS checklists; context-specific issues will also be considered. From model-based cost-utility and cost-effectiveness analyses, we will extract and summarize the model structure, including health states, time horizon, and cycle length. From resource utilization studies, we will extract data about the frequency of resource use (e.g., neurology visits, emergency department visits, admissions to hospital). From utility studies, we will extract the utility for each health state, the source of the preferences (e.g., child, parent, patient, general public), and the method of elicitation.DiscussionDrug-resistant epilepsy in children is associated with important costs to the healthcare system, and decision-makers require high-quality evidence on which to base reimbursement decisions. The results of this review will be useful to both decision-makers considering the decision problem of whether to reimburse cannabis-based products through public formularies and to analysts conducting studies in this area.Systematic review registrationPROSPERO no.: CRD42018099591.

Velocity forecasts using a combined deep learning model in hybrid electric vehicles with V2V and V2I communication

Vehicle velocity forecast is an important clue in improving the performance of energy management in hybrid electric vehicles (HEV). This paper presents a new combined model for predicting vehicle’s velocity time series. The main features of the model are to combine the feature extraction capability of deep restricted Boltzmann machines (DBM) and sequence pattern predicting capability of bidirectional long short-term memory (BLSTM). Hence, the model is named as DBMBLSTM. In addition, the DRMBLSTM model utilizes the vehicle driving information and roadside infrastructure information provided respectively through vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication channels to predict vehicle velocity at various length of prediction horizon. Furthermore, the predictions results of this study are compared with the state of the art of vehicle velocity forecasts. The root mean square error (RMSE) is used as an evaluation criteria of predictions accuracy. Finally, these compared prediction model are applied in model predictive control (MPC) energy management strategy for the verifications of fuel economy improvement of a HEV. Simulation results confirm that the proposed combined deep learning model performs better than other five prediction methods. Therefore, it is a means of arriving at a reliable forecast model for HEV.

Autonomous soaring using a simplified MPC approach

ABSTRACTThe need for efficient propulsion systems allied to increasingly more challenging fixed-wing UAV mission requirements has led to recent research on the autonomous thermal soaring field with promising results. As part of that effort, the feasibility and advantages of model predictive control (MPC)-based guidance and control algorithms capable of extracting energy from natural occurring updrafts have already been demonstrated numerically. However, given the nature of the dominant atmospheric phenomena and the amplitude of the required manoeuvres, a non-linear optimal control problem results. Depending on the adopted prediction horizon length, it may be of large order, leading to implementation and real-time operation difficulties. Knowing that, an alternative MPC-based autonomous thermal soaring controller is presented herein. It is designed to yield a simple and small non-linear programming problem to be solved online. In order to accomplish that, linear prediction schemes are employed to impose the differential constraints, thus no extra variables are added to the problem and only linear bound restrictions result. For capturing the governing non-linear effects during the climb phase, a simplified representation of the aircraft kinematics with quasi-steady corrections is used by the controller internal model. Flight simulation results using a 3 degree-of-freedom model subjected to a randomly generated time varying thermal environment show that the aircraft is able to locate and exploit updrafts, suggesting that the proposed algorithm is a feasible MPC strategy to be employed in a practical application.

Forecast horizon of dynamic lot size model for perishable inventory with minimum order quantities

We consider the dynamic lot size problem for perishable inventory under minimum order quantities. The stock deterioration rates and inventory costs depend on both the age of the stocks and their periods of order. Based on two structural properties of the optimal solution, we develop a dynamic programming algorithm to solve the problem without backlogging. We also extend the model by considering backlogging. By establishing the regeneration set, we give a sufficient condition for obtaining forecast horizon under without and with backlogging. Finally, based on a detailed test bed of instance, we obtain useful managerial insights on the impact of minimum order quantities and perishability of product and the costs on the length of forecast horizon.

Bitcoin at High Frequency

This paper studies the behaviour of Bitcoin returns at different sample frequencies. We consider high frequency returns starting from tick-by-tick price changes traded at the Bitstamp and Coinbase exchanges. We find evidence of a smooth intra-daily seasonality pattern, and an abnormal trade- and volatility intensity at Thursdays and Fridays. We find no predictability for Bitcoin returns at or above one day, though, we find predictability for sample frequencies up to 6 h. Predictability of Bitcoin returns is also found to be time–varying. We also study the behaviour of the realized volatility of Bitcoin. We document a remarkable high percentage of jumps above 80 % . We also find that realized volatility exhibits: (i) long memory; (ii) leverage effect; and (iii) no impact from lagged jumps. A forecast study shows that: (i) Bitcoin volatility has become more easy to predict after 2017; (ii) including a leverage component helps in volatility prediction; and (iii) prediction accuracy depends on the length of the forecast horizon.

Novel Formulation for Optimal Schedule with Demand Side Management in Multiproduct Air Separation Processes

In this article, we address the optimal scheduling of continuous air separation processes with electricity purchased from the day-ahead market, for which we propose a generalized framework to represent different operating states. Specifically, a discrete-time mixed-integer linear programming (MILP) model is developed based on this representation for operating states, which has proven to provide a tight LP relaxation for handling industrial-scale instances. The computational efficiency of the model is demonstrated with data from real industrial production. The response of the scheduling and production level is also tested with various interval lengths for the electricity pricing and length of the time horizon.

Bitcoin at High Frequency

This paper studies the behaviour of Bitcoin returns at different sample frequencies. We consider high frequency returns starting from tick-by-tick price changes traded at the Bitstamp and Coinbase exchanges. We find evidence of a smooth intra-daily seasonality pattern, and an abnormal trade- and volatility intensity at Thursdays and Fridays. We find no predictability for Bitcoin returns at or above one day, though, we find predictability for sample frequencies up to 6 hours. Predictability of Bitcoin returns is also found to be time-varying. We also study the behaviour of the realized volatility of Bitcoin. We document a remarkable high percentage of jumps above 80%. We also find that realized volatility exhibits: i) long memory, ii) leverage effect, and iii) no impact from lagged jumps. A forecast study shows that: i) Bitcoin volatility has become more easy to predict after 2017, ii) including a leverage component helps in volatility prediction, and iii) prediction accuracy depends on the length of the forecast horizon.