Well-established nonlinear control methods often rely on mathematical models that can be imprecise or uncertain. Classical robust control techniques address this by designing conservative control laws, which inherently limit performance. As modern applications demand higher performance, learningbased controllers have emerged as a promising alternative; however, they frequently lack stability guarantees, rendering them unsuitable for safety-critical systems. To bridge this gap, this paper introduces a novel control strategy that integrates a robust controller with an Echo State Network (ESN)-based adaptive law to stabilize a class of uncertain nonlinear systems under persistent disturbances. The robust controller is first designed to ensure Input-to-State Stability (ISS) of the closed-loop system, irrespective of the ESN’s actions or disturbances. Subsequently, the ESN is trained online to compensate for the disturbances and improve the system’s output performance. A key feature of this architecture is that the closed-loop stability is guaranteed a priori by the robust controller, regardless of the ESN’s learning process, while the online adaptation significantly enhances performance. This framework also maintains relatively low computational complexity. Numerical simulations demonstrate the strategy’s effectiveness: in one example, the average performance improves from 60.6% to 92.7% over time, and in another, from 22.8% to 92.3%. These results confirm the controller’s ability to learn and adapt online, showing a significant attenuation of disturbance effects compared to using the robust controller alone.

Sparse convex optimization involves optimization problems where the decision variables are constrained to have a certain number of entries equal to zero. In this paper, we consider the case in which the objective function is decomposed into a sum of different local objective functions and propose a novel fully-distributed scheme to address the problem over a network of cooperating agents. Specifically, by taking advantage of a suitable problem reformulation, we define an Augmented Lagrangian function associated with the reformulated problem. Then, we address such an Augmented Lagrangian by suitably interlacing the Gradient Tracking distributed algorithm and the Block Coordinated Descent method giving rise to a novel fully-distributed scheme. The effectiveness of the proposed algorithm is corroborated through some numerical simulations of problems considering both synthetic and real-world data sets.

In a BRT system, for an efficient planning and management of fleets and drivers, precision of operations is required at the intervals provided. However, sometimes actions are taken that change the order of buses on the circuit, such as overtaking. Despite their impact on operations, these changes are not fully addressed by main headway control algorithms found in the literature and, when considered, they are dealt with enumerating processes to decide on the reordering of buses. To mitigate this issue, this paper presents a generalized formulation for auto-sequencing buses in headway control strategies. The proposed methodology allows the modeling of overtaking, injection, and removal of buses, without resorting to enumerating methods for bus ordering. The obtained results demonstrate that the proposed formulation can handle the events of change of order in a BRT circuit while, at the same time, controlling the headway between the buses and reducing the waiting time for passengers. Thus, the generalized auto-sequencing method is a framework that can be used to augment other control strategies already present in the literature.

Nonlinear model predictive control (NMPC) of industrial processes is changeling in part because the model of the plant may not be completely known but also for being computationally demanding. This work proposes an extremely efficient reservoir computing (RC)-based control framework that speeds up the NMPC of processes. In this framework, while an echo state network (ESN) serves as the dynamic RC-based system model of a process, the practical nonlinear model predictive controller (PNMPC) simplifies NMPC by splitting the forced and the free responses of the trained ESN, yielding the so-called ESN-PNMPC architecture. While the free response is generated by the forward simulation of the ESN model, the forced response is obtained by a fast and recursive calculation of the input-output sensitivities from the ESN. The efficiency not only results from the fast training inherited by RC but also from a computationally cheap control action given by the aforementioned novel recursive formulation and the computation in the reduced dimension space of input and output signals. The resulting architecture, equipped with a correction filter, is robust to unforeseen disturbances. The potential of the ESN-PNMPC is shown by application to the control of the four-tank system and an oil production platform, outperforming the predictive approach with a long-short term memory (LSTM) model, two standard linear control algorithms, and approximate predictive control.

The problem of optimizing the operation of Bus Rapid Transit (BRT) lines that share a common corridor is analyzed. The goal is to minimize passengers travel time, including waiting at stations. In the common corridor, stations are shared by several bus lines which present riding alternatives for passengers with origins and destinations within the corridor. Since buses from different lines interact in the corridor, optimal multi-line headways that consider headways across all lines can alter single-line headways so as to maximize user benefit. We develop the system model in a mathematical programming fashion, embedded with the control strategy as bus holding decision variables. The method, named Multi-line Integrated Holding Control Strategy (M-IHCS), implements the required bus headway regulation for multiple lines in an integrated and simultaneous manner. Numerical tests of the optimization approach developed for the system model show the distinctive features of the solutions, indicating the superiority of multi-line headway control over its single-line counterpart.

This paper presents a real-time integrated holding and priority control strategy for bus rapid transit (BRT) and high-frequency segregated transit systems, in a network approach (considering the whole transit circuit). The integrated control strategy implements simultaneously the required bus headway corrections and the bus priority through signalized intersections with the objective of minimizing the total delay of passengers that are onboard and at stops. The developed model, in the form of a mathematical programming problem, presents peculiarities for describing the behavior of BRT and transit systems such as bus capacity, differentiation between passenger boarding and alighting processes in single stops, terminals and stations, and bus priority at signalized intersections. The iterative procedure proposed for solving the optimization problem brings about simplicity and efficiency for real-time applications. The results obtained by simulation for a transit scenario of the trunk line 10 of the city of Blumenau, Santa Catarina, Brazil, show the efficiency of the proposed strategy and potential of practical application.

A methodology is proposed for production optimization of oilfields consisting of multiple offshore reservoirs. In such complex systems, several production units are interconnected by a subsea pipeline network that transfers fluids to onshore terminals. A graph-based model of production units, pipelines, and nonlinear phenomena leads to a mixed-integer nonlinear problem for production optimization. Owing to its sheer size, the proposed approach relies on piecewise-linear proxy modeling of the production units and fluid flow. The end result is a mixed-integer linear problem to which robust algorithms can be applied. By means of simulation, the methodology is tested for an offshore oilfield in the Santos Basin, encompassing multiple reservoirs and production platforms.

A linear dynamic network is a system of subsystems that approximates the dynamic model of large, geographically distributed systems such as the power grid and traffic networks. A favorite technique to operate such networks is distributed model predictive control (DMPC), which advocates the distribution of decision-making while handling constraints in a systematic way. This paper contributes to the state-of-the-art of DMPC of linear dynamic networks in two ways. First, it extends a baseline model by introducing constraints on the output of the subsystems and by letting subsystem dynamics to depend on the state besides the control signals of the subsystems in the neighborhood. With these extensions, constraints on queue lengths and delayed dynamic effects can be modeled in traffic networks. Second, this paper develops a distributed interior-point algorithm for solving DMPC optimization problems with a network of agents, one for each subsystem, which is shown to converge to an optimal solution. In a traffic network, this distributed algorithm permits the subsystem of an intersection to be reconfigured by only coordinating with the subsystems in its vicinity.

A multiple control-point strategy for holding control of a bus transit system is presented. The model developed is deterministic and assumes the availability of real-time information and historical data of the system. Stochastic effects are disturbances to be compensated by the feedback nature of the control. The objective is to minimize total user delay, which is modeled by a nonconvex cost function and nonlinear constraints. To efficiently solve the problem, simplifications of the original model are introduced, together with an iterative quadratic programming (IQP) optimization procedure. A numerical example illustrates the application of the method, indicating its feasibility for real-time applications and the good approximation of the global optimum provided by the heuristic solution.

The complexity of real-time systems has substantially increased in the past few years regarding both hardware and software aspects. The use of modern sensors, able to capture image and audio data, demands predictable multimedia-like data processing. Moreover, applications like autonomous robots, surveillance, or modern multimedia players may well be characterized by several operation modes, each one associated with light conditions, vision angle, change in user requirements, etc. In this paper, we describe suitable scheduling mechanisms that address these aspects. Application modes are characterized by their required processing bandwidth and benefit values. By using bandwidth reservation schedulers, dynamic reconfiguring scheduling parameters is seen as an optimization problem whose goal is to maximize the overall system benefit subject to schedulability constraints. Two different models for the problem are defined, Discrete and Continuous. The former gives rise to an NP-Hard problem for which efficient approximate solutions are derived. An optimal and polynomial solution to the Continuous model is derived. Both models are then extended to incorporate task execution times described as probability distributions. Making use of this stochastic modeling one is able to dynamically reconfigure the scheduler subject to probabilistic schedulability guarantees. The derived solutions are evaluated by extensive simulation, which indicates the good performance of the proposed reconfiguration mechanisms.

Reservation-based scheduling mechanisms have successfully been used for supporting real-time applications whose tasks exhibit high variability in their execution or release times. Indeed, such mechanisms are able to preallocate system bandwidth to the application tasks so that temporal isolation between them is ensured. However, bandwidth allocation is usually based on off-line policies, which may not be suitable for real-time applications that are structured as having several modes of operation, each one requiring a distinct level of system bandwidth. Variations in light conditions, the changing of energy levels, error-detection, or operator commands are examples of events that may trigger a different mode of operation in multi-mode adaptive real-time applications. In this paper we address the problem of dynamically reconfiguring scheduling parameters of reservation-based mechanisms, offering support for multi-mode adaptive real-time applications. Assuming that each reconfiguration option gives a benefit for the system, reconfiguration is seen as an optimization problem whose objective is to maximize the overall system benefit. Two different models for the problem are formulated, the Integer Programming (IP) and the Linear Programming (LP) formulations. The IP formulation gives rise to an NP-Hard problem for which we give efficient approximate solutions. Also, an optimal and polynomial solution to the LP formulation is derived. Results obtained from extensive simulation indicate the good performance of the proposed reconfiguration mechanisms.

Distributed model predictive control (DMPC) advocates the distribution of sensing and decision making to operate large, geographically distributed systems such as the power grid and traffic networks. This paper presents a distributed optimization framework for DMPC of linear dynamic networks with constraints on each network node. A linear dynamic network can be thought of as a directed graph, whose nodes have local dynamics that depend on the local and upstream control signals and are subject to constraints on state and control variables. The distributed algorithm is based on interior-point methods and can be shown to converge to a globally optimal solution. Some theoretical results are stated and a preliminary application to green-time control in urban traffic networks is described.

The distribution of a limited rate of high-pressure gas to gas-lifted wells, while respecting injection bounds and activation precedence constraints, consists of a mixed-integer nonlinear programming problem. This paper proposes a mixed-integer linear formulation obtained by piecewise-linearizing the nonlinear functions, thereby allowing the use of integer programming algorithms. Valid inequalities for the convex hull of feasible solutions are derived from knapsack covers, for which exact and approximate lifting procedures yield stronger inequalities. Numerical results show that these cover-based cuts reduce the number of nodes explored in a branch-and-bound search.

A linear-dynamic network consists of a directed graph in which the nodes represent subsystems and the arcs model dynamic couplings. The local state of each subsystem evolves according to discrete linear dynamics that depend on the local state, local control signals, and control signals of upstream subsystems. Such networks appear in the model predictive control (MPC) of geographically distributed systems such as urban traffic networks and electric power grids. In this correspondence, we propose a decomposition of the quadratic MPC problem into a set of local subproblems that are solved iteratively by a network of agents. A distributed algorithm based on the method of feasible directions is developed for the agents to iterate toward a solution of the subproblems. The local iterations require relatively low effort to arrive at a solution but at the expense of high communication among neighboring agents and with a slower convergence rate.

Modern real-time systems must be designed to be highly adaptable, reacting to aperiodic events in a predictable manner and exhibiting graceful degradation in overload scenarios whenever needed. In this context, it is useful to structure the system as a set of multiversion tasks. Task versions can be modeled to implement services with various levels of quality. In overload scenarios, for instance, a lower quality service may be scheduled for execution keeping the system correctness and providing graceful degradation. The goal of the reconfiguration mechanism is to select the versions of tasks that lead to the maximum benefit for the system at runtime. In this paper, we provide a schedulability condition based on which we derive an optimal pseudo-polynomial solution for this problem. Then, a faster approximation solution is described. Results from simulation indicate the effectiveness of the proposed approach.

Lift-gas allocation concerns the distribution of a limited rate of high pressure gas to oil wells with the aim of maximizing profit from hydrocarbon selling. The problem is becoming more complex as oil fields mature and constraints are imposed to deliver near-optimal operations. Today, systems engineers take account of bounds on gas injection, discontinuous production functions, and activation precedence constraints. To this end, this paper discusses a mixed-integer linear formulation obtained from a piecewise linear reformulation of well performance curves for lift-gas allocation under activation precedence constraints. The model handles injection bounds, a limited lift-gas rate, and can be extended to incorporate facility constraints. A somewhat rigorous analysis of the integer polyhedron is carried out to produce valid inequalities based on 1-configurations. Not only can these inequalities be used in cutting-plane and branch-and-cut algorithms, but also applied in combination with evolutionary algorithms to produce upper bounds.

Model predictive control (MPC) has become one of the leading technologies to control complex processes, to a great extent, as a result of its flexibility and explicit handling of constraints. Given a dynamic problem (DP), MPC converts DP into a series of static optimization problems, thereby allowing the use of standard optimization techniques to compute the control signals. The reliance of MPC on centralized computations, however, stands as a barrier to its use in the real-time operation of large dynamic networks. To this end, this paper proposes an extension to MPC by decomposing DP into a network of small but coupled subproblems and solving them with a network of asynchronous agents. The net result, after each agent applies MPC to its dynamic subproblem, is a series of sets of static subproblems. Our focus is on the simultaneous solution of these sets of static subproblems. The paper delivers a framework to carry out the decomposition and develops conditions under which the iterative synchronous processes of the agents converge to solutions. Furthermore, it proposes heuristics for asynchronous convergence and reports experimental results from prototypical dynamic networks, demonstrating the effectiveness of the proposed extension.

The continuous gas-lift method is one of the most used artificial lifting techniques in which the allocation of gas injection rates is a very important optimization problem. In this work, we develop a dynamic programming (DP) algorithm that solves the profit maximization problem for a cluster of oil wells producing via gas lift, with multiple well performance curves (WPCs) and constrained by the amount of lift gas available for injection. The algorithm is a low-cost and high-efficiency decision support tool that outperforms alternative methods found in the literature.

The pressure from today’s economic and energy markets demands further distribution of decision making in large, dynamic networks. To this end, distributed model predictive control (MPC) divides the task of operating a dynamic network into a set of small, localized subtasks, one for each distributed control agent. To guide the decomposition of the overall task, which plays a central role in the quality of the operation yielded by the distributed agents, this paper proposes a model for problem decomposition based on the design of agent communication networks. The model gives rise to the communication-network design problem: A bicriteria optimization problem whose objectives are the maximization of the influence perceived by the agents and the minimization of the communication cost. The benefit of the model is the potential to concentrate the effort of implementing distributed MPC, and measuring its actual performance, on a reduced number of decompositions, namely those that are Pareto efficient for the model. The paper gives an account of related work, including the computational complexity of finding Pareto efficient solutions, an integer programming formulation, and families of valid inequalities. Its main contribution is the demonstration that the model can be effective, meaning that Pareto efficient solutions to the model tend to induce efficient problem decompositions. The experimental evidence was gathered by applying the model to decompose control problems of two representative networks, namely arrays of pendulums and electric power networks, and measuring the quality of the operation delivered by the distributed control agents.

This paper demonstrates the effectiveness of applying optimization techniques to solve traffic engineering (TE) problems in IP networks over multiprotocol label switching (MPLS). Our approach models TE tasks as mathematical programming problems and proposes heuristic algorithms. Another contribution of this work is the combination of Lagrangean relaxation with heuristics to compute near-optimal solutions quickly. Numerical experiments contrast the solutions produced by our algorithm with optimal ones, which were obtained with a top-notch optimization software package. All in all, the results indicate that the Lagrangean-based routing method outperforms standard algorithms with respect to a number of performance criteria, including throughput and packet-loss rate.