Massimo Vaccarini

Cooperative Control of Underwater Glider Fleets by Fault Tolerant Decentralized MPC

Abstract A fleet of Autonomous Underwater Vehicles (AUVs) moving together in a prescribed pattern can form an efficient data acquisition network. The problem is to control several AUVs such that after transients, they form a required formation and move along a desired trajectory. The capability to accomplish the mission even in case of faults, is a fundamental requirement for these kind of missions. A completely decentralized predictive control and FDI strategy is here proposed for allowing cooperation. Trough an underwater communication channel, each vehicle broadcasts its position, its future behavior and its actuator/sensor fault situation. Based both on the local and the received information, each vehicle selects the desired formation to keep and plans its future actions. Simulation results are provided to validate the approach.

Formation control of marine veihicles via real-time Networked Decentralized MPC

Traditional ocean research and exploration usually require a lot of time, funds and, in general, resources. In the last years, mainly thanks to the GPS technology, unmanned small vehicles have been proposed for marine exploration and investigation. In this paper, a cooperation scheme (either for surface and underwater vehicles) is outlined for a multi-robot framework with a certain mission. The objective is translated into a formation control problem and the coordination is achieved through underwater sonar communication. A real-time implementable formation control strategy based on Networked Decentralized MPC is presented that drive each single vehicle. MPC allows to take into account for constraints, nonlinearities and exchanged information about future behaviours of the single agents. The decentralized strategy makes use of either local and received information for computing the best behaviour that optimises some objective function. A reconfigurable hierarchical structure is assumed for managing the formation. The real-time implementability is achieved using Multiplexed MPC and/or stochastic optimization algorithms for reducing computation times. Simulation results are provided for comparing the presented algorithms.

NETWORKED DECENTRALIZED MPC FOR UNICYCLE VEHICLES FORMATION

Abstract The formation control of a system of multiple autonomous vehicles is here considered. A decentralized Model Predictive Control is developed, analyzed and compared with the centralized solution. A coordination strategy based on an information exchange through a local network is here proposed for improving the global control performance. Preliminary stability analysis results for the considered architecture are also provided. The developed simulation tests verify that the proposed decentralized control is able to guarantee the same performance of the centralized solution with a significant reduction of the computational efforts.

NETWORKED DECENTRALIZED MPC FOR FORMATION CONTROL OF UNDERWATER GLIDER FLEETS

Abstract Exploration by means of unmanned multi-robot formations is gaining interests in different fields. In particular, the marine exploration by means of autonomous glider fleets allows to save money and to safely perform dangerous missions. The considered vehicles are highly nonlinear and the feasible control actions are constrained. A Networked Decentralized Model Predictive Control strategy is here considered for the formation control in the horizontal plane. Each control agent communicates with its leader in order to keep a desired formation. Some considerations on the stability of the global decentralized system are provided.

State space stability analysis of unconstrained decentralized model predictive control systems

This paper analyzes stability properties of a decentralized MPC strategy. Assuming that control agents can exchange information with each other only once for every sampling interval, stability conditions on model and controller matrices are derived for the unconstrained case. Stability properties are analyzed either from a local or a global point of view and a comparison between the decentralized and the centralized control is introduced

Bayesian Networks for Supporting Model Based Predictive Control of Smart Buildings

Optimal behaviour is one the most desired features of contemporary technological systems. Challenges like secure operation, energy efficiency, and reliable performance call for the optimised behaviour of any systems that operate and interact in our living environment. The challenge in achieving optimised performances resides in the uncertainty that qualifies the environment surrounding technical systems. Whatever model drives the systems’ behaviour, it must be able to face unforeseen events, to manage the vagueness of the sensing apparatus and the errors of the control devices. Bayesian statistics is one of the theoretical backgrounds that support the construction of systems which are able to act effectively inside complex environments. Bayesian statistics is grounded on the fundamental premise that all uncertain‐ ties should be represented and measured by probabilities. Then, the laws of probabilities apply to produce probabilistic inferences about any quantity, or collection of quantities, of interest. Bayesian inference can provide predictions about probability values pertaining time series or can model parameters in terms of probability distributions that represent and summarize current uncertain knowledge and beliefs. Bayesian inference uses a kind of direct causal or model-based knowledge to provide the crucial robustness needed to make the optimised behaviour of technical systems feasible in the real world [1]. Once this kind of models have been built, then theoretically sound evidence propagation algorithms are used to update the belief set about the external environment and about the system performance, on the basis of acquired evidence. This is the fundamental mechanism that drives the construction and the operation of intelligent systems based on Bayesian inference. This chapter describes a sample engineering application of this approach on a large scale. It concerns the design and the development of an intelligent building energy management system (smart BEMS) that is able

APPLICATION OF A NETWORKED DECENTRALIZED MPC TO SYNGAS PROCESS IN OIL INDUSTRY

Abstract The performance of a networked decentralized model predictive control is analyzed and tested on an industrial process characterized by strong interactions. The proposed control architecture is based on a set of independent agents which implement independent model predictive control strategies and exchange a reduced set of information by a local network. The experimental results are satisfactory and comparable with those obtained by the classical centralized model predictive control strategy.

A Wireless System for Real-time Environmental and Energy Monitoring of a Metro Station: Lessons Learnt from a Three-year Research Project

Optimal energy management of underground transportation systems is widely recognized as a key aspect for significant energy savings at regional level. The three-year long EU funded “Seam4us” research project (2011 to 2014) aims to create a system for optimized integrated energy management, relying on the Passeig de Gracia metro station in Barcelona as the pilot station. One of the outcomes of the project was the installation of a wireless sensor network, in order to track in real time both environmental and energy parameters. The data collected by the network have been exploited to inform an intelligent control system about the state of the station. In fact, a reliable real-time sensing is critical for implementing advanced control policies in any kind of environment. For that reason, this paper will report on the Seam4us findings regarding real-time sensing of that quite harsh domain. The design criteria and constraints that led to the final installation will be argued. Then, post-processing functions, i.e. those algorithms turning raw data into the variables which can be processed by the controller, will be described. Finally, the general performance of the system will be discussed, in terms of reliability of data exchange and energy efficiency.

Interoperable Approach in Support of Semi-Automated Construction Management

Construction progress monitoring has always been a major concern for managers. Several are the advantages provided by a reliable approach in this field, among which we cite efficient project performance control and quality control, timely on-site inspections, better control of health and safety prescriptions against job injuries and fatalities, which are still too high. This paper addresses the development of a semi-automated approach for construction management, based on two main features: on one side a project management oriented re-organization of project’s information based on interoperable BIM (Building Information Modeling) and 4D modeling protocols, which aims to facilitate the control at the execution phase; on the other side a monitoring platform for real-time collection of data, relative to work progress and resources usage, by means of low intrusive technologies. Preliminary experiments on the capability by the low intrusive and real-time monitoring system to acquire and filter the data collected on resources usage are carried out. In addition, a general framework implementing several high-level logics is proposed. Its application would lead to seamless production of reports, to a semi-automated control of materials procurement, that is to say with limited or no human intervention. Preliminary application cases are shown, too. Among the expected benefits, not only economic savings for builders, but also better quality of recorded data and real-time availability of information have been pointed out.

A Real-Time and Non Intrusive Structural Health Monitoring System for Buildings

The constant innovation in the development of sensing devices for detecting health status of buildings, highlights the importance of this professional field. In fact, correct diagnosis would prevent many pathologies, driving maintenance interventions towards timely and well targeted operations. This paper describes the development of a monitoring system prototype, conceived to be very low intrusive, that can be applied also on existing buildings, without any need to predispose chases for power or communication cables. Its adoption is expected to be useful not only for standard health status monitoring of buildings during their lifecycle, but also for automated monitoring of old buildings during the conduction of works for their renovation. As a matter of fact in this case accidents due to unexpected collapses during work execution are very common, hence the static condition of buildings should be continuously checked automatically and alerts sent in case of high risk of failures. The sensors under development have been programmed in such a way to be able to monitor rotations and send alerts in case critical thresholds are overcome.