This thesis presents a new approach to local decision-making in multi-agent systems with varying amounts of communication. Here, local decision-making refers to action choices which are made in a decentralized fashion by individual agents based on the information which is locally available to them. The work described here is set within the multi-agent decision process framework. Unreliable, faulty or stochastic communication patterns present a challenge to these settings which usually rely on precomputed, centralised solutions to control individual action choices. Various approximate algorithms for local decision-making are developed for scenarios with and without sequentiality. The construction of these techniques is based strongly on methods of Bayesian inference. Their performance is tested on synthetic benchmark scenarios and compared to that of a more conservative approach which guarantees coordinated action choices as well as a completely decentralized solution. In addition, the method is applied to a surveillance task based on real-world data. These simulation results show that the algorithms presented here can outperform more traditional approaches in many settings and provide a means for flexible, scalable decision-making in systems with varying information exchange between agents.

This dissertation addresses a fundamental question of multi-agent coordination: what infor- mation should be sent to whom and when, with the limited resources available to each agent? Communication requirements for multi-agent systems can be rather high when an accurate picture of the environment and the state of other agents must be maintained. To reduce the impact of multi-agent coordination on networked systems, e.g., power and bandwidth, this dissertation introduces new concepts to enable Intelligent Knowledge Distribution (IKD), including Constrained-action POMDPs (CA-POMDP) and concurrent decentralized (CoDec) POMDPs for an agnostic plug-and-play capability for fully autonomous systems. Each agent runs a CoDec POMDP where all the decision making (motion planning, task allocation, asset monitoring, and communication) are separated into concurrent individual MDPs to reduce the combinatorial explosion of the action and state space while maintaining dependencies between the models. We also introduce the CA-POMDP with action-based constraints on partially observable Markov decision processes, rewards driven by the value of information, and probabilistic constraint satisfaction through discrete optimization and Markov chain Monte Carlo analysis. IKD is adapted real-time through machine learning of the actual environmental impacts on the behavior of the system, including collaboration strategies between autonomous agents, the true value of information between heterogeneous systems, observation probabilities and resource utilization. / Doctor of Philosophy / This dissertation addresses a fundamental question behind when multiple autonomous sys- tems, like drone swarms, in the field need to coordinate and share data: what information should be sent to whom and when, with the limited resources available to each agent? Intelligent Knowledge Distribution is a framework that answers these questions. Communication requirements for multi-agent systems can be rather high when an accurate picture of the environment and the state of other agents must be maintained. To reduce the impact of multi-agent coordination on networked systems, e.g., power and bandwidth, this dissertation introduces new concepts to enable Intelligent Knowledge Distribution (IKD), including Constrained-action POMDPs and concurrent decentralized (CoDec) POMDPs for an agnostic plug-and-play capability for fully autonomous systems. The IKD model was able to demonstrate its validity as a "plug-and-play" library that manages communications between agents that ensures the right information is being transmitted at the right time to the right agent to ensure mission success.

Dissertation abstract Title: Cross-entropy based combination of discrete probability distributions for distributed de- cision making Author: Vladimíra Sečkárová Author's email: seckarov@karlin.mff.cuni.cz Department: Department of Probability and Mathematical Statistics Faculty of Mathematics and Physics, Charles University in Prague Supervisor: Ing. Miroslav Kárný, DrSc., The Institute of Information Theory and Automation of the Czech Academy of Sciences Supervisor's email: school@utia.cas.cz Abstract: In this work we propose a systematic way to combine discrete probability distributions based on decision making theory and theory of information, namely the cross-entropy (also known as the Kullback-Leibler (KL) divergence). The optimal combination is a probability mass function minimizing the conditional expected KL-divergence. The ex- pectation is taken with respect to a probability density function also minimizing the KL divergence under problem-reflecting constraints. Although the combination is derived for the case when sources provided probabilistic type of information on the common support, it can applied to other types of given information by proposed transformation and/or extension. The discussion regarding proposed combining and sequential processing of available data, duplicate data, influence...

Tesis por compendio / [ES] Algunos productos se caracterizan por su falta de homogeneidad, lo que significa que productos con diferentes características pueden ser obtenidos de un mismo proceso de producción debido a factores incontrolables como la naturaleza de las materias primas o las condiciones ambientales durante la producción. Hay cuatro aspectos que caracterizan la falta de homogeneidad en el producto: los subtipos homogéneos que se obtienen de un mismo lote de producción, la cantidad de productos que componen cada subtipo, el valor de cada uno de los subtipos, y el estado de los productos. La falta de homogeneidad en el producto dificulta la gestión de los procesos de las empresas y cadenas de suministro en el momento en el que los clientes requieren homogeneidad entre las unidades de producto que adquieren. Un ejemplo de esto se produce en el sector de la cerámica, en el que los clientes requieren que todas las unidades que van a ser ensambladas juntas tengan el mismo color, espesor y calidad por razones estéticas y de seguridad. Otro ejemplo es el extraído del sector agroalimentario, en el que el mercado final requiere productos que cumplan con un tamaño mínimo, un color particular, o sabor en el caso de las frutas. Además, el sector agroalimentario tiene la complejidad añadida producida por el deterioro de los productos a lo largo del tiempo, y la necesidad de los mercados de ofrecer a los clientes productos con una mínima duración tras su venta. En esta Tesis, se define como productos heterogéneos a aquellos productos que se pueden clasificar en subtipos homogéneos con una cantidad variable, mientras que los productos perecederos son aquellos que, además de ser heterogéneos, tienen falta de homogeneidad en su estado. De acuerdo con estos conceptos, el sector cerámico comercializa productos heterogéneos mientras que el sector agroalimentario comercializa productos perecederos. Esta Tesis propone marcos conceptuales y modelos de Investigación Operativa que soporten la gestión de cadenas de suministro con productos heterogéneos y perecederos en la toma de decisiones centralizada y distribuidas relacionadas con los niveles de decisión estratégica, táctica y operativa. El objetivo es mejorar la competitividad, sostenibilidad y flexibilidad de la cadena de suministro para adaptarse a los requerimientos del mercado bajo condiciones de incertidumbre. Para esto, se han propuesto modelos de Investigación Operativa deterministas e inciertos, cuyos resultados se comparan concluyendo que los resultados obtenidos con los modelos inciertos se adaptan mejor al comportamiento real de las cadenas de suministros.Los modelos de Investigación Operativa propuestos han contribuido a tres áreas de investigación: problemas operativos en el sector cerámico, problemas estratégicos en el sector agroalimentario y problemas de planificación en el sector agroalimentario. Las principales novedades en los problemas operativos en el sector cerámico son el modelado de las características de las baldosas cerámicas, la consideración de los requerimientos de homogeneidad entre unidades de diferentes líneas de pedido, y la posibilidad de realizar entregas parciales y entregas con retraso. Esta Tesis contribuye a los problemas estratégicos en el sector agroalimentario al diseñar una cadena de suministro completa de productos agroalimentarios frescos considerando el aspecto perecedero de los productos e integrando decisiones tácticas, y determinando el impacto real que tiene considerar el aspecto perecedero de los productos durante el diseño de la cadena de suministro ... / [CA] Alguns productes es caracteritzen per la seua falta d'homogeneïtat, el que significa que productes amb diferents característiques poden ser obtinguts d'un mateix procés de producció degut a factors incontrolables com la naturalesa de les matèries primeres o les condicions ambientals durant la producció. Hi ha quatre aspectes que caracteritzen la falta d'homogeneïtat en el producte: els subtipus homogenis que s'obtenen d'un mateix lot de producció, la quantitat de productes que componen cada subtipus, el valor de cada un dels subtipus, i l'estat dels productes. La falta d'homogeneïtat en el producte dificulta la gestió dels processos de les empreses i cadenes de subministrament en el moment en què els clients requerixen homogeneïtat entre les unitats de producte que adquirixen. Un exemple d'açò es produïx en el sector de la ceràmica, en el que els clients requerixen que totes les unitats que seran acoblades juntes tinguen el mateix color, grossària i qualitat per raons estètiques i de seguretat. Un altre exemple és l'extret del sector agroalimentari, en el que el mercat final requerix productes que complisquen amb una grandària mínima, un color particular, o sabor en el cas de les fruites. A més, el sector agroalimentari té la complexitat afegida produïda pel deteriorament dels productes al llarg del temps, i la necessitat dels mercats d'oferir als clients productes amb una mínima duració després de la seua venda. En aquesta Tesi, es definix com a productes heterogenis a aquells productes que es poden classificar en subtipus homogenis amb una quantitat variable, mentres que els productes peribles són aquells que, a més de ser heterogenis, tenen falta d'homogeneïtat en el seu estat. D'acord amb aquests conceptes, el sector ceràmic comercialitza productes heterogenis mentres que el sector agroalimentari comercialitza productes peribles. Aquesta Tesi proposa marcs conceptuals i models d'Investigació Operativa que suporten la gestió de cadenes de subministrament amb productes heterogenis i peribles en la presa de decisions centralitzada i distribuïdes relacionades amb els nivells de decisió estratègica, tàctica i operativa. L'objectiu és millorar la competitivitat, sostenibilitat i flexibilitat de la cadena de subministrament per adaptar-se als requeriments del mercat sota condicions d'incertesa. Per a açò, s'han proposat models d'Investigació Operativa deterministes i incerts, els resultats es comparen concloent que els resultats obtinguts amb els models incerts s'adapten millor al comportament real de les cadenes de subministraments. Els models d'Investigació Operativa proposats han contribuït a tres àrees d'investigació: problemes operatius en el sector ceràmic, problemes estratègics en el sector agroalimentari i problemes de planificació en el sector agroalimentari. Les principals novetats en els problemes operatius en el sector ceràmic són el modelatge de les característiques de les rajoles ceràmiques, la consideració dels requeriments d'homogeneïtat entre unitats de diferents línies de comanda, i la possibilitat de realitzar lliuraments parcials i lliuraments amb retard. Aquesta Tesi contribueix als problemes estratègics en el sector agroalimentari al dissenyar una cadena de subministrament completa de productes agroalimentaris frescos considerant l'aspecte perible dels productes, integrant decisions tàctiques, i determinant l'impacte real que té considerar l'aspecte perible dels productes durant el disseny de la cadena de subministrament ... / [EN] Some products are characterised by their lack of homogeneity, what means that products with different characteristics can be obtained from the same production process due to uncontrollable factors such as the nature of raw materials or the environmental conditions during production. There are four aspects that characterize the lack of homogeneity in the product: the homogeneous subtypes to be obtained from a production lot, the quantity of products that belong to each subtype, the value related to each of the subtypes and the state of the products. The lack of homogeneity in the product hinders the management of the supply chain or company's processes at the time customers require the homogeneity among the acquired units of product. An example of this is produced in the ceramic tile sector, in which customers need all acquired ceramic tiles that are going to be jointly assembled to have the same colour, thickness and quality for aesthetic and safety reasons. Another example is the extracted from the agri-food sector, in which final markets require products that meet some characteristics such as a minimum size, a particular colour or flavour in the case of fruits. In addition, the agri-food sector has the added complexity produced by the deterioration of products over time, and the need of markets to offer to end consumers products with a minimum durability after sale. In this Thesis, heterogeneous products are defined as products for which different subtypes can be obtained in a variable quantity while perishable products are those that, apart from being heterogeneous, have a lack of homogeneity in their state. According to these concepts, ceramic sectors would commercialize heterogeneous products while the agri-food sector would do so with perishable products. This Thesis proposes conceptual frameworks and Operations Research models to support the management of supply chains with heterogeneous and perishable products in centralized and distributed decision-making processes related to strategic, tactical and operative decisional levels. The objective is to improve the supply chain competitiveness, sustainability and flexibility to adapt to market requirements under uncertain conditions. For this, both deterministic and uncertain Operations Research models have been proposed, whose results are compared concluding that results obtained with uncertain models better fit with the behaviour of real supply chains. The proposed Operations Research models have contributed to three research areas: operational problems in the ceramic sector, strategic problems in the agri-food sector and planning problems in the agri-food sector. Main novelties in the ceramic operational problems are the modelling of the characteristics of ceramic tile products, the consideration of homogeneity requirements between units from different order lines, and the possibility of making partial deliveries and delayed deliveries. This Thesis contributes to strategic problems in agri-food products by designing an entire fresh agri-food supply chain considering the perishability of products and integrating tactical decisions, and by determining the real impact that considering the products' perishability has on the supply chain design process ... / This Thesis has been developed in the Research Centre of Management and Production Engineering (CIGIP, for its acronym in Spanish “Centro de Investigación en Gestión e Ingeniería de Producción”) of the Universitat Politècnica de València with the support of the predoctoral grant Programme of Formation of University Professors (FPU, for its acronym in Spanish “Formación de Profesorado Universitario”) from the Spanish Ministry of Science, Innovation and Universities (Ref. FPU15/03595). The supervisors of this Thesis are Dr. Angel Ortiz, and Dra. María del Mar Alemany Diaz that are Professors in the Research Centre of Management and Production Engineering (CIGIP) of the Universitat Politècnica de València. The FPU grant has been endorsed by the supervisor Dr. Ángel Ortiz. This Thesis has also been supported by the project ‘RUC-APS: Enhancing and implementing Knowledge based ICT solutions within high Risk and Uncertain Conditions for Agriculture Production Systems’ (Ref. 691249) funded by the EU under its funding scheme H2020-MSCA-RISE-2015, the project ‘Methods and models for operations planning and order management in supply chains characterised by uncertainty in production due to the lack of product uniformity’ (PLANGES-FHP) (Ref. DPI2011- 23597) funded by the Spanish Ministry of Economy and Competitiveness. The projects RUC-APS and PLANGES-FHP have been led by the one of the supervisors of this Thesis Dr. María del Mar Eva Alemany. In order to obtain the international mention for this Thesis, three months of research stages have been made in the research agency Agenzia Lucana di Sviluppo e di Innovazione in Agricoltora, located in Metaponto (Italy) / Esteso Álvarez, A. (2020). Operations research models for the management of supply chains of perishable and heterogeneous products in uncertain contexts. Application to the agri-food and ceramic sectors [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/141099 / TESIS / Compendio