Thesis (M.S.)--Rochester Institute of Technology, 2009. / Typescript. Includes bibliographical references (leaves 54-56).

Supplier selection is an important problem in supply chain management (SCM), and has attracted the attention of many researchers. Most previous research on supplier selection is based on the assumption that a single product is required. For the few supplier selection models for multiple products, they handled the problem on a product-by-product manner. In such cases, the synergy effect between products which could impact the choice of cooperative suppliers is not taken into account. However, it is practical for the purchasing company to procure multiple products simultaneously and benefit from the synergy effect between products. It is necessary to incorporate the synergy effect between products in multi-product supplier selection. This thesis presents a multi-product supplier selection model incorporating the synergy effect between products. The model is composed of three sub-models, i.e., the synergy determination sub-model, the supplier pre-selection sub-model and the negotiation-based final selection sub- model. As the agent-based technology is a natural tool for modeling distributed systems, the proposed multi-product supplier selection model is realized as a multi-agent system (MAS) with agents representing the relevant parties and functions of the proposed model. Agents of the MAS are able to interact with each other through the respective agent interaction protocols defined specifically for the three sub-models. The synergy determination sub-model is to determine the synergy effect between products. The term complementarity is used to represent the synergy effect between products. The product complementarity measure criteria are formulated based on the activities of automobile manufacturers. Complementarity measure methods are then proposed. The product bundle determination algorithm is presented to generate preferred product bundles. The interaction of agents involving in the sub-model is governed by the synergy determination protocol. The supplier pre-selection sub-model is to shortlist the qualified and competitive suppliers for multiple products. The pre-selection criteria catering for the multi-product environment are formulated. Both the general characteristics and performances of suppliers, and the capabilities supporting multi-product transactions are included in the pre-selection criteria. The TOPSIS-based supplier pre-selection algorithm is established to evaluate suppliers on these criteria. The interaction of agents involving in the sub-model is governed by the pre-selection protocol. The negotiation-based final selection sub-model is to select the cooperative suppliers for multiple products. In order to cater for the multi-product environment, multiple bids are allowed in the negotiation model. The corresponding bid utility function and negotiation strategies are presented. The B&B-based winner determination algorithm is presented to determine the cooperative suppliers. The hybrid protocol of combinatorial procurement auction and multi-bilateral bargaining is established to govern the interaction of agents in the sub-model. A case study has been executed to demonstrate the feasibility, effectiveness and usefulness of the supplier selection model for multiple products with synergy effect. The results indicate that the proposed supplier selection model is able to select suppliers for multiple products simultaneously and incorporate the synergy effect between products. In addition, the agent interaction protocols and related algorithms used in the agent-based system supporting the multi-product supplier selection model are suitable and effective. / published_or_final_version / Industrial and Manufacturing Systems Engineering / Doctoral / Doctor of Philosophy

To implement green supply chain, a company has to consider sustainability impacts in assessing potential suppliers. Thereby, the supplier evaluation and selection criteria would include various key performance indicators such as price, quality, delivery, as well as environmental and corporate social responsibility aspects. Researchers have proposed numerous multi-criteria decision making (MCDM) approaches for evaluating the multiple conflicting criteria in supplier selection. However, most of the existing approaches have ignored some important issues in business environment such as supplier qualification, supplier autonomy, negotiation between manufacturer and supplier, etc. In this thesis, a multi-agent system (MAS) is proposed for supplier selection in green supply chain. It comprises two types of autonomous agents, namely, buyer agent and seller agents, to represent the interests of manufacturer and suppliers, respectively. The proposed MAS presents three prominent features. First, the proposed supplier selection criteria incorporate the conventional, environmental and social aspects in the supplier selection process. The criteria are classified into negotiable and non-negotiable criteria. Initially, all criteria are included to evaluate and rank all the candidate suppliers. Subsequently, the top-ranked candidates are invited to participate in the bargaining process. In this regard, the negotiable criteria are used for assessing the quality of an offer, while the non-negotiable criteria, i.e. environmental and social criteria, influence the manufacturer’s negotiation attitude to candidate suppliers. The classification enables the manufacturer to fully utilize the performance values of all criteria. Secondly, supplier selection is implemented in a two-stage methodology. The TOPSIS method is devised in the first stage to shortlist some suitable candidate suppliers for entering negotiation in the next stage. In the second stage, the agent-based negotiation process is adopted for selecting the final supplier. Representing the manufacturer and the shortlisted suppliers respectively, the buyer and seller agents bargain on a number of negotiable issues in the multi-round negotiation. A multi-issue and multilateral agent interaction protocol, which is an extension of the contract net protocol, is implemented in the MAS. Accordingly, the buyer agent coordinates with the seller agents to exchange offers and counteroffers. Thirdly, a novel preference-based negotiation strategy is used to govern the behavior of agents during negotiation. A heuristics model with the Particle Swarm Optimization (PSO) algorithm and Adaptive Penalty Function has been designed and implemented to realize the proposed negotiation strategy. The strategy guides the autonomous agents to narrow down the discrepancies in the values of the negotiable criteria (price, delivery days, contract length) in their offers, and simultaneously search a mutually beneficial and acceptable agreement. The negotiation payoffs and negotiation time are improved. Experimental results indicate that the proposed agent-based model could help the manufacturer to identify the most appropriate supplier and improve the quality of final agreement. In addition, the model successfully integrates supplier qualification and automated negotiation, and promotes supplier autonomy in the supplier selection process. / published_or_final_version / Industrial and Manufacturing Systems Engineering / Master / Master of Philosophy

Given the continuing improvements in design and manufacturing processes in addition to improvements in artificial intelligence, robots are being deployed in an increasing variety of environments for longer periods of time. As the number of robots grows, it is expected that they will encounter and interact with other robots. Additionally, the number of companies and research laboratories producing these robots is increasing, leading to the situation where these robots may not share a common communication or coordination protocol. While standards for coordination and communication may be created, we expect that any standards will lag behind the state-of-the-art protocols and robots will need to additionally reason intelligently about their teammates with limited information. This problem motivates the area of ad hoc teamwork in which an agent may potentially cooperate with a variety of teammates in order to achieve a shared goal. We argue that agents that effectively reason about ad hoc teamwork need to exhibit three capabilities: 1) robustness to teammate variety, 2) robustness to diverse tasks, and 3) fast adaptation. This thesis focuses on addressing all three of these challenges. In particular, this thesis introduces algorithms for quickly adapting to unknown teammates that enable agents to react to new teammates without extensive observations. The majority of existing multiagent algorithms focus on scenarios where all agents share coordination and communication protocols. While previous research on ad hoc teamwork considers some of these three challenges, this thesis introduces a new algorithm, PLASTIC, that is the first to address all three challenges in a single algorithm. PLASTIC adapts quickly to unknown teammates by reusing knowledge it learns about previous teammates and exploiting any expert knowledge available. Given this knowledge, PLASTIC selects which previous teammates are most similar to the current ones online and uses this information to adapt to their behaviors. This thesis introduces two instantiations of PLASTIC. The first is a model-based approach, PLASTIC-Model, that builds models of previous teammates' behaviors and plans online to determine the best course of action. The second uses a policy-based approach, PLASTIC-Policy, in which it learns policies for cooperating with past teammates and selects from among these policies online. Furthermore, we introduce a new transfer learning algorithm, TwoStageTransfer, that allows transferring knowledge from many past teammates while considering how similar each teammate is to the current ones. We theoretically analyze the computational tractability of PLASTIC-Model in a number of scenarios with unknown teammates. Additionally, we empirically evaluate PLASTIC in three domains that cover a spread of possible settings. Our evaluations show that PLASTIC can learn to communicate with unknown teammates using a limited set of messages, coordinate with externally-created teammates that do not reason about ad hoc teams, and act intelligently in domains with continuous states and actions. Furthermore, these evaluations show that TwoStageTransfer outperforms existing transfer learning algorithms and enables PLASTIC to adapt even better to new teammates. We also identify three dimensions that we argue best describe ad hoc teamwork scenarios. We hypothesize that these dimensions are useful for analyzing similarities among domains and determining which can be tackled by similar algorithms in addition to identifying avenues for future research. The work presented in this thesis represents an important step towards enabling agents to adapt to unknown teammates in the real world. PLASTIC significantly broadens the robustness of robots to their teammates and allows them to quickly adapt to new teammates by reusing previously learned knowledge. / text

This dissertation proposes a suite of novel approaches for solving multiagent decision and optimization problems based on the Collaborative Design Network (CDN), a framework for multiagent decision making. The framework itself is distributed, decision-theoretic and was originally proposed for multiagent component-centred design. This application is a novel use of the CDN, and demonstrate the generality of the CDN framework for general decision-theoretic planning. First, the framework is applied towards tackling a multiagent decision problem outside of collaborative design called multiagent expedition (MAE), a testbed problem which abstracts many of the features of real-world multiagent decision-making problems. We formally introduce MAE, and show it to be a subclass of a decentralized partially observable Markov Decision process (Dec-POMDP). We apply the CDN to the online MAE planning problem. We demonstrate that the CDN can plan in MAE with conditional optimality given a set of basic assumptions on the structure and organization of the agent team. We introduce a set of knowledge representational aspects to achieve conditionally optimal planning. We experimentally verify our approach on a series of benchmark problems created for this dissertation to test the various aspects of our CDN solution. We also investigate further methods for scalability and speedup in MAE. The concept of \emph{partial evaluation} (PE) is introduced, based on the assumption that an agent has an intended effect given an agent's action and considers all other effects unintended. This assumption is used to derive a bound for planning that partitions the set of joint plans into a set of fully evaluated and a set of partial evaluated plans. Plans which are partially evaluated can significantly speed up planning in the centralized case. PE is also applied to the CDN, to both public decisions between agents and private decisions local to an agent. We demonstrate that applying PE to public decisions in the CDN results in either intractable communication or suboptimal planning. When applied to private decisions, we show PE can still be very effective in decreasing planning runtime.

This thesis provides an artificial intelligence approach for the problem of resource allocation in time-critical dynamic environments. Motivated by healthcare scenarios such as mass casualty incidents, we are concerned with making effective decisions about allocating to patients the limited resources of ambulances, doctors and other medical staff members, in real-time, under changing circumstances. We cover two distinct stages: the Ambulance stage (at the location of the incident) and the Hospital stage (where the patient requires treatment). Our work addresses both determining the best allocation and supporting decision making (for medical staff to explore possible options). Our approach uses local search with social welfare functions in order to find the best allocations, making use of a centralized tracking of patients and resources. We also clarify how sensing can assist in updating the central system with new information. A key concept in our solution is that of a policy that attempts to minimize cost and maximize utility. To confirm the value of our approach, we present a series of detailed simulations of ambulance and hospital scenarios, and compare algorithms with competing principles of allocation (e.g. sickest first) and societal preferences (e.g. egalitarian allotment). In all, we offer a novel direction for resource allocation that is principled and that offers quantifiable feedback for professionals who are engaged in making resource allocation decisions.

Within the artificial intelligence subfield of multiagent systems, one challenge that arises is determining how to efficiently allocate resources to all agents in a way that maximizes the overall expected utility. In this thesis, we explore a distributed solution to this problem, one in which the agents work together to coordinate their requests for resources and which is considered to be ex-ante rational: in other words, requiring agents to be willing to give up their current resources to those with greater need by reasoning about what is for the common good. Central to our solution is allowing for preemption of tasks that are currently occupying resources; this is achieved by introducing a concept from adjustable autonomy multiagent systems known as a transfer of control (TOC) strategy. In essence a TOC strategy is a plan of an agent to acquire resources at future times, and can be used as a contingency plan that an agent will execute if it loses its current resource. The inclusion of TOC strategies ultimately provides for a greater optimism among agents about their future resource acquisitions, allowing for more generous behaviours, and for agents to more frequently agree to relinquish current resources, resulting in more effective preemption policies. Three central contributions arise. The first is an improved methodology for generating transfer of control strategies efficiently, using a dynamic programming approach, which enables a more effective employment of TOCs in our resource allocation solution. The second is an important clarification of the value of integrating learning techniques in order for agents to acquire improved estimates of the costs of preemption. The last is a validation of the overall multiagent resource allocation (MARA) solution, using simulations which show quantifiable benefits of our novel approach. In particular, we consider in detail the emergency medical application of mass casualty incidents and are able to demonstrate that our approach of integrating transfer of control strategies results in effective allocation of patients to doctors: ones which in simulations re- sult in dramatically fewer patients in a critical healthstate than are produced by competing MARA algorithms. In short, we offer a principled solution to the problem of preemption, allowing the elimination of a source of inefficiencies in fully distributed multiagent resource allocation systems; a faster method for generation of transfer of control strategies; and a convincing application of the system to a real world problem where human lives are at stake.

Motivated by shortcomings in recent natural disaster responses; this paper reports on a computational approach that offers techniques for matching social demands of a disaster type with the strengths of cultural traits among rescue teams.

Normative systems have been proposed as a useful abstraction to represent ideals of behaviour for autonomous agents in a social context. They specify constraints that agents ought to follow, but may sometimes be violated. Norms can increase the predictability of a system and make undesired situations less likely. When designing normative systems, it is important to anticipate the effects of possible violations and understand how robust these systems are to violations. Previous research on robustness analysis of normative systems builds upon simplistic norm formalisms, lacking support for the specification of complex norms that are often found in real world scenarios. Furthermore, existing approaches do not consider the fact that compliance with different norms may be more or less important in preserving some desirable properties of a system; that is, norm violations may vary in severity. In this thesis we propose models and algorithms to represent and reason about complex norms, where their violation may vary in severity. We build upon existing preference-based deontic logics and propose mechanisms to rank the possible states of a system according to what norms they violate, and their severity. Further, we propose mechanisms to analyse the properties of the system under different compliance assumptions, taking into account the severity of norm violations. Our norm formalism supports the specification of norms that regulate temporally extended behaviour and those that regulate situations where other norms have been violated. We then focus on algorithms that allow coalitions of agents to coordinate their actions in order to minimise the risk of severe violations. We propose offline algorithms and heuristics for pre-mission planning in stochastic scenarios where there is uncertainty about the current state of the system. We then develop online algorithms that allow agents to maintain a certain degree of coordination and to use communication to improve their performance.

We present tools for a network topology based characterization of heterogeneity in multiagent systems, thereby providing a framework for the analysis and design of heterogeneous multiagent networks from a network structure view-point. In heterogeneous networks, agents with a diverse set of resources coordinate with each other. Coordination among different agents and the structure of the underlying network topology have significant impacts on the overall behavior and functionality of the system. Using constructs from graph theory, a qualitative as well as a quantitative analysis is performed to examine an inter-relationship between the network topology and the distribution of agents with various capabilities in heterogeneous networks. Our goal is to allow agents maximally exploit heterogeneous resources available within the network through local interactions, thus exploring a promise heterogeneous networks hold to accomplish complicated tasks by leveraging upon the assorted capabilities of agents. For a reliable operations of such systems, the issue of security against intrusions and malicious agents is also addressed. We provide a scheme to secure a network against a sequence of intruder attacks through a set of heterogeneous guards. Moreover, robustness of networked systems against noise corruption and structural changes in the underlying network topology is also examined.