Home Page

Public servers

TAC SCM 2008

TAC SCM 2009

TAC SCM 2010

TAC SCM 2011

TAC SCM Server

Team Members

MinneTAC source code

MinneTAC-05 binary

MinneTAC-06 binary

MinneTAC-09 binary

Other TAC-SCM binaries

Papers

Links





Members Home

Papers

Many enterprises that participate in dynamic markets need to make product pricing and inventory resource utilization decisions in real-time. We describe a family of statistical models that address these needs by combining characterization of the economic environment with the ability to predict future economic conditions to make tactical (short-term) decisions, such as product pricing, and strategic (long-term) decisions, such as level of finished goods inventories. Our models characterize economic conditions, called economic regimes, in the form of recurrent statistical patterns that have clear qualitative interpretations. We show how these models can be used to predict prices, price trends, and the probability of receiving a customer order at a given price. These ``regime'' models are developed using statistical analysis of historical data, and are used in real-time to characterize observed market conditions and predict the evolution of market conditions over multiple time scales. We evaluate our models using a testbed derived from the Trading Agent Competition for Supply Chain Management (TAC SCM), a supply chain environment characterized by competitive procurement and sales markets, and dynamic pricing. We show how regime models can be used to inform both short-term pricing decisions and long-term resource allocation decisions. Results show that our method outperforms more traditional short- and long-term predictive modeling approaches.

The TAC Energy Trading Broker is a design framework on which to build working software agents. The framework will include sufficient capability to act as a broker in the TAC Energy scenario with minimal capabilities. The broker framework is designed to lower the barrier to entry that, in other TAC scenarios, imposes huge challenges for agent developers who are not strong software developers. The framework leverages the Repast Simphony toolkit so that the developers can build their brokers totally using the Repast's graphical interface. The framework also provides APIs for those developers who want to build their brokers from scratch. Additionally, the framework design is flexible, extensible, and its components will be independently testable. This makes the framework easy for framework developers to understand and maintain.

Over the years, competitions have been important catalysts for progress in Artificial Intelligence. We describe one such competition, the Trading Agent Competition for Supply Chain Management (TAC SCM). We discuss its significance in the context of today's global market economy as well as AI research, the ways in which it breaks away from limiting assumptions made in prior work, and some of the advances it has engendered over the past six years. TAC SCM requires autonomous supply chain entities, modeled as agents, to coordinate their internal operations while concurrently trading in multiple dynamic and highly competitive markets. Since its introduction in 2003, the competition has attracted over 150 entries and brought together researchers from AI and beyond in the form of 75 competing teams from 25 different countries.

Configurable, metadata-driven systems provide an attractive design philosophy for large, modular applications due to their scalable, easy-to- test natures. However, their size still presents challenges to developers in that it is difficult to avoid duplicating code, and to users in that one must be familiar with a very large number of components to be able to configure such a system effectively. In this paper I present a method of si- multaneously reducing the sizes of these lists and the amount of redundant component code using source code annotations. I use these annotations as a basis for an ontology that in turn helps reduce the number of components one must be familiar with to create a system configuration. This helps reduce the cognitive burden of using the system to a reasonable level. I additionally employ these annotations to eliminate a large amount of redundant, uncoordinated code and enforce a link between components’ behavior and semantic metadata.

An effective Decision Support System (DSS) should help its users improve decision-making in complex, information-rich, environments. We present a feature gap analysis that shows that current decision support technologies lack important qualities for a new generation of agile business models that require easy, temporary integration across organizational boundaries. We enumerate these qualities as DSS Desiderata, properties that can contribute both effectiveness and flexibility to users in such environments. To address this gap, we describe a new design approach that enables users to compose decision behaviors from separate, configurable components, and allows dynamic construction of analysis and modeling tools from small, single-purpose evaluator services. The result is what we call an “evaluator service network” that can easily be configured to test hypotheses and analyze the impact of various choices for elements of decision processes. We have implemented and tested this design in an interactive version of the MinneTAC trading agent, an agent designed for the Trading Agent Competition for Supply Chain Management.

An autonomous trading agent is a complex piece of software that must operate in a competitive economic environment. We identify the problem of decision coordination as a crucial element in the design of an agent for TAC SCM, and we review the published literature on agent design to discover a wide variety of approaches to this problem. We believe that the existence of such variety is an indication that much is yet to be learned about designing such agents.

We describe an approach to building a highly configurable, semi-autonomous agent that can support users playing a variety of roles in a supply-chain trading environment. The agent's decision processes are composed of networks of simple services that are described using an OWL ontology. The ontology describes both the abstract data structures that are produced and consumed by individual services, as well as the business meaning of these data elements. This approach supports goal-directed composition of services in order to generate performance dashboards, as well as direct injection of user input at arbitrary points in the network.

We propose a variation of the TAC SCM supply-chain trading competition, in which human decision-makers compete with fully-autonomous agents. Because of the complexity and time pressures of the competition environment, humans may be assisted by semi-autonomous agents, which could be modifications of existing agents. The research goal is to discover what kinds of decision support will make a human decision-maker most effective in this environment. We show how an existing agent might be modified to operate in this new competition by updating our MinneTAC agent into a highly configurable, semi-autonomous agent that can support human users playing a variety of roles in the modified competition environment. The agent's decision processes are composed of networks of simple services that are described using an OWL ontology. The ontology describes the structure of the service network, along with the structure and semantics of the data elements that are produced and consumed by individual services.

The energy sector will undergo fundamental changes over the next ten years. Prices for fossil energy resources are continuously increasing, there is an urgent need to reduce CO2 emissions, and the United States and European Union are strongly motivated to become more independent from foreign energy imports. These factors will lead to installation of large numbers of distributed renewable energy generators, which are often intermittent in nature. This trend conflicts with the current power grid control infrastructure and strategies, where a few centralized control centers manage a limited number of large power plants such that their output meets the energy demands in real time. As the proportion of distributed and intermittent generation capacity increases, this task becomes much harder, especially as the local and regional distribution grids where renewable energy generators are usually installed are currently virtually unmanaged, lack real time metering and are not built to cope with power flow inversions (yet). All this is about to change, and so the control strategies must be adapted accordingly. While the hierarchical command-and-control approach served well in a world with a few large scale generation facilities and many small consumers, a more flexible, decentralized, and self-organizing control infrastructure will have to be developed that can be actively managed to balance both the large grid as a whole, as well as the many lower voltage sub-grids. We propose a competitive simulation test bed to stimulate research and development of electronic agents that help manage these tasks. Participants in the competition will develop intelligent agents that are responsible to level energy supply from generators with energy demand from consumers. The competition is designed to closely model reality by bootstrapping the simulation environment with real historic load, generation, and weather data. The simulation environment will provide a low-risk platform that combines simulated markets and real-world data to develop solutions that can be applied to help building the self-organizing intelligent energy grid of the future.

Modern electronic commerce creates significant challenges for decision-makers. The trading agent competition for supply-chain management (TAC SCM) is an annual competition among fully-autonomous trading agents designed by teams around the world. Agents attempt to maximize profits in a supply-chain scenario that requires them to coordinate Procurement, Production, and Sales activities in competitive markets. An agent for TAC SCM is a complex piece of software that must operate in a competitive economic environment. We report on results of an informal survey of agent design approaches among the competitors in TAC SCM, and then we describe and evaluate the design of our MinneTAC trading agent. We focus on the use of evaluators - configurable, composable modules for data analysis, modeling, and prediction that are chained together at runtime to support agent decision-making. Through a set of examples, we show how this structure supports Sales and Procurement decisions, and how those decision process can be modified in useful ways by changing evaluator configurations.

We show how an autonomous agent can use observable market conditions to characterize the microeconomic situation of the market and predict future market trends. The agent can use this information to make both tactical decisions, such as pricing, and strategic decisions, such as product mix and production planning. We develop methods to learn dominant market conditions, such as over-supply or scarcity, from historical data using Gaussian mixture models to construct price density functions. We discuss how this model can be combined with real-time observable information to identify the current dominant market condition and to forecast market changes over a planning horizon. We forecast market changes via both a Markov correction-prediction process and an exponential smoother. Empirical analysis shows that the exponential smoother yields more accurate predictions for the current and the next day (supporting tactical decisions), while the Markov correction-prediction process is better for longer term predictions (supporting strategic decisions). Our approach offers more flexibility than traditional regression based approaches, since it does not assume a fixed functional relationship between dependent and independent variables. We validate our methods by presenting experimental results in a case study, the Trading Agent Competition for Supply Chain Management.

The MinneTAC trading agent is designed to compete in the Supply-Chain Trading Agent Competition. It is also designed to support the needs of a group of researchers, each of whom is interested in different decision problems related to the competition scenario. The design of MinneTAC breaks out each basic behavior into a separate, configurable component, and allows dynamic construction of analysis and modeling tools from small, single-purpose “evaluators”. The agent is defined as a set of “roles”, and a working agent is one for which a component is supplied for each role. This allows each researcher to focus on a single problem and work independently, and it allows multiple researchers to tackle the same problem in different ways. A working MinneTAC agent is completely defined by a set of configuration files that map the desired roles to the code that implements them, and that set parameters for the components. We describe the design of MinneTAC, and we evaluate its effectiveness in support of our research agenda and its competitiveness in the TAC-SCM game environment.

An autonomous trading agent is a complex piece of software that must operate in a competitive economic environment. We report results of an informal survey of agent design approaches among the competitors in the Trading Agent Competition for Supply Chain Management (TAC SCM). We identify the problem of decision coordination as a crucial element in the design of an agent for TAC SCM, and we review the published literature on agent design to discover a wide variety of approaches to this problem. We believe that the existence of such variety is an indication that much is yet to be learned about designing such agents.

An autonomous trading agent is a complex piece of software that must operate in a competitive economic environment and support a research agenda. We describe the structure of decision processes in the MinneTAC trading agent, focusing on the use of evaluators - configurable, composable modules for data analysis and prediction that are chained together at runtime to support agent decision-making. Through a set of examples, we show how this structure supports sales and procurement decisions, and how those decision process can be modified in useful ways by changing evaluator configurations.

We present the design of a service oriented architecture which facilitates flexible managerial decision making in dynamic business networks. We have implemented and tested this architecture in the MinneTAC trading agent, which is designed to compete in the Supply Chain Trading Agent Competition. Our design enables managers to break out decision behaviors into separate, configurable components, and allows dynamic construction of analysis and modeling tools from small, single-purpose “evaluator” services. The result of our design is that the network can easily be configured to test a new theory and analyze the impact of various approaches to different aspects of the agent's decision processes, such as procurement, sales, production, and inventory management. Additionally we describe visualizers that allow managers to see and manipulate the configuration of the network, and to construct economic dashboards that can display the current and historical state of any node in the network.

We present a semi-parametric model that describes and predicts pricing behaviors in a market environment using a Gaussian mixture model and a Markov process. We show how the model can be used to guide resource allocation and pricing decisions in an autonomous trading agent. We validate our model by presenting results obtained in the Trading Agent Competition for Supply Chain Management.

We present a semi-parametric model that describes pricing behaviors in a market environment, and we show how that model can be used to guide resource allocation and pricing decisions in an autonomous trading agent. We validate our model by presenting experimental results obtained in the Trading Agent Competition for Supply Chain Management.

We consider the problem of rational, self-interested, economic agents who must negotiate with each other in a market environment in order to carry out their plans. Customer agents express their plans in the form of task networks with temporal and precedence constraints. A combinatorial reverse auction allows supplier agents to submit bids specifying prices for combinations of tasks, along with time windows and duration data that the customer may use to compose a work schedule. We describe the consequences of allowing the advertised task network to contain schedule infeasibilities, and show how to resolve them in the auction winner-determination process.

Supply chain management is commonly employed by businesses to improve organizational processes by optimizing the transfer of goods, information, and services between buyers and suppliers. Traditionally, supply chains have been created and maintained through the interactions of human representatives of the various companies involved. However, the recent advent of autonomous software agents opens new possibilities for automating and coordinating the decision making processes between the various parties involved. Autonomous agents participating in supply chain management must typically make their decisions in environments of high complexity, high variability, and high uncertainty since only limited information is visible. We present an approach whereby an autonomous agent is able to make tactical decisions, such as product pricing, as well as strategic decisions, such as product mix and production planning, in order to maximize its profit despite the uncertainties in the market. The agent predicts future market conditions and adapts its decisions on procurement, production, and sales accordingly. Using a combination of machine learning and optimization techniques, the agent first characterizes the microeconomic conditions, such as over-supply or scarcity, of the market. These conditions are distinguishable statistical patterns that we call economic regimes. They are learned from historical data by using a Gaussian Mixture Model to model the price density of the different products and by clustering price distributions that recur across days. In real-time the agent identifies the current dominant market condition and forecasts market changes over a planning horizon. Methods for the identification of regimes are explored in detail, and three different algorithms are presented. One is based on exponential smoothing, the second on a Markov prediction process, and the third on a Markov correction-prediction process. We examine a wide range of tuning options for these algorithms, and show how they can be used to predict prices, price trends, and the probability of receiving a customer order. We validate our methods by presenting experimental results from the Trading Agent Competition for Supply Chain Management, an international competition of software agents that has provided inspiration for this work. We also show how the same approach can be applied to the stock market.

Market simulations, like their real-world counterparts, are typically domains of high complexity, high variability, and incomplete information. The performance of autonomous agents in these markets depends both upon the strategies of their opponents and on various market conditions, such as supply and demand. Because the space for possible strategies and market conditions is very large, empirical analysis in these domains becomes exceedingly difficult. Researchers who wish to evaluate their agents must run many test games across multiple opponent sets and market conditions to verify that agent performance has actually improved. Our approach is to improve the statistical power of market simulation experiments by controlling their complexity, thereby creating an environment more conducive to structured agent testing and analysis. We develop a tool that controls variability across games in one such market environment, the Trading Agent Competition for Supply Chain Management (TAC SCM), and demonstrate how it provides an efficient, systematic method for TAC SCM researchers to analyze agent performance.

We present methods for an autonomous agent to predict price distributions and price trends in the customer market of the Trading Agent Competition for Supply Chain Management. We describe how these predictions can then be used by the agent to make strategic and tactical sales decisions

We present methods for an autonomous agent to identify dominant market conditions, such as over-supply or scarcity, and to forecast market changes. We show that market conditions can be characterized by distinguishable statistical patterns that can be learned from historic data and used, together with real-time observable information, to identify the current market regime and to forecast market changes. We use a Gaussian Mixture Model to represent the probabilities of market prices and, by clustering these probabilities, we identify different economic regimes. We show that the regimes so identified have properties that correlate with market factors that are not directly observable. We then present methods to predict regime changes. We validate our methods by presenting experimental results obtained with data from the Trading Agent Competition for Supply Chain Management.