The MarCon Algorithm: A Systematic Market Approach to Distributed Constraint Problems

Abstract

MarCon (Market-based Constraints) applies market-based control to distributed constraint problems. It offers a new approach to distributing constraint problems that avoids challenges faced by current approaches in some problem domains, and it provides a systematic method for applying markets to a wide array of problems. Constraint agents interact with one another via the variable agents in which they have a common interest, using expressions of their preferences over sets of assignments. Each variable integrates this information from the constraints interested in it and provides feedback that enables the constraints to shrink their sets of assignments until they converge on a solution. MarCon originated in a system for supporting human product designers, and its mechanisms are particularly useful for applications integrating human and machine intelligence to explore implicit constraints. MarCon has been tested in the domain of mechanical design, in which its set-narrowing process is particularly useful.

What Is Set‐Based Design?

“…The first group of experiments used the Responsible Agents for Product/Process Integrated Development (RAPPID) product, developed by The Center for Electronic Commerce of the Environmental Research Institute of Michigan (ERIM). RAPPID facilitated negotiation between the human agents to foster SBD and was designed to help human designers manage product characteristics across different functions and stages in the product life cycle (Parunak et al 1998, 1999a, b).…”

Designing for incentives

“…The computational market could in principle be extended to represent the preferences of multiple stakeholders, but was studied only in the context of a single customer and thus one set of preferences and avoiding the need to address any incentive challenges. Parunak et al [15] introduced the MarCon algorithm for the support of a design process between multiple human product designers, but without providing any formal incentive properties. Both papers contribute the framing of preference elicitation and constrained optimization, but consider only static problems in which participants commit to a one-time decision about the design of an engineering artifact.…”

“…But it seems unnatural for payments to be made in this direction, except perhaps when the PPs are independent contractors and looking for tasks in order to gain experience and/or reputation. For example, suppose that there are levels L and H for feature 1 and the values are (0,20), (5,5) and (15,10), for level L and H for each of the first PP, second PP and the PM respectively. The optimal decision is level H, with total value 35.…”

Contributions to an Electronic Institution Supporting Virtual Enterprises' Life Cycle

“…One of the requirements for the negotiation protocol we are here proposing, besides dealing with attributes intra-dependencies, is the capability to deal with attributes' interdependencies. This is an important requirement to be considered in our scenario, because in the VE formation process interdependent negotiations take place simultaneously, and The distributed dependencies satisfaction problem has been the subject of attention of other researchers, addressing the study of both single (Yokoo, Durfee, Ishida, & Kuwabara, 1992) and multiple dependent variables (Armstrong & Durfee, 1997;Parunak, Ward, & Sauter, 1999). In the VE formation process, dependencies may occur between multiple variables, making the latter approaches more relevant to our research.…”

“…The first mentioned paper, (Armstrong & Durfee, 1997) describes algorithms to reach one possible solution, not the optimal one. The second paper (Parunak et al, 1999) introduces an algorithm that, although reaching the optimal solution, imposes that all agents involved in the mutual dependencies resolution process have to know all agents' private utility functions.…”