Abstract: Virtual organization refers to the temporary teaming of enterprises. By sharing physical, human and knowledge resources via information technologies, a virtual organization enables member enterprises to share skills, costs, access to one another’s markets and, at the same time decrease the risk of investments. To realize this new generation of business model, the ability to form and operate virtual enterprise is very important. The paper describes our experience gained by implementing a multiagent system that simulates an artificial marketplace, for which we have derived several decision-making mechanisms in various stages of a virtual organization. We presented a negotiation protocol and a bid selection algorithm for agents to form a virtual organization. We adopted the Motivational Quantities framework to support the agent’s local reasoning process. In order to better understand the organizational problem, we adapted a statistical model that predicts the expected rewards of individual agents and the performance of the virtual organization. The comparison and analysis of the results from both the simulation and the model prediction are also presented in this paper.
Introduction
In a globalized economy, companies often need to form temporary "Virtual Organizations" (VOs) to share costs, skills, and market access for specific projects. However, forming these manually is slow and complex. The challenge is twofold: first, coordination—how to automatically select the best partners from a pool of bidders to maximize profit; and second, incentives—how to ensure self-interested agents fulfill their promises when outside market opportunities arise.
Methodology
The authors implement a multi-agent system (MAS) that simulates an artificial marketplace using three core components:
- Recursive Best-First Search (RBFS): For the "Initiator" agent forming the VO, an RBFS algorithm with heuristics selects the optimal group of partners. It groups bids into "bins" based on task types and searches for the combination that maximizes expected profit.
- Motivational Quantities (MQ) Framework: This serves as the decision-making engine for individual agents. Instead of focusing solely on immediate cash rewards, agents use MQ to quantify "utility," factoring in intangible benefits like relationship maintenance and the avoidance of commitment penalties.
- Utility Mapping Functions: A specific mathematical function maps task rewards to local utility. By adjusting parameters ($a, b,$ and $c$), the system can model different penalty policies, such as linear penalties or progress-based thresholds.
Experimental Setup and Results
The study utilized the MASS Simulator and Java Agent Framework (JAF) to create a synthetic benchmark: the Virtual Building Organization Scenario. This involved a real estate developer and subcontractors performing sequential tasks: framing, foundation, electrical, plumbing, and finishing.
Key Insights
- Non-Monetary Utility: By treating "organizational health" as a resource via the MQ framework, agents naturally gravitate toward more stable and reliable VO commitments.
- The "b" Parameter as a Policy Knob: In the utility function, the parameter $b$ acts as a control for agent attitude. A $b$ value near 0.5 mimics a linear penalty, while a value near 0.9 is better for progress-based penalties, encouraging agents to complete critical early-stage tasks.
Performance Verification
The statistical model's predictions for conflict probability closely matched simulation results (e.g., predicting a conflict probability of 0.26 vs. an actual 0.21). Furthermore, the MQ mapping function successfully incentivized agents to stay with the VO even when outside tasks became more lucrative (reward range 1000–1500).
Conclusion and Future Work
While the framework successfully automates VO formation and stabilizes commitments, limitations remain. The RBFS algorithm may become computationally intractable as more variables (like quality of work) are added. Additionally, the current model assumes agents have prior knowledge of task frequencies, which is rare in volatile markets. Future research will focus on making these models more adaptive to unknown environments and exploring more complex initial bidding logic.
Summary updated