Survivability of Complex System – Support Vector Machine Based Approach

Logistic systems which are inherently distributed, in general can be classified as complex systems. Survivability of these systems under varying environment conditions is of paramount importance. Different environmental conditions in which the logistic system resides are translated into several stresses. These in turn will manifest as internal stresses. Logistic systems can be modeled as a collection of software agents. Each agent’s behavior is a result of the stresses imposed. Predicting the agents’ collective behavior is of paramount importance to ensure survivability. Analytical modeling of such systems becomes very difficult, albeit impossible. In this paper, we study a supply chain in which a real life scenario is used. We implement the supply chain in Cougaar (Cognitive Agent Architecture developed by DARPA) and develop a predictor, based on Support Vector Machine. We report our methodology and results with real-life experiments and stress scenarios. INTRODUCTION Logistic systems can be classified as complex systems (Choi et al., 2001, Baranger, http://necsi.org/projects/baranger/cce.pdf). Logistic systems have many components such as suppliers and distributors at several stages. These components are distributed geographically but interdependent. At each component some form of nonlinear decision making process goes on. Typically the system would respond in a stable manner to external disturbances. But due to information delay, inherent feedback structure and nonlinear components unstable phenomena can arise which may ultimately manifest as chaotic behavior. Efficient resource allocation and collective oscillations (of say inventory levels) are some examples of emergent behavior shown by supply chains. They have structure at many scales, each component itself represents a simple supply chain. The components compete due to resource limitation but collobarate/cooperate to maximize their gains which is another characteristic feature of a complex system. The survivability of logistic systems under varying environmental conditions is of paramount importance. Survivability is going to be itself an emergent property of a logistic system and it represents the ability of the system