Negotiation and Monitoring in Open Environments

Abstract

Large scale, distributed, digital environments offer vast potential. Within these environments, software systems will provide unprecedented support for daily life. Offering access to vast amounts of knowledge and resources, these systems will enable wider participation of society, at large. An example is the Smart Energy Grid that increases sustainability and decreases reliance on fossil fuels. Such systems require technology that is capable of negotiating Service Level Agreements (SLAs) between consumers and providers. Multi Agent Systems (MAS) is one such technology that offers a straightforward analog for complex systems of autonomous parties. MAS is based on the notion of autonomous agents that represent human actors (i.e. owners) and are capable of negotiating SLAs and coordinating processes with other agents. They know their owner’s preferences and needs. They are capable of negotiating price, Quality of Service (QoS) characteristics and penalties. They also monitor provisioning of services to detect and penalize service violations. This dissertation presents a MAS framework for automated negotiation and monitoring of SLAs in open environments. In this context, an open environment is a large-scale, distributed environment that is also dynamic and untrusted. This framework enables secure discovery, negotiation and access to distributed resources. Through a process of exchanging messages (e.g. offers, counter-offers), agents together search for a mutually acceptable agreement (e.g. service, price, quality). A negotiation protocol defines the negotiation objects (i.e. offers), language and rules governing interaction. This dissertation presents the WS-Agreement Negotiation protocol with extensions for open environments. This protocol is experimentally validated in the AgentScape middleware. Open environments also present challenges regarding security, trust and privacy. No single authority has complete control over an open environment and no single authority governs the actions of all participants (i.e. agents). Therefore, additional mechanisms are required to ensure security, privacy and promote trust between participants. Automated monitoring mechanisms using a Trusted Third Party (TTP) address issues of security and thus support negotiation in open environments. This dissertation presents a self-adaptive monitoring approach that (1) offers monitoring assurance that agreements are honored, (2) builds a secure audit log of agreement compliance, (3) performs measurements while safeguarding privacy of (sensitive) data, (4) dynamically reacts to changes in risk and (5) enables trust-building between consumers and providers. This monitoring approach is experimentally validated in the AgentScape middleware. Automation of complex tasks, such as negotiation, can increase efficiency. To illustrate these benefits, the framework is applied to two complex systems, including the Smart Energy Grid. In this case, the looming complexity crisis of intermittent generation, real-time pricing and consumer demand shifting requires immediate attention. This domain presents not only technical (e.g. smart-meters) but also social challenges (e.g. user acceptance). The MAS automation framework presented in this dissertation addresses technical challenges by reducing manual labor and increasing efficiency. Automation even enables higher utilization of green resources and reduction of waste (e.g. produced, but unconsumed energy). Transparent, trusted monitoring mechanisms address social challenges by ensuring privacy of (sensitive) data and encouraging user acceptance. Software systems, such as those presented in this dissertation, enable wider participation of society, at large, and offer vast potential.