Akin Günay

AAMAS Conference 2019 Conference Paper

Supple: Multiagent Communication Protocols with Causal Types

• Akin Günay
• Amit K. Chopra
• Munindar P. Singh

A (communication) protocol captures how agents collaborate by specifying the messages they exchange. In particular, since the information content of messages characterizes the interactions a protocol specifies, message types can improve collaboration by strengthening the specification of what each agent may legitimately expect from another agent. In addition, in implementations, typing information can enable improved verification of agents. We introduce Supple, a protocol specification language that expresses message schemas with typed parameters. Supple enables definition of causal types for parameters that constrain how other parameters are computed in a protocol enactment. We give the formal semantics of Supple; characterize the liveness and safety of Supple specifications; and provide decision procedures for them.

AAMAS Conference 2016 Conference Paper

Dynamically Generated Commitment Protocols in Open Systems (JAAMAS Extended Abstract)

• Akin Günay
• Michael Winikoff
• Pinar Yolum

This extended abstract presents our work “Dynamically Generated Commitment Protocols in Open Systems” that appeared in the Journal of Autonomous Agents and Multi-Agent Systems [1].

ECAI Conference 2016 Conference Paper

GPU-Accelerated Value Iteration for the Computation of Reachability Probabilities in MDPs

• Zhimin Wu
• Ernst Moritz Hahn
• Akin Günay
• Lijun Zhang 0001
• Yang Liu 0003

ECAI Conference 2014 Conference Paper

GOSU: computing GOal SUpport with commitments in multiagent systems

• Özgür Kafali
• Akin Günay
• Pinar Yolum

Goal-based agent architectures have been one of the most effective architectures for designing agents. In such architectures, the state of the agent as well as its goal set are represented explicitly. The agent then uses its set of actions to reach the goals in its goal set. However, in multiagent systems, most of the time, an agent cannot reach a goal only using its own actions but needs other agents to act as well. Commitments have been successfully used to regulate those interactions between agents. This paper proposes a framework and an environment for agents to manage the relations between their commitments and goals. More specifically, we provide an algorithm called GOSU to compute if a given set of commitments can be used to achieve a particular goal. We describe how GOSU can be implemented using the Reactive Event Calculus and demonstrate its capabilities over a case study.

AAMAS Conference 2013 Conference Paper

Generating and Ranking Commitment Protocols

• Akin Günay
• Michael Winikoff
• Pinar Yolum

Interaction is a fundamental part of multiagent systems, and is usually regulated by protocols. Typically, these protocols are defined at design-time. However, there exist situations where it is desirable to generate protocols at runtime. We develop an algorithm to generate commitment protocols considering the agent’s goals, its capabilities and its domain knowledge about the other agents’ goals and services. We then provide a method to rank these protocols in terms of their risk and benefit in order to select the one that best suits the agent’s current state.